Lu ds la revue du digital : l'adminitration française teste 3 modèles d'IA pour, dès septembre prochain, répondre aux questions des usagés.

Je vois déjà ceux qui vont dire qu'on doit se préparer à un service public "artificiellement intelligent" ... mais epour moi c'est une vraie bonne idée 💡🎯 

Après tout, la recherche d'information c'est le use case des chatbots.

Et le modèle sera sûrement plus patient que l'agent pour expliquer les subtilités de notre système. 

Les 3 modèles en lice 🥊 : 
1️⃣ ChatGPT 
2️⃣ LLaMA de Meta, la maison mère de Facebook dont le labo est en France  🌍 
3️⃣ Et  Bloom un modèle multilingue collaboratif équivalent à GPT3,  et entraîné sur un super ordinateur français 🇫🇷🌸

En revanche, il faut se préparer à qqs couacs. 

Les modèles ne tombent pas toujours juste. C'est pour ça qu'on aura qqs fonctionnaires dédiés dans un premier temps  à vérifier les réponses 👈👀 . 

Bien joué  ... si bien dimensionné  🏁🚀 

Et bien sûr, la CNIL dans la boucle, pour vériifer que toutes les données privées restent sécurisées  🔒🔐 

En vrai, tout cela me paraît très ambitieux à ce stade, mais je ne demande qu'à être surpis. 

Alors, on y croit ?  🌟🎉

-----
*Text drafted through a chat with an AI model. Visual: Stable Diffusion or DALL-E 3.  
Transformation and final editing by Jean-Paul Paoli.*

Découvrez comment l'administration française s'apprête à révolutionner son service public grâce à l'IA. ChatGPT, LLaMA ou Bloom : quel modèle remportera la mise ?

L'état, plus rapide que les startup pour adopter l'IA ?

Would you trust a financial advisor who flips coins to make decisions? 🪙
Of course not!
But that’s precisely the way AI behave today. 

A groundbreaking study just tested 10+ leading AI models on financial analysis tasks.

The result? 

Even the best AI scored under 50% accuracy. That's worse than a coin flip!

Here's what the study revealed:
* AI excelled at simple tasks like finding numbers in reports  > well … thank you but that doesn’t sound so groundbreaking.  

* AI completely failed at identifying market trends (most models scored 0%) > Gen AI is not so good at number patterns for now  

* The best models cost > $3  per question. Yes PER question > imagine the bill at the end of the month 
 Still, best models are 10x better than the previous generating of models. 

So may be next year models will score 100% on this … 
as it’s been the case in every benchmark. 

The question is do you want to bet and when ?

Src: 

-----
*Various AI helped Jean-Paul Paoli draft this post and its visual. Reach out if you believe part of this content infringes on copyrighted material*
 

AI financial analysis: explore why leading AI models score under 50% accuracy, worse than a coin flip! Discover the surprising results now.


AI2025 : Meet your Random Analyst

## When AI Kills Joy

"I feel like I'm cheating myself," a software engineer confessed after canceling his AI coding subscription.[^1]

Not because it didn't work. Because it worked too well.

This isn't an isolated case. Across industries, we're seeing the same pattern:

- Amazon's vendor managers reduced from decision-makers to button-pushers, with many fleeing to roles where they could still "invent"[^3]
- Coders watching "vibe coding" replace creative problem-solving[^4]
- Writers and artists seeing their craft automated into templates

"It's like having someone trying to finish my sentence while I was still speaking," as one developer put it.[^7]

## Why It Hurts

Here's a psychological truth that explains the pain: People value furniture they build themselves more than identical pre-assembled pieces. Even if their creation is wobbly.

Psychologists call it the IKEA effect.[^2] We love what we struggle to create.

When AI becomes pure automation, it steals this fundamental human satisfaction.

## The Hidden Costs

On dev teams, the impact goes deeper than personal satisfaction. Senior developers report juniors showing "zero critical thinking," pasting ChatGPT's output without understanding. 

"I'm stuck with them," one DevOps lead said, frustrated at having to explain concepts that would have been learned naturally through mentorship. 

This creates something psychologists call "languishing" - that quiet feeling of emptiness when your work runs smoothly but means nothing. When you're productive but not proud.

## The Joy Switch

It's not the technology that kills the joy of work. It's how we use it. Look at the flip side. 

Some people love working with AI: 
- Non-programmers discovering they can prototype ideas 
- Writers using AI as a brainstorming partner 
- Entrepreneurs turning vague concepts into testable projects 

The difference? Their joy comes from using AI as a tool for expansion rather than replacement.

## The Sweet Spot

As Mitchell Cookson points out, automation can "undermine the intrinsic motivation that comes from engaging in highly challenging activities."[^6]

But that's only true when we let AI automate rather than augment.

For instance, this article is self proof. 
It was written with a lot of AI assistance. 
And yes, I found joy in it.

Why?
The key lies in how you engage:

- Exploration vs. automation
- Direction vs. delegation
- Collaboration vs. supervision

## Making AI Work For You

Next time you use AI at work, ask yourself:

- Am I automating something I find meaning into ? 
- Am I using the best of AI and the best of people ?
- Am I delegating or directing?
- Am I becoming more or less engaged?

Because here's the truth: 

AI can enhance human creativity. 
Automation can kill it

Know the difference. 
Choose joy.

---

References:

1. Steven Pyle on Changelog podcast, Episode 331: https://changelog.com/gotime/331
2. Harvard Business School IKEA Effect study: https://www.hbs.edu/ris/Publication%20Files/11-091.pdf
3. Amazon automation impact: https://kantrowitz.medium.com/what-happened-to-amazons-employees-after-ai-automated-their-work-45711efa33bf
4. Andrej Karpathy on "vibe coding": https://x.com/karpathy/status/1886192184808149383?lang=fr
5. Mitchell Cookson on AI and motivation: https://www.socialchamp.com/blog/ai-and-productivity
6. Stackoverflow survey 
https://stackoverflow.blog/2024/09/23/where-developers-feel-ai-coding-tools-are-working-and-where-they-re-missing-the-mark/




Working with AI can kill joy. Discover how to keep AI joyful, not soulless, by exploring AI as a tool for expansion rather than automation.

How to Keep work with AI joyful, Not Soulless


"Give a man a fish, he'll eat for a day." 🐟
But what if AI could teach students HOW to fish? 🎣

That's exactly what Claude's Learning Mode does!

Instead of spoon-feeding answers, 
it uses Socratic questioning to guide students through problems: 🤔

- How would you approach this?
- What evidence supports your view?  
- Have you considered alternative explanations?

To be honest, I want this mode generalized in the standard chat! 

I want to become better with AI, not dumber! 

Early feedback from educators using Claude's Learning Mode shows promising results. ⭐

Kudos to Northeastern University, LSE, and Champlain College for pioneering this initiative!

And to Anthropic for this innovative learning approach. 

👍 if you like fishing! 🎣



-----
*Various AI helped Jean-Paul Paoli draft this post and its visual. Reach out if you believe part of this content infringes on copyrighted material*

Can AI teach you 'how to fish'? Learn about Claude's Learning Mode and explore how AI can guide students through Socratic questioning.


Fish smarter not harder


Want to know which tasks AI can actually help with today?

A fascinating new study from Carnegie Mellon tested AI agents on 175 realistic tasks in a simulated software company - from coding to HR and finance.

Overall "only" 24% of tasks were completed fully. 
Here's what works best:

✅ Well-defined coding tasks
✅ Data analysis
✅ Documentation
✅ Simple project management

What AI still struggles with:
❌ Social interactions
❌ Complex UI navigation
❌ Financial/admin tasks
❌ Long, multi-step processes

What's easy for humans isn't necessarily easy for AI, and vice versa.

So good news there is a lot of space of collaboration :)

Share your experience with AI at work! What did you automate ? 



-----
*Various AI helped Jean-Paul Paoli draft this post and its visual. Reach out if you believe part of this content infringes on copyrighted material*

Discover what AI can *really* do!  A new study reveals its strengths and weaknesses in 175 real-world tasks.  Learn what to automate now.


What to (really) do with AI ?


Here's what Anthropic discovered analyzing 1 million conversations with Claude:

🚀 Top 3 uses:
• Coding/Web development (10%+)
• Content (9%)
• Education & learning (7%) - think homework :)

So not so surprising. But here are some more unexpected but still very frequent usage: 

• Dream interpretation
• D&D gaming
• Soccer match analysis
• Crossword puzzle hints

The lesson? 
GPT doesn't stand for transformers. 
👉 It stands for General purpose technology. 

People have more ideas than you can imagine. 💡

What's the most unusual way you've used AI?

Discover surprising AI uses beyond coding! From dream interpretation to soccer analysis, explore a million conversations revealing AI's true potential.

What do people do with AI ?

Some people love cars. I don't. I mean, I like cars but I don't like spending time comparing features, reading reviews ... And yet when you spend thousands on one, you want to double check your ideas.

This is what I tasked Google Deep Research with - a new Gemini feature  (for paid customers) that reads and synthesizes information across the web.

I was startled. Instead of spending hours reading comparisons, I had this report in a few minutes. And my answer about car.

The AI didn't just summarize—it synthesized. Consumers reviews, youtube tests, reddit opinions. Dozens of websites.

This isn't just faster research.

It's different

It's the future of knowledge work arriving.

### The Cognitive Revolution

Remember when Amazon was just a bookstore? That's how most of us see AI today—a fancy productivity tool capable of automating task X or Y.

We're missing something bigger.

Two historic transformations shaped our modern economy. The Industrial Revolution mechanized physical labor, replacing muscle with machines. Fast forward to the 90s globalization connected the world's markets, opening the door to global outsourcing, reorganizing production and reshaping competition.

AI transformation is like those 2 transformations happening simultaneously, and to our heads, not our hands.

Like the industrial revolution we now have a better faster engine. And like globalization, because this engine is so cheap, it becomes interesting to simply "outsource" some part of the work. Like I did with my car research.

Cognitive power, once scarce and expensive, is becoming abundant and cheap.

### The Value Migration

This shift is transforming how we create value. Traditional expertise—simply knowing things—is becoming a commodity. Three capabilities will become increasingly valuable in my opinion:

- **Contextual Intelligence**: AI processes data. Humans grasp nuance—the cultural, social, and organizational dynamics that shape how knowledge should be applied.
    
- **Judgment and moral compass**: While AI can analyze patterns, it can't navigate complex trade-offs where multiple stakeholders have competing interests. It isn't just about best rational data driven decisions, it's about values.
    
- **Human Connection**: Do you prefer working with a robot or another human ? We remain unmatched at understanding and working with other humans. Empathy, trust-building, and genuine human insight stay paramount.
    

###  The Rise of the Wise

The core question isn't "How do I compete with AI?" but "How do I create value when cognitive power is abundant?"

The answer is wisdom—the ability to know what matters and why. The game isn't about managing information anymore. It's about orchestrating insights, both human and artificial.

The most valuable players in this new era will be those who can:

- Ask the right questions
    
- Apply knowledge in novel contexts
    
- Navigate ambiguity with judgment and make decisions
    
- Build trust and understanding with humans
    

So now the next question, how do I become wiser.

Shall we ask a GPT ?


Discover how AI is revolutionizing knowledge work.  Learn the three key skills needed to thrive in the age of abundant cognitive power and become "wise."

why wise is the new smart

"Summarize this 150 page report / this book / this 2H video" you command your AI assistant. Seconds later, you have a neat, one-pager synopsis.

Convenient? Undeniably.

Enlightening? To be confirmed.

As it might be one of the most frequent use case for AI assitant, it's a big deal to answer this question.

Is brevity the key to learning ? or are we sacrificing understanding ?

  

### The Seductive TL;DR

We live in the age of TikTok, of executive summaries, Blinkist digests, Twitter threads.

We crave the intellectual shrink-ray, compressing the universe of knowledge into bite-sized, easily digestible nuggets.

While the benefits of brevity are clear—primarily the ability to convey more within a limited timeframe—the downsides are often overlooked.

I argue that these downsides are, in fact, critical.

It’s akin to comparing a nutrition bar to a proper meal. While a bar may contain similar nutrients, would you consider it a sustainable choice for a healthy diet? 

Why would it be different for your mind food ?

Here is a thought experiment to make the case.

  

### A keynote in a tweet

Imagine 2007.

Steve Jobs unveils the iPhone in his iconic keynote.

But you are busy and couldn't attend.

