@ylecun is Meta's Chief AI Scientist and Turing Award winner. For 35 years, he's been right about every major AI breakthrough when everyone else was wrong. He championed neural networks during the "AI winter." But his new predictions are his boldest yet...

1. "Nobody in their right mind will use autoregressive LLMs a few years from now." The technology powering ChatGPT and GPT-4? Dead within years. The problem isn't fixable with more data or compute. It's architectural. Here's where it gets interesting...

Every token an LLM generates compounds tiny errors exponentially. The longer the output, the higher the probability of hallucination. This is why ChatGPT makes up facts. Why scaling won't save current models. Mathematical certainty. But LeCun didn't stop there:

2. Video-based AI will make text training primitive LeCun's calculation: A 4-year-old processes 10¹⁴ bytes through vision alone. That equals ALL the text used to train GPT-4. In 4 years. Through one sense. This changes everything about how AI should learn:

Babies learn gravity and physics by 9 months. Before they speak. "We're never going to get human-level AI unless systems learn by observing the world." Companies building video-first AI will leapfrog text-based systems. Here's what Meta is secretly building:

3. Proprietary AI models will "disappear" LeCun's exact words: "Proprietary platforms, I think, are going to disappear." He calls it "completely inevitable." OpenAI's closed approach? Google's secret models? All doomed. His reasoning will shock the industry:

"Foundation models will be open source and trained in a distributed fashion." A few companies controlling our digital lives? "Not good for democracy or anything else." Progress is faster in the open. The world will demand diversity and control. LeCun's timeline will surprise

4. AGI timeline is 2027-2034 @ylecun's exact words: "3-5 years to get world models working. Then scaling until human-level AI... within a decade or so." But it won't come from scaling LLMs.

Every company betting only on GPT-style scaling will be blindsided. LeCun calls the "country of geniuses in a data center" idea "complete nonsense." The smart money is repositioning for the architecture shift.

5. AI assistants replace all digital interfaces Ray-Ban Meta glasses: Look at Polish menu, get translation. Ask about plants, get species ID. That's primitive compared to what's coming. AI will mediate ALL digital interactions. Here's what this means for your business:

The economic implications are massive. Companies building on OpenAI APIs could see foundations crumble in 3-5 years. But early movers positioning for JEPA? They'll capture the next $10 trillion wave. LeCun's advice for surviving this transition:

How to prepare: Researchers: "Don't work on LLMs. Focus on world models and sensory learning." Companies: Build on open-source foundations like PyTorch and Llama. When the shift happens, you adapt instantly. The window to position yourself is closing:

LeCun's warning reveals the hidden opportunity: As companies abandon LLMs for world models, they're creating a massive validation gap. These new architectures aren't just different - they're fundamentally harder to monitor and govern.

While everyone's racing to build next-generation AI, the smart money is positioning for what makes them trustworthy. The companies that survive this transition won't just have better models. They'll have the governance frameworks to validate them at scale.

In a world where AI shapes every business decision, trust isn't optional. It's the only competitive advantage that matters. And there's one thing that builds AI trust faster than anything else:

Proper model validation and governance. Are you an Enterprise AI Leader looking to validate and govern your AI models at scale? http://TrustModel.ai provides the model validation, monitoring, and governance frameworks you need to stay ahead. Learn more:

Thanks for reading. If you enjoyed this post, follow @karlmehta for more content on AI safety. Repost the first tweet to help more people see it: Appreciate the support.

@karlmehta @grok without your own bias, is there any truth to this? How can I find more info

@karlmehta tl;dr no agi. meh ai is dead and this new ai is not possible because computers compute 1's and 0's on silicone using energy at a man made frequency. they will never smell or live in our physical reality.

@karlmehta Input modality is meaningless to the relationship between Information and Intelligence. Visual information is high data but low signal. Lots of noise, low informational content relative to the data set size. Any frame of video contains almost all the information of the previous

@karlmehta @threadreaderapp unroll

@karlmehta He is both right and wrong. Could be right about that new architecture is likely and needed but wrong about LLM technology. It will bring incredible value over the next 10 years and will reshape industries. even with what we have today. and there will be many competing

@karlmehta @grok what are the hardware requirements for World Models, what level of scaling will need to occur?

@karlmehta Can you give a concrete example of what you mean by the following: “But early movers positioning for JEPA? They'll capture the next $10 trillion wave.”

@karlmehta Did he also convince Zuck to make him report to a 28 yo kid?

@karlmehta Unpopular opinion, he's 100% right across the board. LLMs feel like a hack to get AI working. It just doesn't "smell" like the right architecture.

@karlmehta And he got demoted because he’s been wrong in the medium term and he’s not Chinese enough 🤷‍♂️

@karlmehta But everyone is already switching to world models, from google deep minds genie 2, to x training on videos and games, to the rumoured big breakthrough for OpenAI coming from having their ai play endless games. All the SOTA models are already multi modal and not just LLMs anymore.

@karlmehta This guy’s track record is wild. Always good to stay curious and be ready for big shifts; AI’s moving fast and sometimes hype gets ahead of reality, but prepping now never hurts.

@karlmehta This is just Walter Bishop in the Fringe prequel series, no?

@karlmehta The difference between prediction and speculation is when speculation comes to pass. Just sprout theories, one of it will eventually happen.

@karlmehta LeCun's track record is legendary. The real question isn't what he's predicting - it's how fast we can position ourselves before everyone else catches on. That $20B redirect just validated what smart builders already knew. 🧠

@karlmehta he got demoted bc Meta has lagged in AI & now Zuck is forced to open up the checkbook to try to make up ground.

@karlmehta I predicted Neuralink in 1988

@karlmehta But did he have the Chiefs for thr win?

@karlmehta Boring