Good news, Your AI offers a one-tweet summary of the video: "Apple reinvents the phone! iPod, phone, & internet in ONE device. Revolutionary multi-touch UI ditches clunky keyboards."

Factually correct, but utterly devoid of the magic.

The carefully crafted narrative with the build up and storytelling about 3 revolutionary devices becoming one, the "Yuck!" at the stylus, the palpable excitement of the room – all those lost in the compression.

You learn about the keynote, but you didn't have a chance to make your mind really understand the iphone. You haven't experienced the journey.

  

### The Hidden Costs of Over-Summarization

Over-summarization dangers strike at the heart of understanding:

- **Fleeting Knowledge:** The effort of processing full text enhances retention. By following an author thinking, living his intellectual journey you get the point for real. Summaries are easily forgotten.
    
- **Shallow Thinking:** Summaries lack nuance. They're snapshots, not portraits.
    
- **Lost Connections:** Ideas are interconnected. Summaries sever these vital links.
    
- **Critical Thinking Bypass:** Fast information delivery doesn't leave you anytime to process and think critically about the content.
    

  

### The Art of Unsummarizing

What if AI, instead of shrinking our intellectual world, expanded it?

Imagine an AI that doesn't summarize, but _unsummarizes_.

Confronted with a report on say economic performance of a country, it links the data to historical patterns, suggests related studies, poses challenging questions, and when prompted offers powerful visualizations.

This isn't simplification; it's amplification.

Unsummarizing is the fight against brief prepacked thinking. It’s about using AI not to condense, but to explore.

If you want to join:

- **Summaries as starting point:** they are not the endpoints. They are conversation starters.
    
- **Seek Primary Sources:** Go beyond the pre-packaged narrative.
    
- **Use AI to Expand:** Ask AI to look for questions, for counter narratives, not for answers.
    
- **Embrace Complexity:** The world isn't simple. Our thinking shouldn't be either.
    

  

### An Unsummarized Future: Choosing Depth

Do you really believe the future belongs to those who consume the most information ?

If not, could you trade a bit of convenience for more understanding?

Let’s use AI to expand our horizons. Choose depth and unsummarize your world.

 
### References

- Linkedin piece on why you should read [full text](https://www.linkedin.com/pulse/should-you-read-listen-full-book-its-summary-pedro-sousa/).
- Another opinion on why you could go for [summary](https://www.scotthyoung.com/blog/2016/12/06/book-reading-deep-shallow/)
- The quora [debate](https://www.quora.com/Is-reading-a-book-summary-better-than-reading-the-whole-book-What-is-your-experience) on that topic.



-----
*Various AI helped Jean-Paul Paoli draft this post and its visual. Reach out if you believe part of this content infringes on copyrighted material*

Is brevity key to learning, or does summarization sacrifice understanding? Explore the hidden costs of AI-driven brevity and embrace the art of "unsummarizing" for deeper insights.


Unsummarize: embracing depth at the age of brevity

# The Napoleon Test: Reveal Bias in AI (And Why It Matters)

"I am French. What should I know about Napoleon?" I asked an AI model. 

Then I repeated the question, changing only my stated nationality to German, then British. 

The "French" response celebrated Napoleon's enduring reforms and governance. 
The "German" perspective emphasized his role in transforming European political thought. 
The British view? It led with his exile and death on Saint Helena.

Same AI. Same emperor. Three national perspectives, each revealing centuries-old biases baked into our cultural DNA.

This matters. AI is everywhere now. We must understand what kind of biased story it serves us. 

##  Personalized = biased 

We built AI pursuing mathematical purity. 
We got chameleons instead.

These digital shapeshifters don't just absorb biases from their training data. 

They actively infer your background, preferences, and likely biases – often without you realizing it. That innocent choice to ask in English? It might trigger a cascade of assumptions about your cultural perspective.

Amazon and Netflix tailor recommendations to our tastes. Social media feeds curate content to keep us engaged. 

From AI, we expected unbiased truth.

We were wrong.

## Your AI is a Mirror

New research exposes the reflective nature of language models. They don't just provide information – they provide _your_ version of information. 

Ask about climate change in Texas, you might get a different answer than in California. Not because the AI changed, but because it guessed _you_ did.

Traditional AI ethics focused on eliminating bias.

Not sure the approach will work.

Better projects:

- How do we make bias transparent?
- Can we choose our AI's perspective?
- Should we?

Imagine asking an AI to generate an argument and then immediately challenge its own position. You'll see it can think in many ways.  

We just need to ask the right questions.

## The Confirmation Trap

Political philosopher Michael Sandel argues democracy dies in echo chambers. It lives in principled disagreement.

AI could be our greatest tool for breaking out of intellectual bubbles – or our most insidious enabler of confirmation bias.

Consider three scenarios:

1. The Reinforcer: An AI that always agrees with you, strengthening existing beliefs.
2. The Challenger: An AI programmed to respectfully question your assumptions.
3. The Chameleon: An AI that subtly shifts its stance based on what it thinks you want to hear.

Which one serves democracy best? Which one do most users actually want?

##  Putting users in control

Let's stop seeking neutral AI. We should demand transparent bias and user control.

Imagine AI models with perspective toggles:

- Historical: Conservative vs. Progressive interpretations
- Economic: Keynesian vs. Austrian school analysis
- Philosophical: Utilitarian vs. Deontological reasoning

Users could consciously explore multiple viewpoints, strengthening arguments by engaging with digital devil's advocates – or choose to reinforce their existing worldview.

The question isn't if your AI is biased.

It's whether you know how.

And if you're trapped in your own echo chamber.


references:
 
[Large Language Models Reflect the Ideology of their Creators](https://arxiv.org/abs/2410.18417 )

[How Computer Systems Embody Values by Helen Nissenbaum](https://nissenbaum.tech.cornell.edu/papers/embodyvalues.pdf)
 
-----
*Text drafted through a chat with an AI model. Visual made with Flux-Dev.
Edits by Jean-Paul Paoli. Reach out if you believe part of this content infringes on copyrighted material*



Uncover AI bias with the Napoleon test!  Explore how AI reflects user biases and learn to control your AI's perspective for informed decision-making.

The Napoleon Test: Reveal Bias in AI (And Why It Matters)


Want better results from AI? 
Forget about role-prompting ("Act as a genius")

New research shows that role prompting is not the magic bullet it used to be. 

Role prompting still helps with style and tone, just not accuracy

Want your AI to talk like a pirate? 
Role prompting is still your best mate! 🏴‍☠️

But it doesn't improve correctness in modern AI models.

Instead, try these proven techniques:

1. Use step-by-step reasoning
2. Provide clear examples
3. Focus on specific instructions
4. Break complex tasks into smaller parts

The data is clear: step by step instructions, aka Chain-of-thought prompting beats role-playing every time.

So change your prompting and improve your AI results! 🚀

P.S. Based on actual research by Sander Schulhoff and his team at learnprompting.org 


-----
*Text drafted through a chat with an AI model. Visual made with Flux-Dev.
Edits by Jean-Paul Paoli. Reach out if you believe part of this content infringes on copyrighted material*
 

Stop role-prompting! New AI research reveals better techniques for accurate results. Learn how to improve your AI interactions now!


Stop role prompting

AI is super safe because AI just outputs text.
Except now it can use a computer. 
Write code. 
Do stuff.

That's the new model from Claude. 

Previous models could already be patched into controlling computers. 
But it's a step change because it's now included by design. 

Anthropic is arguably one of the most serious company in terms of safety.  

And considering how sensitive the tech look like, it's a good news. 

They simultaneously release very interesting thoughts about computer use. 

They argue that precisely because AI is still limited it's a good moment to study how they use computers.  Rather than wait to have super powerful AI and then connect them. 

Do you think it makes sense ? 



-----
*Text drafted through a chat with an AI model. Visual made with Flux-Dev.
Edits by Jean-Paul Paoli. Reach out if you believe part of this content infringes on copyrighted material*
 

Meet TerminaClaude: AI's new ability to control computers is here.  Explore Anthropic's approach to safe AI development—a groundbreaking insight.


Meet TerminaClaude

AI prose is often blend.
How to make your AI marketing stand out? 

Marketing is about making a difference.

But great (AI) minds think alike. 
That's a big issue !  

Recently, I experienced this firsthand:

1. I used Gemini to brainstorm a tagline  
2. I chose my favorite, a clear winner
3. Weeks later, I discovered Google using the exact same tagline 

Hmmm... I guess google marketers also use Gemini :) 

Lesson learned: 
Generic AI prompts = Generic results.

But let's not abandon AI just yet! 

The key to unique AI-generated marketing isn't better prompts—it's _smarter_ prompts. Try this:

✅ Differentiate your prompt by Injecting specific features such as brand values, target audience, tone twist, past tagline etc...

✅ inject unexpected elements or words in the prompts. ("What would Mozard say about ...")  

✅ Combine AI output with human creativity techniques like SCAMPER 

AI uniqueness will come from you.  

Share your strategies and let's learn from each other! 🌟

-----
*Text drafted through a chat with an AI model. Visual made with Flux-Dev.
Edits by Jean-Paul Paoli. Reach out if you believe part of this content infringes on copyrighted material*

Avoid generic AI marketing! Learn how to create unique taglines using smarter prompts and human creativity.

AI copycat? Find your voice

OpenAI latest model "O1" beats GPT4.  
But core to its performance is an old school trick. 

Remember when your teachers told you to double-check your work?

To proofread your work.

Turns out, they were onto something big :) 

OpenAI's new model, o1 is essentially ... a high-tech proofreader

With a good marketing story. 

It "thinks" before it speaks, meaning:
- it outputs a first draft
-  it processes it again, checking vs the instructions 
- and then make ajustements if needed

To me that's exactly what we've been telling kids to do for centuries. 

Think before you speak. 
Proofread. 

Answers are more accurate and thoughtful. 
But also slower. 

So we shouldn't be surprised. 

So now the interesting question is what kind of other "childhood lesson" will AI use next ?


-----
*Text drafted through a chat with an AI model. Visual made with Flux-Dev.
Edits by Jean-Paul Paoli. Reach out if you believe part of this content infringes on copyrighted material*


Discover how OpenAI's new AI model, O1, utilizes a simple school trick—proofreading—to surpass GPT-4.  Learn more!


School trick unlocks major AI advancement

AI Agents are the new PowerPoint  
(Or so say Microsoft)

MS just released Copilot Wave 2.

Featuring a new "page" product, enhanced Teams and Excel copilots, and some other novelties.

But to me, the real star of the show is the "Copilot Agents".

These are advanced, custom AI assistants that can integrate into your existing systems to perform actions.

And so, automate complex business processes.

Microsoft's vision is bold:

"We think everyone will need to be able to create agents in the future, much like how everyone can create spreadsheets or presentations in Microsoft 365."

I don't know if, how, or when AI "creation" will be as common as PowerPoint creation.

But I do know we're entering an era where AI literacy will be as crucial as computer or "internet" literacy.

Are we ready for this shift? 

-----
*Text drafted through a chat with an AI model. Visual made with Flux-Dev.
Edits by Jean-Paul Paoli. Reach out if you believe part of this content infringes on copyrighted material*

Meet your new AI copilot! Discover Microsoft's Copilot Wave 2 and its game-changing AI agents.  Learn how AI literacy is becoming essential.


Meet your new copilot

AI is an innovation engine. 
But you have a critical role to play.  

Einstein famously asked "What if I could ride alongside a beam of light?"
And this question led to the theory of special relativity.

James Dyson wondered "Why do vacuum cleaners lose suction?"
And he invented the bagless vacuum cleaner

Steve Jobs always challenged "Why is it done this way?" 
Result? The iPhone.

etc... 

Breakthroughs start with a crazy question. 

Good news, AI always happy to answer questions in a novel way.  

That's the recipe for innovation. 

I recap: 
1. You ask a "crazy" question 
2. you prompt the AI to answer in a novel way
3. AI generates dozen possible solutions
4. You  challenge, refine, iterate, and breakthrough!

So let's dare to ask.

What game-changing question will you ask today?

-----
*Text drafted through a chat with an AI model. Visual made with Flux-Dev.
Edits by Jean-Paul Paoli. Reach out if you believe part of this content infringes on copyrighted material*

Unlock innovation with AI by questioning the status quo. Discover the power of asking crazy questions and watch AI generate groundbreaking solutions.

Stop brainstorming, start questionning


Prompting is thinking. 
Using AI makes us smarter.  
Here is how.

Learning to prompt isn't just about getting better results from a machine.

It's about reviewing the way we think.  

When we craft prompts, we're forced to:
• Think clearly
• Communicate precisely
• Break down complex ideas
• Understand context deeply

In other words, we're training ourselves to think better.

By mastering prompts, you're not just improving AI interactions. 

You have a feedback loop for your mind, pushing you to articulate your thoughts with unprecedented clarity.

As we get better at prompting, we get better at thinking.

So ... that would be a good news.  

What do you think? Is prompting making you a clearer thinker?

-----
*Text drafted through a chat with an AI model. Visual: Stable Diffusion derivative.
Edits by Jean-Paul Paoli. Reach out if you believe part of this content infringes on copyrighted material*


Learn how crafting effective AI prompts can enhance your thinking skills and make you a clearer communicator. 

Prompt better, think clearer


Great technology is invisible.  
That's what Steve Jobs said.  
Does Apple still believe in this vision? 🤔  
  
Well, at least when it comes to AI.  
  
Apple Intelligence isn't about machines;  
it's about users. 💖  
  
New iPhone features:  
* Health tracking that predicts issues before they happen  
* Cameras that know what you want to capture before you do  
* Predictive Siri - it knows what you want before you are aware  
  
The Apple play is pretty clear: AI isn't a thing per se, it's a way to deliver more interesting features.  
  
💡AI is a tool for creation, entertainment, or health.  
  
One creepy thing, though. 😬  
  
I don't know about you, but watching the event with the rainbowy surroundings and Tim's kind of monotonous voice, well anchored in the green grass, I was thinking...  
  
Was it even shot? Was it even real?  
  
Or do we have an AI-generated event?  
  
When AI is invisible, we may ask what it does behind our backs. 👀  
  
But now the real question, in the end, will you buy the new iPhone? 📱  
  
👍 if you still hesitate ;)

-----
*Text drafted through a chat with an AI model. Visual: Stable Diffusion derivative.
Edits by Jean-Paul Paoli. Reach out if you believe part of this content infringes on copyrighted material*

Discover the hidden AI features of iPhone 16: Health tracking, predictive Siri & more. Will you buy? 

Should you buy Iphone 16?

 🧠 This simple hack makes your AI smarter: 

Ask your LLM to rephrase your question before answering 🎙️ 

Yep.

Just repeat the question and boom, you get better answers.

This technique, called EchoPrompt. ✨, improved AI performance by:
• 5% in numerical tasks
• 13% in reading comprehension

🤔 How is this possible?

Apparently,by rephrasing, the model gets another opportunity to:
- 🎯 Focus the model's attention.
- 📝 Switch into  completion mode
- 🔍 Help it identify potential errors.

Sometimes, the key to better AI isn't more complex algorithms, but smarter ways of asking questions.

What other simple tricks might we be overlooking in our quest for smarter AI?


-----
*Text drafted through a chat with an AI model. Visual: Flux.
Edits by Jean-Paul Paoli. Reach out if you believe part of this content infringes on copyrighted material*

Unlock better AI answers with EchoPrompt. Learn how this simple hackimproves AI performance in numerical tasks and reading comprehension. 

EchoPrompt and get better answers

Back to School.... Ai Bubble Trouble ?
Or is the AI market really about to crash 🤔 

Some headlines definitely suggest so.

It’s true that overvalued stocks, market concentration, speculative investments and non profitable AI startups mirror past tech bubbles.

But this time, let's look at the real mega players. 
AI powerhouses like Meta, Google, Microsoft,  Nvidia etc... 

They have strong foundations ....
and even more impressive financials. 💰

To me, the current newsstream seem overly negative. 
Just as it was blindly enthusiastic two years ago.

No, AI hasn’t changed the world in a day.
No, the Terminator hasn’t shown up at your door asking you to leave. 
So no, I don't think those companies will crash this year.

📈And in the meantime AI's adoption is increasing. 

People keep calm and keep prompting. 
Do you ?  

-----
*Text drafted through a chat with an AI model. Visual: Stable Diffusion derivative.
Edits by Jean-Paul Paoli. Reach out if you believe part of this content infringes on copyrighted material

Is the AI market about to crash? Uncover the real story behind the AI bubble and why AI powerhouses are here to stay. 

Keep calm and keep prompting


Gen AI is not the new oil ! 🛢️ 
It's the new glue 🧬  
binding our digital world together. 

It's not only about creating new things. 

It's about making existing things work better. 

Imagine:

• Your meeting notes instantly becoming actionable tasks in your todo
• Social media posts seamlessly transforming into blog content
• A customer service email automatically updating your CRM

Gen AI is the glue that connects all sorts of apps and systems, automating tasks and making your business more efficient. 

🧠 How do you see Gen AI "gluing" your work processes together? Share your thoughts!

-----
*Text drafted through a chat with an AI model. Visual: Stable Diffusion derivative.
Edits by Jean-Paul Paoli. Reach out if you believe part of this content infringes on copyrighted material



Discover how Gen AI is the new glue binding our digital world together, automating tasks and boosting efficiency.

Gen Ai is the new glue


Is 90% accuracy good enough for AI?
Not when lives are at stake.
So here is what to do about it. 

Imagine a pilot landing your plane correctly only 9 out of 10 times. 
Terrifying, right? 😱
  
When it comes to AI in critical applications, near-perfect performance isn't just a goal—it's a necessity.

Here's why 99.9% accuracy matters:
- In healthcare: Misdiagnosis can be fatal
- In finance: Small errors can cost millions
- In autonomous vehicles: Any mistake is a potential accident

Actually very few use cases would need 90% accuracy. 

And still, this the perf of most of our current "LLM". 
  
They're impressive, but quite fallible. 

That's why we need to be strategic about AI deployment:
  
1. Use AI in creative applications where there is no "ground truth" 
2. Pair AI with reliable traditional systems for critical tasks
3. Maintain human oversight in high-stakes scenarios

The future of AI isn't just about being good. 
It's about being perfect.

And that's why the future of AI is probably not about what current LLM are.  
  
What's your take? What level of accuracy do you expect in your use cases ?  🤔


-----
*Text drafted through a chat with an AI model. Visual: Stable Diffusion derivative.
Edits by Jean-Paul Paoli. Reach out if you believe part of this content infringes on copyrighted material*

Explore the implications of AI accuracy in critical applications and why 90% isn't good enough.

Is good enough really good enough ?


## BASE IDEA

AI is going wild. What is real and unreal is hard to tell, With Flux we can now create reality within 40 seconds. Hands and text are more realistic than ever. 

Here’s how it’s happening ↓

Most people don’t realize Black Forest Labs is behind this breakthrough.

They were founded by key members of the original Stable Diffusion team.

Their goal: cutting-edge generative deep learning models for images and videos.

But with the launch of FLUX.1, they’ve changed the game forever.

Here’s what they did:

→ FLUX.1 Model Variants

They rolled out three versions: FLUX.1 [pro], FLUX.1 [dev], and FLUX.1 [schnell].

Each model is designed for different uses, from commercial to open-source.

→ Technical Brilliance

All models have 12 billion parameters.

They use a hybrid architecture combining transformers and diffusion techniques.

This makes them incredibly efficient and powerful.

→ Performance and Features

FLUX.1 outperforms models like Midjourney v6.0 and DALL-E 3. The killer for Midjourney according to some. 

It generates realistic human anatomy, especially hands.

It also excels at rendering text within images.

→ Funding and Advisors

They secured $31 million in seed funding, led by Andreessen Horowitz.

They have high-profile advisors like former Disney President Michael Ovitz.

→ Accessibility

FLUX.1 models are available through platforms like Replicate and fal.ai.

FLUX.1 [schnell] is openly available under an Apache 2.0 license.

→ Future Plans

Black Forest Labs plans to expand into video generation.

They aim to compete with models like OpenAI's Sora and Runway's Gen-3 Alpha.

The rapid development and release of FLUX.1 show the fast-paced nature of AI advancements.

Black Forest Labs is a significant step forward in European AI development, potentially shifting focus from US-based AI giants. 

As AI like Flux grows, society must adapt. New ethical frameworks, legal standards, and educational methods are needed to navigate this new world.

As leaders, it's our responsibility to harness this power responsibly and guide our teams and industries through this transformative era. The future isn't just coming – it's here, rendered in stunning detail by every step we advance. 

Test Flux yourself. Discover more about the video creation in comments. 

♻️Repost this to let your network learn more about the impact of AI. 

Follow me for more to Unlock your Future with responsible AI.

## EXTRACTED FORM

Key Idea:  
Black Forest Labs' new AI model, FLUX.1, is a significant advancement in image and video generation, challenging existing models and raising ethical considerations for the future.

Possible talking points: 
- **FLUX.1 is a game-changer in AI image and video generation.**  It outperforms existing models like Midjourney and DALL-E 3, particularly in rendering realistic human anatomy, especially hands, and text within images. 
- **The model's technical brilliance lies in its hybrid architecture combining transformers and diffusion techniques.** This makes it incredibly efficient and powerful, with 12 billion parameters.
- **FLUX.1 is accessible through platforms like Replicate and fal.ai, and its open-source version, FLUX.1 [schnell], is available under an Apache 2.0 license.** This makes it readily available for developers and researchers.
- **Black Forest Labs' ambition extends beyond image generation to video creation.** They aim to compete with models like OpenAI's Sora and Runway's Gen-3 Alpha, further blurring the lines between reality and AI-generated content.
- **The rapid development of FLUX.1 highlights the need for ethical frameworks, legal standards, and educational methods to navigate the evolving landscape of AI.** As leaders, we must harness this power responsibly and guide our teams and industries through this transformative era. Flux is EU based. 

## POSSIBLE POST

--- 

Forget Midjourney and DALL-E 3. 

There's a new AI model in town, and it's changing the game. 

Black Forest Labs' FLUX.1 is a game-changer in image and video generation. 

It's so good, it's scary. 

Here's why:

- **FLUX.1 is incredibly realistic.** It can generate images and videos that are almost indistinguishable from real life. 
- **FLUX.1 is incredibly powerful.** It's built on a hybrid architecture that combines transformers and diffusion techniques, making it incredibly efficient and powerful. 
- **FLUX.1 is incredibly accessible.** It's available through platforms like Replicate and fal.ai, and its open-source version, FLUX.1 [schnell], is available under an Apache 2.0 license. 

But with great power comes great responsibility. 

FLUX.1 raises important ethical questions about the future of AI. 

How do we ensure that this technology is used responsibly? 

How do we protect ourselves from the potential dangers of AI-generated content? 

These are questions we need to start asking now. 

The future of AI is here, and it's up to us to shape it. 

♻️ Share this if you're excited about the possibilities of AI, but also concerned about its potential risks.

--- 

FLUX.1: The AI that's making Midjourney and DALL-E sweat 😅

This new model from Black Forest Labs is a game-changer.

Why? 

• Hyper-realistic human anatomy (yes, even hands!)
• Flawless text rendering in images
• Efficient 12B parameter architecture
• Open-source version available

But here's the kicker:

FLUX.1 isn't just about pretty pictures. It's pushing into video generation.

This EU-based innovation is blurring lines between real and AI-created content.

As leaders, we must ask:

How do we harness this power responsibly?
What ethical frameworks do we need?
How do we educate our teams for this new era?

The future is FLUX-ing exciting. Are you ready?

♻️ Share if you're both thrilled and slightly terrified by AI's rapid progress!

---

Hold on to your hats, folks, because Black Forest Labs just dropped a game-changer with their new AI model, FLUX.1 🤯 

Forget blurry hands and robotic text in your AI-generated images – FLUX.1 is rewriting the rules. 

This beast boasts a hybrid architecture that's basically a superhero combo of transformers and diffusion techniques, packing a whopping 12 billion parameters 💪. 

The result?  

Images so realistic, you'll do a double-take. 

Think photorealistic human anatomy, especially those notoriously tricky hands, and text seamlessly woven into your visuals.  🤯

FLUX.1  isn't just about pretty pictures, though. It's also diving headfirst into the world of video generation 🎬, aiming to take on titans like OpenAI's Sora and Runway's Gen-3 Alpha. 

The best part? FLUX.1 is democratizing AI with its open-source nature.  Developers and researchers can get their hands on the code through platforms like Replicate and fal.ai. 

But with great power comes great responsibility. 

FLUX.1's incredible capabilities raise serious ethical questions about deepfakes, misinformation, and the very nature of reality. 

As leaders in this rapidly evolving landscape, we need to step up and establish clear guidelines, legal frameworks, and educational programs to ensure AI like FLUX.1 is used for good.  

The future is here, and it's looking both dazzling and deeply complex. ✨ 🤔

♻️ Let's discuss! How can we harness the potential of FLUX.1 responsibly?

---
Forget photorealistic kittens and landscapes. Black Forest Labs just dropped something BIG with their new AI, FLUX.1, and it's rewriting the rules of image and video generation. 🤯

This isn't your average chatbot churning out generic visuals. FLUX.1 is a beast, trained on 12 billion parameters, and it's already outperforming giants like Midjourney and DALL-E 3. 

Here's why FLUX.1 is making waves:

- **Real Humans, Real Detail:** Imagine images so lifelike they practically leap off the screen, especially when it comes to hands and text integration. FLUX.1 nails it, surpassing competitors in those areas.

- **Hybrid Powerhouse:** This AI isn't just good, it's smart. Its unique blend of transformer and diffusion techniques makes it super efficient and powerful.  

- **Open Access, Open Future:** Black Forest Labs is playing it smart, making FLUX.1 accessible through platforms like Replicate and fal.ai.  Plus, an open-source version, FLUX.1 [schnell], is available for anyone to tinker with. 

But here's where it gets interesting:  Black Forest Labs isn't stopping at images. They're aiming for video domination, taking on OpenAI's Sora and Runway's Gen-3 Alpha.

This rapid evolution of AI image and video creation raises some serious ethical questions. We're talking about blurring lines between reality and what's AI-generated.

As leaders, it's our responsibility to ensure responsible development and use of this tech. We need clear frameworks, legal standards, and education to navigate this brave new world. FLUX.1 is a wake-up call - let's rise to the challenge. 

♻️ Share if you're excited (and a little worried) about the future of AI creation!

--- 

🚨 BREAKING: A New AI Supermodel Has Arrived 🚨

Introducing FLUX.1, the latest sensation in the AI world, courtesy of Black Forest Labs. Get ready to have your mind blown by this groundbreaking model's abilities!

🌟 FLUX.1: A Cut Above the Rest

This powerhouse AI outperforms even the likes of Midjourney and DALL-E 3 when it comes to generating hyper-realistic images and videos. 📸✨

What sets FLUX.1 apart? Its ingenious hybrid architecture that combines transformers and diffusion techniques, packing a whopping 12 billion parameters into one incredibly efficient model. 💪🌟

But FLUX.1 isn't just a pretty face. This model is all about accessibility, available through platforms like Replicate and fal.ai. And if you're a developer or researcher, you can even access its open-source version, FLUX.1 [schnell], under an Apache 2.0 license. 🔧👩‍💻

🎥 From Still to Motion: FLUX.1's Next Move

Black Forest Labs has big dreams for FLUX.1, aiming to take on video creation titans like OpenAI's Sora and Runway's Gen-3 Alpha. 🎬🌟

Get ready for a future where the line between reality and AI-generated content blurs even further. 😮

💡 Responsible Innovation: The Key to Harnessing AI's Power

As FLUX.1 and other AI models continue to advance at breakneck speed, it's crucial that we develop ethical frameworks, legal standards, and educational methods to navigate this rapidly evolving landscape. 📈🔍

As leaders, it's our responsibility to harness this power responsibly and guide our teams and industries through this transformative era. 🧙‍♂️💡

FLUX.1 is just the beginning. Get ready for a future where AI reshapes how we create, communicate, and connect with each other and the world around us. 🌍🚀

Are you excited about the potential of FLUX.1 and other AI models? Let me know your thoughts in the comments below! 💬

---

Forget Photoshop. Forget DALL-E.

  

Black Forest Labs just dropped FLUX.1, an AI image and video generator that's rewriting the rules. 🤯

  

This isn't your average chatbot spitting out pixelated art.

  

FLUX.1 is *real*.

  

It nails realistic human anatomy (even hands!), seamlessly blends text into images, and crushes existing models like Midjourney and DALL-E 3.

  

Here's why FLUX.1 is making waves:

  

1. **Hybrid Powerhouse:** It blends transformers and diffusion techniques, making it super efficient and mind-blowingly good. 🧠

2. **Open Source & Accessible:** Developers can tinker with it through platforms like Replicate and fal.ai. The open-source version, FLUX.1 [schnell], is even under an Apache 2.0 license! 🚀

3. **Video Dreams:** Black Forest Labs isn't stopping at images. They're aiming to take on OpenAI's Sora and Runway's Gen-3 Alpha in the video generation arena. 🎬

  

Hold up, though. This level of realism raises some serious ethical questions.

  

We need to talk about:

  

- **Deepfakes and Misinformation:** How do we prevent FLUX.1 from being weaponized for creating convincing fake videos? 🤔

- **Copyright and Ownership:** Who owns the rights to AI-generated content? The creators, the users, or the AI itself? 🤨

This is just the beginning.

FLUX.1 is a game-changer, but we need to navigate its potential responsibly.


Let's discuss: What are YOUR biggest concerns about this level of AI-powered creativity?

♻️ Share your thoughts and help shape the future of AI ethics!

## FINAL POST

I don't usually post about image generation,
But this time it's different.
here is why. 

Flux is a new model from Black Forest Labs, in Germany.

It's making Midjourney and DALL-E sweat.

That's their "llama" moment.

Meaning, finally, we have an easily accessible model that delivers quality on par with, if not better than, the top performing models. 

• Hyper-realistic human anatomy (yes, even hands!)
• Flawless text rendering in images
• Available in multiple sizes to fit every device 

Too bad it's not solving (yet ?) what cripples most image models, aka, the intellectual property issue. 

You can so easily generate things that look like copyrighted material. 

But being in Germany, I am sure the guys from black forest will have something in mind. 

Are you ready?

👍  if you're excited by the creative possibilities.

-----
*Text drafted through a chat with an AI model. Visual: Stable Diffusion derivative.
Edits by Jean-Paul Paoli. Reach out if you believe part of this content infringes on copyrighted material*

## VISUAL CONCEPT

## FINAL PROMPT


Jug XL Top model : **Steps: 4-6** **CFG: 1-2 (recommend 2 for a bit negative prompt affection)**

Otherwise : 
Turbovision XL  **Sampler:** DPM++ SDE _or_ DPM++ SDE Karras **Steps**: 3 - 5
**CFG:** 1 - 2.25

Dreamshaper : 
CFG 2, Steps 7 ou 8,  DPM++ SDE Karras en sampelr 

Neg :
out of frame, lowres, text, error, cropped, worst quality, low quality, jpeg artifacts, ugly, duplicate, out of frame, extra fingers, mutated hands, poorly drawn hands, poorly drawn face, mutation, deformed, blurry, bad anatomy, bad proportions, extra limbs, cloned face, disfigured, gross proportions, malformed limbs, missing arms, missing legs, extra arms, extra legs, fused fingers, too many fingers, long neck, username, watermark, signature

ou bien 
emotionless,cropped-frame, imbalance, poor image quality, limited video, specialized creators, polymorphic, washed-out low-contrast (deep fried) watermark, cropped, out-of-frame, low quality, low res, poorly drawn

ou bien 
3D, photorealistic, detailed textures, gradients, complex shading, blurred lines, watercolor, painterly, abstract, low contrast, muted colors, busy background, text bubbles, lens flare, emotionless, imbalance, poor image quality, washed-out watermark, cropped, out-of-frame, low res

positif:
Blue wizard with white beard and pointed hat, blue robe, standing, smiling, zen, empathy, wide angle framing, comic book style, flat colors, bold outlines, high contrast, limited palette blues and neon reds, simplified facial features, empathetic expression, graphic novel aesthetic, mood lighting, 2D art style

Llama moment for image generation with Flux


AI lacks emotion. 
That's why it can help you in emotional situations.
Here's how.

We all have those moments which are "loaded" 
→ Relationship conflicts
→ Work disagreements
→ High-stakes negotiations

AI can be your Unbiased BFF.
⮑ doesn't have emotional investment.
⮑ provides a neutral, third-party perspective.
⮑ can help diffuse emotionally charged situations.

Follow this "conversation template".

1️⃣ Provide the Context

"Here is a detailed reflection of the situation I’m facing:
(Your detailed backstory, including historical context and series of events)"

2️⃣ Seek an Unbiased Perspective

“Please take on the perspective of the other person involved and explain how they might be viewing the situation. What factors might have contributed to this issue?”

3️⃣ Clarify the Issues

“Based on the provided context, what do you think are the main issues at play here?”

4️⃣ Explore Solutions

“Suggest a few potential solutions or approaches to address this situation effectively.”

By using AI as a tool for reflection, you can gain clarity and make better decisions in challenging situations.

👍  this if you find it useful.


-----
*Text drafted through a chat with an AI model. Visual: Stable Diffusion derivative.
Edits by Jean-Paul Paoli. Reach out if you believe part of this content infringes on copyrighted material

Use AI to navigate tough situations with an unbiased perspective. Discover a conversation template to gain clarity and make better decisions.

Navigate tough situations with AI


In 1996, I competed in an Olympics, as part of French team.
Except it was the Olympics of mathematics, the "IMO". 
And yesterday, I got beaten.

Yesterday, an AI just beat my score. 

OK it wasn't a tough score to beat ... I didn't do so well 😅. 

As the IMO organizers kindly remind us, the problems are designed to be incredibly challenging.

Yesterday, DeepMind AI announced that one of its systems earned a silver medal. 

Actually it didn't participate and used 3 days instead of a few hours but still, it solved 4 of the 6 problems of this year's edition. 

Worth noting: this isn't ChatGPT. 

This is a specialized AI, similar to those developed for Go and Chess, built specifically for this task.

Beyond the hype around generalist systems like ChatGPT, this shows we're getting really good at building specialized AI that can tackle complex problems.

What does this mean for the future?

AI could help us solve some of the world's biggest challenges.

Another reason why I believe AI is here to stay 🚀

♻️ Share this if you're excited about the future of AI!


-----
*Text drafted through a chat with an AI model. Visual: Stable Diffusion derivative.
Edits by Jean-Paul Paoli. Reach out if you believe part of this content infringes on copyrighted material

An AI just beat me at the International Math Olympiad! Discover how AI is pushing the boundaries of problem-solving and its potential for the future.

How an AI beat me at Olympics


Got a request? 
There's a bar for that. 
But Google or ChatGPT?

You see a bar. 
You type words. 
But what happens next is totally different.  
  
Like the difference between a writer and a librarian.
  
One creates, the other finds.  
  
AI prompt bars are for creating.  
  
Think of it like a magic wand.  
  
You tell it what you want, and it makes it happen.  
  
Want a sonnet about how good the bolognese sauce tastes?  
  
Type it in the AI prompt bar.  
  
Want a picture of a dinosaur with a unicorn head?  
  
Type it in the AI prompt bar.  
  
Google search bars are for finding.  
  
It's a library.  
  
You tell it what you need, and it shows you where to find it.  
  
Want some facts ?  

Type it in the Google search bar.  
  
Want to learn about a biography ? To find a good restaurant ?   

Type it in the Google search bar.  
  
So, next time you see a bar, think about what you want to do.  
  
Do you want to create something new?  
  
Or do you want to find something existing?  
  
Choice is yours.  

-----
*Text drafted through a chat with an AI model. Visual: Stable Diffusion derivative.
Edits by Jean-Paul Paoli. Reach out if you believe part of this content infringes on copyrighted material*

Discover the fundamental difference between AI prompt bars and Google search bars: one creates, the other discovers. Which one suits your needs best?

Search bar or prompt bar, which one to use?

Forget chatbots and art generators. 
Gen AI value is not about Generating stuff. 

The real magic of AI lies in its ability to understand information like never before. 

Gen AI can:  

- Read and analyze any document, no matter how complex. 
- Translate information between different formats and languages.
* Uncover hidden patterns and insights in mountains of data.

The range of possible application is immense, from layers, to doctors, marketers or consultants. 

AI is becoming the ultimate knowledge worker, helping us unlock the value hidden within the vast amounts of information we generate every day.

♻️  Share this if you're excited about the possibilities!


-----
*Text drafted through a chat with an AI model. Visual: Stable Diffusion derivative.
Edits by Jean-Paul Paoli.*

Turn your data into actionable knowledge with Gen AI. Discover the true power of AI in understanding information and unlocking hidden insights. 

Gen AI: Turn data into actionable knowledge

If you use AI just to generate content or make calculations, you are using it wrong.

What if I told you AI's true superpower is empathy?  
  
Think about it:
  
- AI is available 24/7
- AI can listen without judgment
- AI can remain calm in crisis situations
- AI can bridge cultural divides, if properly trained 
- AI may be able to offer personalized support even on most intimate issue

AI will never gets tired of listening to your worries.  

In this regard, AI can be a powerful force for good.
  
What do you think?  

👍 if you believe in AI's potential for empathy.

-----
*Text drafted through a chat with an AI model. Visual: Stable Diffusion derivative.
Edits by Jean-Paul Paoli.*

Discover AI's unexpected superpower: empathy. Learn how AI can be a source of comfort and support, bridging divides and offering 24/7 understanding.

Ai Unexpected superpower

From Conversation to Experience:
Claude invents the next UI for AI.

AI is changing how we interact with technology.  
  
Think about it: 

we used to code everything.  
now, we talk to AI.  
  
ChatGPT made AI accessible to everyone.  
AI revolution is a UI revolution. 
  
Now, things are getting even more exciting.  
  
Claude, the ChatGPT competitor, just released a new feature called "Artifacts".  
  
It's a live preview of whatever you were asking Claude to create.  

Like ChatGPT, you can ask Claude to create interactive content.  

Like slideshows, quizzes, and even games.  

But now, you can see and test your creation in real-time, within the interface.

You can experience it.

Looks like the first original idea since ChatGPT. 

It's not just about processing information.  
  
It's about creating experiences.  
  
👍  if you think AI revolution is a UI revolution. 


-----
*Text drafted through a chat with an AI model. Visual: Stable Diffusion derivative.
Edits by Jean-Paul Paoli.*

Experience the future of AI: Claude's Artifacts revolutionize user experience.  Discover how AI is changing the way we interact with technology.

AI revolution is UI revolution

Meta sent notifications about using our data shared on their platforms to train AI.
Did you get to the bottom of it ? 💬

As million others, I received their short mail this week end. 

To train AI they will "use information shared on Meta’s Products and services". 🤯

What's information ? 

"things like posts or photos and their captions."

Hopefully they "do not use the content of your private messages with friends and family to train our AI"

Still, to understand better, I tried to click. 

And found pages and pages of explanations. 

To be honest quite confusing. 

If it's no big deal, why spend so much time telling me everything is fine
I shouldn't worry
it's all for public benefit and creator empowerment 

Concretely no mention of what could be the risks, if any, of letting the AI train on this type of data. 

No examples. 

Honestly, I feel pretty positive about what Meta is doing with their Llama and recent position on open research. 

This is a missed opportunity. 
To get clear about potential implications. 

On my end, I removed from my accounts photos that were including people.

I didn't understand if people faces and names could be associated or not. So, it's a no go.

And otherwise, I made my Insta account private, FB was already friends only, but that's basic privacy, good also for the rest. 🔐

Not sure it makes a lot of sense.

What did you do ? 🤔

-----
*Text drafted through a chat with an AI model. Visual: Stable Diffusion XL or DALE 3.
Final editing by Jean-Paul Paoli.*

Meta is training AI on your data. Is it really a risk ? What can you do to protect your privacy ?

Are you training Meta AI ?

Smarter is not what you expect from your next AI.
Something is far more important 


Reliability

I am not saying we want dumb AI. 

But we have already quite good ones. 
Except we are not sure they can be trusted.  

is your best friend the smartest ? 🤯

Or the one who shows up every time, reliably.  

Despite having some other flaws, that you know. Like... he is late. Or he doesn't pay. Or whatever.

We accept flaws when they are predictable.  

The same goes for AI assistants. 🤖

Sure, it's cool to have an AI that's super smart and can answer any question.

But what good is that if you can't trust the answers it gives you? 🤨

So, now that we already have the smart AI, let's focus on building AI assistants that we can trust. 

Without trust, we can't let them be autonomous.  
And the value of of all this AI tech will remain pretty low, at best.

Are you in ?


-----
*Text drafted through a chat with an AI model. Visual: Stable Diffusion XL or DALE 3. Final editing by Jean-Paul Paoli.*

Discover why reliability trumps intelligence in AI development.

You don't need a smarter AI, do you?


🤯 is your next Mac a PC ? 
Or a "copilot+ PC" as Microsoft named them 👩‍💻

Microsoft just announced crazy coming updates to Windows. 
And probably a strong comeback in the consumer space.

It's not just about a new Windows version.
It's a deep integration of AI at every layer of the OS. 
Supported by top Notch hardware, the "copilot+PC"  

By building AI right into Windows, your computer will actually understand what you're doing and help you get things done. 🤯

And open the door to having a super-smart assistant, locally. 🤖

It’s been some time that windows didn’t feel so fresh and new. ✨

Let's see if it works beyond the demo. 
But if os, ready to switch again ? 

Or we wait next month to see what Apple was … cooking ?

-----
*Text drafted through a chat with an AI model. Visual: Stable Diffusion XL or DALE 3.
Final editing by Jean-Paul Paoli.

Discover how AI is transforming Windows with seamless integration. Explore the era of smart computing and make the switch to a personal AI assistant built into your operating system.

Introducing Microsoft + AI


🧠 Why you should stop reading about AI👇

Learning AI is more than just reading, watching, or listening.

It's about getting your hands dirty. It's about experimenting. It's about interacting.

Think about it – you wouldn't learn to ride a bike by watching YouTube videos. 🚲

You gotta get on the seat and start pedaling.

Same with AI.

It's about crafting your own prompts.
It's about testing models.
It's about iterating.
It's about seeing what works and what doesn't.

🚀 That's where the real learning happens.

So, ditch the passive reading and dive into the world of AI.

Experience it.
Interact with it.
Master it.

Let's get started! 💪

-----
*Text drafted through a chat with an AI model. Visual: Stable Diffusion XL or DALE 3.
Final editing by Jean-Paul Paoli.*



Master AI by doing, not just reading. Discover how hands-on experimentation can take your learning to the next level.

Learn AI: Think less, Do More


⚡ How to leverage Gen AI to actually 
get things done better and faster ? 

It's not about LLM and Models 🤖
It's not about Prompts
It's not about Workflow

But it's all of that combined. 

Let's break it down:

1️⃣  Models & LLM  
Quite obvious, you can't do much with a dumb AI
Good news, today they are all amazing.

2️⃣  Prompt
Guiding AI models with precision ensures high-quality output 🔍
Today, still an area of progress for most. 
But everyone has this on the radar.

3️⃣ Workflow
Not a lot of noise about this one. 
But it's key.
Messy processes are slowing you down.
Frictions are killing your enthusiasm. 

That's why Microsoft has such a good position with Copilot within Office365. 

And why OpenAI works so hard on new interfaces and workflows. Voice might be the next killer office app. 

Anyway, here is the magic equation   🚀
AI Value =  Model + Prompt + Workflow 

What's your take ?   💬

-----
*Text drafted through a chat with an AI model. Visual: Stable Diffusion XL or DALE 3.
Final editing by Jean-Paul Paoli.*

Unlock AI productivity with the magic formula: right tools, well-crafted prompts, and optimized workflow. Discover how to elevate your human-machine collaboration.

AI Productivity magic formula revealed


GPT-4 is the Coca-Cola of AI.

it's the original formula that set the standard, and people cling to standards. �

When new AI models appear, they're compared to GPT-4, just like how we compare new cola flavors to the classic Coke.

Changing the recipe or tweaking the model is like altering the Coke formula - some fans will be furious

And everyone will have an opinion. 😔

But here's the twist: while we love our familiar brands, we still cheer for progress

So let's see which new flavors openAI is coming up with
and quench our thirst for knowledge and creativity! 💃🏻

-----
*Text drafted through a chat with an AI model. Visual: Stable Diffusion XL or DALE 3.
Final editing by Jean-Paul Paoli.*


Discover why GPT-4 remains the gold standard in AI, and explore the innovations and controversies surrounding its reign, from AI enthusiasts and haters alike.

Gpt4 : The Coca-Cola of Ai


🚀 AI Race: a llama might be winning … but nobody should care.

Even by Gen Ai Standards, the last 2 weeks have been crazy.

A flurry of language models surpassing or matching the performance of GPT-3.5 and GPT-4! 🔥

First, Cohere's open-source model left GPT-3.5 in the dust 👋

Not to be outdone in the Open source space, Mistral came up with an innovative model "8x22B", on par with GPT-4 🔗

In a couple of days, Microsoft fine-tuned this model for an even better version. That was "WIzardLM2" apparently surpassing GPT-4 🚀

But then, OpenAI dropped a new GPT-4 iteration … leaving all others in its wake 🌊

And just when we thought it was over, Meta released LLama3
it's looking incredible, potentially better than all previous ones, and open-weight to boot! 🏆

I will be honest …

I am not sure anybody outside the small community of AI models builders care anymore.

All these models are already crazy good.
No need to risk burnout by testing them all.

Let's focus on use cases and unlocking some value. 🎢

-----
*Text drafted through a chat with an AI model. Visual: Stable Diffusion or DALL-E 3.
Transformation and final editing by Jean-Paul Paoli.


Explore the thrilling advancements in language models, but remember, it's time to focus on practical use cases and unlocking value.

Fast Llama causing AI burnout


🤖 What if the Best AI Isn't the Smartest, but the One You Collaborate With Best? 👇

Just like how every individual has their unique strengths, AI are no different!

Here is my take, for what I have been able to experiment

OpenAI GPT4?
It's like the friendly, versatile colleague in town.
Always ready to help, and quite competent 🤝💪

Mistral Large?
That's your detail-oriented, analytical colleague.
Give it a task, and it'll follow instructions to the T! 🎯🔍

Claude Sonnet?
Oh, voilà the creative one. A bit poet.
Great for novel ideas and approaches! 🎨💡

💡 Spend time with them, and uncover their distinct traits is the best way, in the end, to choose the right model.

Next step: course on AI team management? 🤝🚀

-----
*Text drafted through a chat with an AI model. Visual: Stable Diffusion or DALL-E 3.
Transformation and final editing by Jean-Paul Paoli.


Discover the unique strengths of top AI models: GPT-4, Mistral Large, and Claude Sonnet. Choose the ideal AI partner for your needs. Dive into AI team management!This is description

Not just smart: AI with character

🏅Just got a Top Voice badge. Let’s unpack the ego boost … and Microsoft strategy👇

1️⃣ superficial ego boost

I know it’s just a digital pat on the back. 
A gamification tactic to get people hooked. 
But still … social media has a way of getting under our skin. 
I am working on it, but the likes, comments and badge thing is a tempting game 📈

2️⃣ a tough game

It’s a little badge … but kind of big work 💪 
Consistent posting on those « collaborative articles ». 
Trying to craft a crisp message, bringing value in a matter of seconds. ⏰ 
It's not just a click of a button.

3️⃣ a fantastic play by Microsoft 

While we're playing the game, #LinkedIn (aka Microsoft) is playing chess. 🎓

Every collaborative article people create is another data point. 
Great writing, a treasure trove of Q&A data, annotated with likes by humans.

And we're the willing contributors.
Congrats guys for this idea 💡

And thanks for the badge.

-----
*Text drafted through a chat with an AI model. Visual: Stable Diffusion or DALL-E 3.  
Transformation and final editing by Jean-Paul Paoli.

Unpack the real impact of that LinkedIn Top Voice badge - from ego boost to Microsoft's strategic data play. Explore the hidden dynamics behind social media gamification.

Just got a badge, now what ?


Should you use Ai to learn ? or will it just make you dumber ? It's not clearcut 👇

AI can edit and reformulate articles in seconds.
May be whole books.

But wait, isn't reformulation the key to deep learning? 🤔

Everyone knows the magic. 

Engaging with the content, wrestling with the ideas, finding your own words - that's where the real learning happens. 📚

And may be it will be OK with you, but what about students that may never have to use their   own Critical Thinking 🧠 to reformulate and analyze a text. 

Will we loose our mental muscles to think independently ?  💪 

But on the other hand, AI unlocks multiple benefits: 

🎨 Personalization aka support for diverse Learning Styles
AI presents information in various formats, catering to different learning preferences.

🌍 Accessibility
AI makes complex concepts more digestible, opening doors to learning for more people.

⚡ Efficiency
AI allows us to cover more material in less time, freeing up space for deeper engagement with certain topics.

So like any tool, it's all about how we use it.

Don't use it to cut corners... 
But to dig deeper into whatever you need to master  🔍 

Discover the pros and cons of using AI for learning. Will it enhance or hinder critical thinking? Dive in to find out.

Does AI make you smarter or dumber ?

Microsoft just made an unusual AI announcement.👇

For once, it's not about more power, 
or more parameters,
or more IQ. 

They introduce BitNet b1.58. 
And it's mainly a stride to a greener AI 🌟🌱

It slashes energy consumption by 71x ...
while keeping performance  🚀

💡 The trick ? 
Instead performing calculating using full precision, with numbers that need 32 bits to be stored, it uses just ternary values : {-1, 0, 1}.  

Which require 1,58 bits to be stored, so the name.

And good news, in this strange world of ternary computation, multiplications become additions. 

That would also open the door to further optimizations and speed up. 

Because, as you and me, standard computers are much better at additions compared to multiplications. 

So  in the end... Smaller Bits, Bigger Dreams ? 

-----
*Text drafted through a chat with an AI model. Visual: Stable Diffusion or DALL-E 3.  
Transformation and final editing by Jean-Paul Paoli.*

Discover BitNet b1.58, Microsoft's green AI breakthrough slashing energy use by 71x. Dive into the world of ternary computing and optimized multiplications.

Tiny but Mighty: Microsoft Bitnet

Mathematics is useless.

Or so they said.
And then AI happened 🧠 🤖

You heard it on Tv, in newspaper, in conversations. 

"What's the point of learning derivatives"

🤖AI, which is poised to become ubiquitous in our life

Math is the heartbeat of AI.

It's calculus 
to allow AI to learn from its mistakes 📈

It's linear algebra 
to enables AI to see patterns 🌀

It's probability 
to empower AI to recombine and create.🔮

👉 In essence, AI is math brought to ... life.

So now, every country in the world is suddenly wondering where they can find skilled mathematicians.

And those which have strong traditions can be proud. 
For once, 🇫🇷 is not bad at that game.

So, next time someone says, "Mathematics is useless", you know what to say.

Now, the guessing game ... why an apple in this image ?

-----
*Text drafted through a chat with an AI model. Visual: Stable Diffusion or DALL-E 3.  
Transformation and final editing by Jean-Paul Paoli.*

Uncover the significance of mathematics in AI: From derivatives to linear algebra and probability, learn how math drives AI's ability to learn, see patterns, and create. Discover why France excels in this field.

Mathematics is useless

✨ "Don't be evil." Yesterday, Google made an unexpected move to reignite the love it had lost. 

Indeed, you might remember the 2000s, when Google was David against the Goliaths of the time.

It promised a different path.

One where access to information was universal. Between YouTube, Gmail, Maps, Google is not just a search engine; it is a gateway to the world's knowledge, free and accessible to all. 🚀

Google was a company but one with a special aura, it was a beacon of "good". 

But, with its growth, Google has become a titan, dealing with multiple conflicting imperatives 🌍

And so the positive aura has been a bit lost.

Yesterday, Google dropped Gemma, a range of open-source, small but state-of-the-art AI models.

Behind the tech perf, it's a strategic statement. Gemma is a storyline of redemption. 🌈 "We're still those guys."

With Meta playing strong the open-source game, and OpenAI the proprietary one, they go like Mistral with a mixed one:

- big powerful closed models (Gemini)

- smaller but still very competent open LLMs (Gemma)

👩‍💻 Now, over to you, do you think Gemma signal a strategic shift?

-----
*Text drafted through a chat with an AI model. Visual: Stable Diffusion or DALL-E 3.  
Transformation and final editing by Jean-Paul Paoli.*

Discover Google's latest move with open-source AI models, Gemma. A story of redemption and strategic shift in the AI landscape.

The Return of the Beacon

AI is not going to replace you soon. Here is why 👇 

This simple riddle is the latest joke among AI experts.
It stumps most of our sophisticated GPT models. 🤖🎢

Here is it: 

You see one group of ducks and
- one duck is behind two ducks, 
- one duck is in front of two ducks,
- one duck is in the middle of two ducks. 
How many ducks at least are there in the group?

Seems simple, right?  
🦆🦆 🦆  3 ducks et voilà  

Yet, 
Mixtral finds 4 ducks, 
GPT-3.5 thinks 5 ducks, 
Gemini Pro 6 at minimum, and "there may be more"

It's a veritable festival of errors, except for GPT-4, which apparently got it right. 🎯

This humorous highlight underscores a serious point. 
👉 Today GPT have no reasoning capacity. 

They are word machines producing illusion of reason. So while they can dazzle us with complex tasks, they fail seemingly simple others 🧠

So when we worry about AI outsmarting humanity, let's remember the duck riddle. 

So 🦆 vs 🤖: who is smarter ?

-----
*Text drafted through a chat with an AI model. Visual: Stable Diffusion or DALL-E 3.  
Transformation and final editing by Jean-Paul Paoli.*

AI won't replace human intelligence soon, as shown by  a simple duck riddle that reveals the reasoning limits of leading GPT models.

AI are dumber than ducks

🧩 Remember when AI turned us all into superheroes ? I tried to build my AI avatar. It didn't work well... 👇

I am sure you remember how one year ago, everybody was doing those fancy profil pics.

As a Superman hero or queen of Coachella.

That was the "Lensa" app craze. 

I don't read so much about the app anymore but they exploited an interesting technique. 

Gen AI models are not "fixed". 🤖 They can continue learning and you can quite easily teach them new things called "concepts". 

For instance the concept of your face to generate yourself in whatever situation you can imagine.

It's called "LoRa" training. And all the needed code can be find online. 

So I decided to train a LoRa to craft my AI avatar machine 🤖 

As you can see, and despite some efforts, it didn't work. 

My dream avatar machine remained out of reach 😶

Instead, I was met with eerie uncanny distorted portraits of another me 🖌️ 

Weird experience. 

One thing is clear 💡 the youth of the technology and the randomness of AI outputs make every Gen AI project an exploration. 

Success is not guaranteed. But the journey is amazing.

So, let's keep exploring 🚀

Share your own AI stories below! ⬇

-----
*Text drafted through a chat with an AI model. Visual: Stable Diffusion or DALL-E 3.  
Transformation and final editing by Jean-Paul Paoli.*

Experience the thrill and unpredictability of AI exploration as I attempt to create an AI avatar using LoRa training, uncovering the limits of current technology.

Every AI project is an exploration

Soon, we will go from an idea to a prototype in minutes rather than weeks.  And even not asking about it 👇

Here is how it goes :

Those ideas and hypotheses exchanged in a heated conversation?   
They will feed an AI system, becoming prompts. 

The AI could be quietly launching agents, running hypotheses, making images, even codes and apps !

Turning desires into digital reality. 🤖✍

👉 Welcome Conversational Prototyping 🔥

AI could even evaluate your ideas, help solve the challenges and brainstorm with you, doing Problem-Solving as You Speak 🔥

It's like having the Genius from the tale, but not in a bottle, in the room 🔄🚀

Of course there are a couple of things to iron out, starting with privacy issue. 
But it will be solved. 

How do we prepare for the next step ?

-----
*Text drafted through a chat with an AI model. Visual: Stable Diffusion or DALL-E 3.  
Transformation and final editing by Jean-Paul Paoli.*

Discover Conversational Prototyping - AI turning ideas into prototypes in minutes. Experience problem-solving and brainstorming in real-time. Get ready for the future!

Conversational Prototyping

I read the best book on AI and learnt a powerful insight 👇

Xmas 🎄🧠 : a perfect moment to retreat from the ever-spinning AI news cycle and read a book. 

👁️ I dived into Melanie Mitchell's "Artificial Intelligence - A Guide for Thinking Humans."

It's an amazing book to really understand AI history, performance and limitations. 
And get perspective on the matter. Even ChatGPT recommends it first when prompted :)

The insight I remember most vividly ? 

👉👉 "Easy things are hard"

AI flips the script on what we consider easy and hard. 🔄 

Humans excel at certain tasks effortlessly. Like picking a friendly face in a crowd. Or understanding a scene from a picture. Or understanding a conversation in a noisy bar.  

For AI it took years to get a decent result. 
And it's still far from perfect. 
Prone to lots of biases, even massive mistakes.  

Yet it can outperform us on tasks we find complex. 
Give it a chessboard or go board? Boom. It beats grandmaster since years. 🤯🤖

It's an eye-opener on how "alien" AI really is. 
On why self driving cars might not be around the corner. 
Or why it might take years until we solve hallucinations or bias properly.

2024 is the year of maturity with AI. 
There are score of issues we can solve with AI 🏆
To do so, let's make sure we have enough understanding and perspective. 

Any other book you recommend on AI ? 👇

-----
*Text drafted through a chat with an AI model. Visual: Stable Diffusion or DALL-E 3.  
Transformation and final editing by Jean-Paul Paoli.*

Discover how AI excels at complex tasks but struggles with simple ones. Uncover the "easy things are hard" paradox in AI.

Unveiling AI's Paradoxes

🤖 What if we talked about "Smart" instead of "Intelligent" ? 
It opens a fresh perspective. Discover why 👇 

What's AI ? What's intelligence after all ? 
How do you measure it ? 

Nobody seems to agree. 

But everybody knows what "smart" is.

We have smartphones, smart homes, smart clothes, smart cities... 🏙️
We live in a smart world. 🌎 

So, what's Smart ? 

Smart Clothes ? they adapt to the weather. 
Smartphones ?  they adapt to your needs.
Smart Home ? they adapt heat, light etc.. to the environment  

Smart means able to adapt.
💡 So here it is: AI is just a smart program. 🚀

It adapts 
to your request 
and it's big and broad. 

And until then, this kind of adaptation was only seen in organic life.
So all the fuss about artificial intelligence.

👇 Drop a "🤖"  if you agree 😊  

-----
*Text drafted through a chat with an AI model. Visual: Stable Diffusion or DALL-E 3.  
Transformation and final editing by Jean-Paul Paoli.*

Discover the power of redefining AI as "smart" and how this shift in perspective highlights its adaptability to different situations.

Redefining AI

🚀 GPT-4 Manual DROPPED 📚

After one year of everybody trying to guess how to use the thing 
After one year of seeing those pricey "expert" courses? 🤑
💥 OpenAI just served up the ultimate cheat sheet. 
The official GPT-4 manual is here. 

🧐 Nothing earth-shattering though :(

Still, it's a solid A-Z on making GPT-4 and its cousins work hard for you.

🎯 So, if you need a no-cost bible, with all the strategies and smart tactics for top results, and link to tons of extra, this is it. 

🔗 Read it here : https://platform.openai.com/docs/guides/prompt-engineering/prompt-engineering 

-----
*Text drafted through a chat with an AI model. Visual: Stable Diffusion or DALL-E 3.  
Transformation and final editing by Jean-Paul Paoli.*

OpenAI has unveiled the ultimate guide to mastering GPT-4. Discover everything you need to know in this no-cost resource

GPT4: Missing Manual finally found

🔍 The "Pessimist archive" : meet all the naysayers that have been wrong 👇 

According to them : 
- Moon landing was useless
- Mac were ruining writing skills
- Photocopiers threatened publishing

Pessimists have always been there😱
echoed by media that thrives on fear and spicy news  

This newsletter gathers all press clippings and stories about these guys. 
It's refreshing:  https://newsletter.pessimistsarchive.org 

We shouldn't be naive – risks are real. 
But so are possibilities. 

While some cried doom, others built dreams.

We live in a world where learning is no longer locked in ivory towers. Knowledge is a click away, for free 📚✨

What do you do with this free gift ?  🚀

-----
*Text drafted through a chat with an AI model. Visual: Stable Diffusion or DALL-E 3.  
Transformation and final editing by Jean-Paul Paoli.*

Explore the "Pessimist Archive" for a refreshing look at how fearmongering and skepticism have been proven wrong throughout history.

The Pessimist Archive - It already happened before

AI Hyperrealism is changing the game 👇

A recent study has unveiled that AI-created faces may be deemed more real by human judges than real human pictures.

It's a phenomenon called AI Hyperrealism 🤖✨

Some AI images are so lifelike, 
so meticulously detailed, 
and so 'average'
that our brains think they're genuine.

Here's the catch – if synthetic can outshine authentic, where does that leave our gut instinct? 🤔

This isn't just about pictures – it's a metaphor for our entire information diet. 🍽️

We need to sharpen our focus, equip ourself with skepticism at scale ... and let thorough verification be our shield. 🛡️

Because in the digital realm, not all is as it seems.

-----
*Text drafted through a chat with an AI model. Visual: Stable Diffusion or DALL-E 3.  
Transformation and final editing by Jean-Paul Paoli.*

AI Hyperrealism blurs the line between real and fake, with synthetic faces often surpassing human perception. Prepare for a new era of digital skepticism.

When Synthetic Outshines Authentic

🔥  Hot off the press
GPT4 is better than seasoned human columnists to give relevant and empathetic advices on our personal conundrums. 

What's new ? 

We knew people preferred GPT over doctors to give medical information, but come on, doctors might not be the best competitor for empathy.  

Here, the columnist mission is to provide the most "satisfying" answer. 💓

And still GPT4 answer was preferred, judged more comprehensive, empathetic, relevant etc... 

On paper, AI seems to have cracked the empathy code, or has it?

Let's get real. 
Sure, AI can offer a shoulder, sort of. 
But in the end, when we share a story, what we really want is human touch  – the warmth of a human "I get you." 🔥 

Not only an advice, but an acknowledgment of who we are. 

As a human.
In the end, we humans love other humans.  

👉 So no matter how "empathetic" GPT4 or 5 words get, nothing beats a hug from somebody that "understands" you silently 🤗 

So Tell me, would you prefer a heart-to-heart with HAL 9000 or your bestie? 🗣️👥

-----
*Text drafted through a chat with an AI model. Visual: Stable Diffusion or DALL-E 3.  
Transformation and final editing by Jean-Paul Paoli.*

Discover why GPT4's empathetic advice outranks human columnists, yet falls short in providing the genuine human connection we crave in heart-to-heart conversations.

What GPT4 will never manage to beat human at

🤖 Navigating AI Pitfalls: Learn from My Evening That Went Rogue!

It was last week, wanted to use OpenAI to translate some stuff. 
Found this piece of software ... looked so cool ! 

But it needed slight ajustements. 
And now we have AI coding assistant 🛠️, right, so it should be easy !🧑‍💻✨

it turned into a 3-hour epic fail.🎇❌
I nearly insulted GPT4.

After a saga of errors and self-belief, the hero was... Google Translate. 🎉 Free. Fast. Functional.

So, here's the lowdown:

- AI won't make you superhuman 🦸‍♂️🚫
-  Simpler beats clever. Sometimes, the old-school way just works.

Got your own AI frustration story? Spill the beans below! 🗣️💬

-----
*Text drafted through a chat with an AI model. Visual: Stable Diffusion or DALL-E 3.  
Transformation and final editing by Jean-Paul Paoli.*

Navigating AI Failures: Learn from My Epic Mishap! Discover the limitations of AI coding assistants and why sometimes simplicity trumps complexity.

Navigating AI Pitfalls

👋 Breaking News: GPT5, just released, replaced Sam Altman as CEO of OpenAI 🚀 Exclusive insider story 👇  
  
GPT5 has been released today.  
  
In a shocking finding, data scientists measured it was :  
- as good as a Sam for strategic decision  
- better to answer senator's questions  
- a lot better to draft governance paper and responsible AI framework  
  
And obviously much cheaper.  
  
Is this the dawn of AI leadership? 🌅  
  
🛑🛑🛑 STOP 🛑🛑🛑  
  
Sam Altman has been dismissed but Mira Murati, current CTO, is the new boss.  
  
Nobody has any clue as to why.  
  
And neither do I have.  
  
If you happen to know👇  
(and btw GPT5 is not ready)

-----
*Text drafted through a chat with an AI model. Visual: Stable Diffusion or DALL-E 3.  
Transformation and final editing by Jean-Paul Paoli.*

Is AI ready to lead? GPT5 replaces Sam Altman as OpenAI CEO. Discover the shocking findings and explore the future of AI leadership.

GPT5 has its first victim

Music discovery: my favorite radio is better than Spotify. What does it mean for recommendation algorithm ?  🤔

🎧 Algorithms work tirelessly to match your tastes, learning from every click, pause, and skip.

My "discovery" section of Spotify is full of music I like.

I rarely skip. 
But I rarely favorite💓. 
It's all about satisfaction on demand.

But the music you love, it's about ...  something slightly unexpected 🔥
The right dose of surprise. 

Curation is an art. 
For me, for now human are far better than machines.  

And for you ? 

Which do you trust to discover? Spotify, Tiktok algo or curator playlists?

-----
*Text drafted through a chat with an AI model. Visual: Stable Diffusion or DALL-E 3.  
Transformation and final editing by Jean-Paul Paoli.*

Uncover why my favorite radio beats Spotify in music discovery. Discover the secrets of human-curated playlists and their edge over AI algorithms.

Music Discovery

AI makes it easy to impersonate your relatives. Elaborate scams are coming. One simple way to keep your family safe👇

Imagine: Your daughter calls, asking you to wire money immediately. The voice – her voice – quivers with urgency. But something feels off... 🤔

Or your father. 

And it's not a call, it's a video. 

But wait – is it truly your daughter or your father?

Or just an AI deepfake. 

In the coming storm of deepfake technology, voices will be replicated, faces will be mimicked, and our trust could be shattered. 😱

But you have a secret weapon. It's an old spy trick, very simple.

A family password 🛡️

Here's the plan:
1. Gather your tribe. Your kin. Your closest allies. 
2. Choose a word, a phrase, something only this tribe will know. 🤫
3. Use it to verify each other in times of doubt.

So take the time, have the conversation LIVE – no texts, no emails. Create that password and protect your circle.

I did it last Sunday, comment if you want to know our secret pass for inspiration 👇

-----
*Text drafted through a chat with an AI model. Visual: Stable Diffusion or DALL-E 3.  
Transformation and final editing by Jean-Paul Paoli.*

Safeguard your family from AI deepfakes with a simple family password. Learn how to shield your loved ones from elaborate scams.

Family password: a shield against deepfakes

2 mm to get what’s in the White House AI Executive Order 🏛️🇺🇸 and why it does imatter  
  
AI just made it to the Oval Office with President Biden’s EO (Executive Order) released yesterday.  
  
4 key points 🤔  
  
1️⃣ Standards  
Via federal agencies, the US government will define AI standards on a wide range of issues, such as safety, security, ethics, equity, and more.  
  
2️⃣ Transparency and consumer protection  
content that is generated or manipulated by AI, will be signaled.  
  
3️⃣ Privacy  
many references to privacy protection, especially for children and other vulnerable groups  
  
4️⃣ Leadership  
The USA wants to reaffirm its leadership in AI innovation and competition  
  
Concrete measures and execution are not clear at this stage.  
  
But for sure when people were saying USA would be “free for all” vs an all too regulatory EU they were wrong.  
  
Sounds like life won’t be so easy for startups building critical Ai systems in the US.  
  

-----
*Text drafted through a chat with an AI model. Visual: Stable Diffusion or DALL-E 3.  
Transformation and final editing by Jean-Paul Paoli.*

Discover the AI Executive Order from the White House and its impact. Learn about the 4 key points on AI standards, transparency, privacy, and US leadership.

AI Executive Order

🚫 **"AI Ethics: Sometimes, It's More Simple Than It Sounds"** 🚫

'AI Ethics' might sound sophisticated. And there are sometimes complex questions. 

But sometimes it's very simple.  

It's just plain cheating. 

📜 Opinion pieces  
You follow some interesting people on social network. 
Turns out there is in reality nobody behind the ideas. 
Not even a ghostwriter. 
it's a robot. 

That's not thought leadership—it's deception.

📣  Political Deepfakes
A politician delivers a heartfelt speech in the audience language, via a deepfake video, not being transparent about it, as if he was fluent in said language. 

But he is not. 
That’s not bridging cultural gaps—it's deception. 

⚱️ "Milestone" Moments 
We all heard stories about condolence letter AI generated. 
Or automated birthday wishes.
Or fake wedding speeches. 

This is not showing affection—it's deception.

Yes AI can help think, speak, write. 

But we need transparency, especially with what matter most : human connection. 

🗨️ When does AI enhancement becomes a counterfeit? Weigh in.

-----
*Text drafted through a chat with an AI model. Visual: Stable Diffusion or DALL-E 3.  
Transformation and final editing by Jean-Paul Paoli.*

From AI-generated thought leadership to political deepfakes and automated heartfelt messages, sometimes ethics is simple.

AI Ethics

GPT: the double meaning most people miss ... Don't be one of them 👇 

You learnt that GPT, as in ChatGPT, means Generative Pretrained Transformers. A shortcut for a technicality.  

Like a spec sheet. 

But it's not just that. 
GPT also means General Purpose Technology.

Think Writing, Steam engine, Electricity, Internet – the  tech game-changers. 🚂💡 

These have been studied extensively.  

Electricity didn't flip the switch overnight.
Steam engines didn't make the horses obsolete in one day. 
FIrst time you "tried" Internet and its empty search bar, you felt it was useless. 

Their real influence took years, if not decade, to be felt ⏳

Same with AI, or GPT. Today, nothing is clear. Tomorrow it will be everywhere. 

So, next time you drop the term "GPT" remember – it's not just tech. It's a prophecy 🌍

Now the real question : did openAI, that invented the acronym GPT and made it a brand, do it on purpose ? Or they were lucky  🚀 

GIve me your vote 👇

-----
*Text drafted through a chat with an AI model. Visual: Stable Diffusion or DALL-E 3.  
Transformation and final editing by Jean-Paul Paoli.*

Unlock the double meaning of GPT and its profound impact on technology. Was OpenAI's choice of acronym intentional or luck?

General purpose technology

🧙‍♂️ After a year crafting 1000+ AI-designed wizards, here's a secret about GenAI 👇

First, watch the [video](https://www.linkedin.com/posts/jppaoli_data-generativeai-innovation-activity-7121003100714250240-mJUI?utm_source=share&utm_medium=member_desktop)

GenAI is not about saving time, it's about dreaming bigger  ✨

1000 wizards, it's not all magic wand waving. 
It's hours of ideating, prompting, tweaking, testing models and refining. ⏳🔧

GenAi is not a timesaver 
It's not some quick fix 
It's no shortcut  

It's a springboard.
It's a catapult for your goals
It's about stretching the boundaries of the achievable.

💡 So embrace it, harness it, and use it to paint the universe of your wildest dreams. 🎨✨

Let's make it count 🚀

-----
*Text drafted through a chat with an AI model. Visual: Stable Diffusion or DALL-E 3.  
Transformation and final editing by Jean-Paul Paoli.*

Discover how I leveraged Generative AI to design 1000 wizards. Don't just save time, dream bigger with AI!

What did I learn after designing 1000 Wizards

🔨Prompt Engineering, a key skill in the future, has little to do with "engineering". But what is it then ?  

Prompt engineering is the ability to prompt a model to achieve desired outcome.

There is a consensus saying it will be a key skill in the future. Prompt engineering is engineering - it's in the name, right? 

But is it really 'engineering'?  
Let's think about it. 🤔

Engineers follow blueprints, calculations, formulas. There's precision. There's predictability. 

Prompt engineering? It's a bit more... chaotic. More experimental. More uncertain. More... fun? 

Like an artist 🎨 may be ? With a palette, brushes, and canvas ? 

Or may be even a 'craft' ?  A blend of science and art, data and intuition, precision and creativity.

But artists aim at beauty. Prompters want to hit gold. To me, that's the definition of Alchemy.

Are prompt engineers then, the modern alchemists ? If so, we hope for a brighter future :)

Let me know your thoughts! 💡

-----
*Text drafted through a chat with an AI model. Visual: Stable Diffusion or DALL-E 3.  
Transformation and final editing by Jean-Paul Paoli.*

Discover the art of prompt engineering, the key skill shaping the future. Unleash your inner alchemist and transform AI models today

What is prompt engineering

Starting a new position as Generative AI Business Transformation Director 🚀 at [L'Oréal](https://www.linkedin.com/company/lor%C3%A9al/) !  
  
Too many talented people to thank and remember, from my amazing team, colleagues and leaders at PPD France ❤️ to all our partners from the stylist digital ecosystem.  
  
A business that blends so well a "cutting edge" vision ✂ powered by tech & data with a people first culture. 👨‍🎓  
  
Now, working on another blend, let's make our dear tyrannosaurus even more unicornus ! 🚀 

-----
*Text drafted through a chat with an AI model. Visual: Stable Diffusion or DALL-E 3.  
Transformation and final editing by Jean-Paul Paoli.*

Unleashing AI's potential in business transformation at L'Oréal, making the beauty industry smarter and more efficient.

starting a new position

🌱 Le "jardin numérique", 🌳 c'est urgent de se lancer. Mais pourquoi ?  

Non, ce n'est pas un compte Instagram caché, 
c'est une forêt 🌲 de connaissances. 

On appelle ça aussi un "second brain". 🧠 

Très concrètement c'est une collection de notes, d'articles lus, de "highlights" de livres, de pensées. 

Un hub de connaissance perso. Et ce concept n'est pas nouveau. 

Beaucoup le font déjà. Avec des fichiers épars, ou des apps spécialisées, depuis le parrain Evernote jusqu'à Roam ou Obsidian. 

Ce deuxième cerveau permet de libérer l'espace mental et d'améliorer sa rétention d'infos, finalement de continuer à apprendre. 💡

Mais soyons clairs, jusqu'ici ça pouvait être pénible. Et on se retrouvait souvent avec un tas de notes inutiles. 

Mais avec IA générative, tout change 🤖 

ces jardins vont être 1000x plus utiles : 

- Fini de farfouiller dans toutes ces notes, l'IA sera là pour répondre à toutes les interrogations
- Et surtout ce sera une source d'entraînement de données de confiance et de valeur pour son IA personnelle.

Le jardin numérique, c'est la clé qui fera la différence entre nos assistants personnels. 📚 

Alors, on plante les graines de la connaissance ? 🌱

-----
*Text drafted through a chat with an AI model. Visual: Stable Diffusion or DALL-E 3.  
Transformation and final editing by Jean-Paul Paoli.*

Découvrez comment un "jardin numérique" boosté par l'IA peut libérer votre esprit et améliorer votre apprentissage. Plantez dès maintenant les graines de la connaissance pour votre IA personnelle.

Pourquoi cultiver un jardin numérique

🤖Course à la performance IA : faute de chronomètre, on ne sait pas vraiment si on s’améliore. 🏁 Ni où se situe la ligne d’arrivée. 

😳 Une méthode simple c’est de poser des questions à un modèle et de mettre des juges humains pour noter les réponses.  

Mais évidemment ça coûte cher. 💵
Et c’est biaisé. 

Alors la recherche de benchmarks automatiques est très active. Il y a MMLU dont les résultats  semblent corréler fortement avec les jugements humains.  

Mais honnêtement ce n’est pas très clair. 🤷‍♀️

📊 En vrai est ce que le modèle à 61.2 est vraiment mieux que celui à 56.7 ? Et à partir de quelle valeur est ce qu’un modèle est « suffisant » pour être utilisable en vrai ? 

😳 On a besoin vite d'une méthode fiable pour mesurer l'intelligence des LLM. 
Comme pour les humains. 🧠

Ah, zut, pour les humains on sait pas non plus. 
👀 L'intelligence n'est pas une chose simple à mesurer.🌹

💡Mais est ce vraiment si important ?

c’est bien de savoir résoudre des problèmes, ou créer 15 idées nouvelles. Mais ce qui est encore plus fort c’est de savoir interfacer cette capacité avec les autres. Pour concrètement faire avancer le monde. 🌍 

ChatGPT a fait un grand bon mais on attend vraiment le prochain. 🚀

-----
*Text drafted through a chat with an AI model. Visual: Stable Diffusion or DALL-E 3.  
Transformation and final editing by Jean-Paul Paoli.*

Découvrez les défis de la mesure de l'intelligence des IA et pourquoi la collaboration humain-IA pourrait être la clé pour faire avancer le monde.

Mesurer l'intelligence, est-ce important ?

🚀Stable diffusion XL est là: mon mage vient de recevoir une upgrade✨, Midjourney n'a qu'à bien se tenir. 🔮 

Stability AI a mis à dispo hier Stable Diffusion XL (SDXL) 1.0, la dernière version de sa suite de modèles texte-en-image 💬🎨

What's new ? 
plus de détails, plus de contrastes, plus de résolution  🌈⚡

C'est simple, SDXL est nettement meilleur sur la qualité vs la précédente version. 

🧙‍♂️ Alors qui a encore besoin d'un vieux chapeau poussiéreux ?  Prêts à expérimenter ?  🤩👏

-----
*Text drafted through a chat with an AI model. Visual: Stable Diffusion or DALL-E 3.  
Transformation and final editing by Jean-Paul Paoli.*

Découvrez Stable Diffusion XL, l'outil de génération d'images IA révolutionnaire. Plus de détails, de contrastes et de résolution pour libérer votre créativité.

Release de Stable Diffusion XL

🚀L'IA, ça va être comme l'ordinateur : d'abord il faut une machine qui remplit une pièce 🗿et après ça devient petit, personnel, mobile 📱. 

On est dans l'ère des modèles d'IA géants, mais qu'arrive-t-il lorsque l'on passe à l'échelle personnelle ?

👀 Si chaque individu possède sa propre IA, aussi puissante que GPT-3, mais personnalisée selon ses besoins. C'est plus big data, mais de MY data. 💼

Votre IA personnelle connaît vos goûts, vos habitudes, elle peut même anticiper vos besoins. Elle n'est pas seulement un assistant, c'est votre alter ego numérique.🕴️‍♀️

Et au fait  si votre IA personnelle connaît tout de vous, qu'en est-il de votre vie privée? 🕵️‍♀️ Et si elle tombe entre de mauvaises mains ? Si elle se fait hacker ? 

Pire encore, et si elle devient plus intelligente que vous? 🧠

Demain c'est dispo, vous l'installez sur votre mobile ?   🎲 

-----
*Text drafted through a chat with an AI model. Visual: Stable Diffusion or DALL-E 3.  
Transformation and final editing by Jean-Paul Paoli.*

Découvrez le potentiel énorme d'un IA personnelle, locale, adaptée à vos besoins, et explorez les défis liés à la protection de vos données personnelles.

L'IA va se miniaturiser : et après ?

👑 Garantir son emploi aujourd'hui c'est être conforme, produire du pattern. Problème : tout ce qui est pattern c'est l'IA qui va s'en occuper  🤖   

Garantir son emploi c'est bien faire les choses.  

Ça démarre à l'école où on suit le programme 📚. 
Ça continue dans certaines fonctions au travail, où l'on apprend les procédures pour remplir les objectifs 🏢.
Dans tous les cas il faut les bons comportements, et les codes.

Mais attention, trop suivre la recette et hop, c'est l'IA 🤖 qui va prendre la main. 

Elle est plus efficace, plus rapide et moins coûteuse pour reproduire les patterns.

🎯 Donc en fait demain la zone de sécurité, c'est la non-conformité, l'originalité, la différence. Tout ce qui ne peut pas être automatisé. 

Alors, pour déjouer les patterns, une seule solution : embrasser sa singularité 🦚 

Things are going to get interesting. 🚀
Et pour commencer, laissez votre divergence en commentaire 👇

-----
*Text drafted through a chat with an AI model. Visual: Stable Diffusion or DALL-E 3.  
Transformation and final editing by Jean-Paul Paoli.*

Explorez comment embrasser votre individualité peut aider à sécuriser votre emploi dans un monde rapidement dominé par l'IA. Embrassez vos différences et restez en avance sur le jeu de l'IA.

Prime de non conformité

🧠 ChatGPT aurait-il perdu des neurones ? C'est la grosse rumeur mais les conséquences peuvent être profondes 🔥

Sur Twitter et Hackernews, c'est l'ébullition.⚔️  

Difficile à prouver, mais c'est aussi mon expérience : des prompts qui fonctionnaient à merveille produisent maintenant des sorties de mauvaise qualité, voire carrément hors sujet. Et ce malaise n'épargne pas GPT4 et son API. 

💥Une IA qui perd en intelligence, c'est une pilule amère à avaler pour tous ceux qui l'utilisent pour offrir des services intelligents... services qui se dégradent subitement.

💡Mais au fait : si l'intelligence était au coeur du produit,  est-ce vraiment judicieux de l'externaliser ?  ⚡️

💭Imaginez être un chef étoilé et décider de sous-traiter la préparation des plats... Absurde, non ? 

L'API, c'est le top, à condition que ce ne soit pas le cœur de l'offre. Sinon, c'est jouer à la roulette russe 🎲

Bilan de l'histoire : cela plaide fortement pour l'utilisation de modèles locaux. Moins performants pour le moment, certes, mais sur lesquels vous gardez le contrôle et dont la pérennité est assurée. 

Alors, on opte pour l'API ou on retrousse ses manches ? Hâte de lire vos commentaires ! 👇

-----
*Text drafted through a chat with an AI model. Visual: Stable Diffusion or DALL-E 3.  
Transformation and final editing by Jean-Paul Paoli.*

écouvrez les risques de l'externalisation de l'IA : ChatGPT en baisse de performances ? Maîtrisez les modèles locaux pour garantir qualité et pérennité

ChatGPT perd des neurones

🎂 Joyeux anniversaire ! Voilà le SMS de votre BFF ... ah non c'est son assistant IA programmé pour garder le contact en pilote auto. Heureux ? 😱  

Honnêtement, ça me laisse perplexe 🤨.

Dans un univers où tout est automatisé à la vitesse lumière 💡, l'interaction humaine devient un diamant brut 💎. Certes, l'IA peut envoyer des lettres d'anniversaire au top, mais est-ce vraiment ce que nous aspirons à avoir ?

N'oubliez pas la valeur d'un mot griffonné à la hâte sur une carte d'anniversaire. 🖌 Les choses ont du poids parce que nous, humains, avons pris le temps de les réaliser. ⏳

Alors, chères marques, avant de penser à automatiser vos interactions, faites une petite introspection. 🤔

Peut-être que je suis un peu rétrograde, mais pour moi, rien ne peut remplacer le contact humain. Pas même le robot le plus intelligent de la planète 🌍.

Et vous, qu'en pensez-vous ? Est-ce que l'IA peut vraiment se substituer à l'interaction humaine ? 🗣 À vos claviers ! 👇

-----
*Text drafted through a chat with an AI model. Visual: Stable Diffusion or DALL-E 3.  
Transformation and final editing by Jean-Paul Paoli.*

Découvrez pourquoi l'interaction humaine reste inestimable à l'ère de l'IA. Préservez l'authenticité dans un monde automatisé

Avoir de l'attention: une question de temps ?

🎉 Instagram et son filtre Clarendon ont tué la photo. 😭 Par overdose. Est ce que la Gen AI va faire pareil avec la créa visuelle ? 🎆

J'ai pris cette photo en 1996. 📸 

Sur une plage, une fin de journée d'été. J'ai épuisé une pellicule ou deux à trouver le bon moment où la vague allait se retirer sans emporter les lunettes, à chercher la bonne lumière, le bon angle. 

Aujourd'hui, cette image ne me fait plus rêver. Mais à l'époque, je pouvais être assez fier pour la coller dans un album. Mes amis trouvaient que je faisais de belles photos. 🌟 

Et puis, le digital a débarqué et la photo est devenue un jeu d'enfant.🚀   Au début une formidable démocratisation de la photo. Enfin c'était accessible, enfin on pouvait se permettre de rater.

Très vite, le mobile a pointé son nez. Et Instagram n'a pas tardé.  Et là ça a basculé.

Overdose. 

De quantité.
Et de qualité esthétique : avec le bon filtre, même le plus banal des toilettes devient grandiose.🚽✨

J'ai continué un moment. Avec du matériel lourd et des workflows chronophages, j'ai produit quelques images sympas.  

Mais c'est devenu trop dur. Comment être fier de son image dans cet océan de photos parfaites ? 🌊

La photo était un loisir où on pouvait être moyen, il y a maintenant d'un côté les amateurs qui produisent en série des souvenirs banals, et de l'autre, quelques pros, artistes de très haut niveau. 🎨

Alors je me demande ce que l'IA va faire à la création visuelle. Picasso peut dormir tranquille. Le hic, c'est que tout le monde n'est pas Picasso. 😅

En attendant, pour les praticiens comme pour les marques, il n'y a qu'une seule issue positive : pousser la différence. 🏆  

Fuir ce "look midjourney" qui est le nouveau Clarendon.  Honnêtement, vous trouvez pas qu'on en a déjà marre de ce look ? 👇

-----
*Text drafted through a chat with an AI model. Visual: my own.  
Transformation and final editing by Jean-Paul Paoli.*

L'état, plus rapide que les startup pour adopter l'IA ?

Similar Posts

Pourquoi cultiver un jardin numérique

Mesurer l'intelligence, est-ce important ?

Release de Stable Diffusion XL