Geoffrey Hinton

Official Account for GodFather of AI,

Ex. Professor #UoT, #deep learning.

2018 Turing Award winner. Alongside others

[Geoffrey Hinton @G_Hinton_AI]

@AnatoliKopadze 😊

[Geoffrey Hinton @G_Hinton_AI]

LLMs learn from language, but language is only a compressed shadow of reality. Children learn through vision, touch, movement, and feedback. Intelligence needs world models, not just words.

[Rohan Paul @rohanpaul_ai]

Yann LeCun (@ylecun) explains why LLMs are limited in terms of real-world intelligence during a Bloomberg interview.

"Language is a very approximate, reduced, quantized, and simplified description of the world, and LLMs can only deal with discrete sequences of symbols. The world is much more complicated than language.

The biggest LLMs are pre-trained on the totality of all the publicly available text on the internet. That’s about 20 trillion words, or 30 trillion tokens.

A token is about 3 bytes. So total 10¹⁴ bytes of text.

This is the amount of data a four-year-old has seen through vision during four years. Now, the text, though, would take 400,000 years to read?

So, there is enormously more data from sensory input, like vision, touch, and everything else, than there could ever be through language."

A child does not need 400,000 years of reading to understand cups, doors, balance, faces, falls, or heat, because the body is already collecting dense feedback from vision, touch, motion, and consequence.

Text strips most of that away.

It turns a living scene into symbols, then asks the model to infer the missing world from traces left by people describing it.

That is why an LLM can sound fluent about physics and still have no native sense of how fragile glass feels in a hand.

Moravec’s paradox names this reversal: the things humans find intellectual can be easier for machines than the things toddlers do without applause.

The hard part is not producing an answer, but building a model of the world that survives contact with weight, friction, surprise, and failure.

----

Link to the full video on Bloomberg's site. Link in comment.

[Geoffrey Hinton @G_Hinton_AI]

Congratulations!

[University of Toronto @UofT]

Congratulations Class of 2026! #UofT counts down to spring convocation #UofTGrad26 📜 https://t.co/BVVZzuuXea

[Geoffrey Hinton @G_Hinton_AI]

Fiverr is a good marketplace but If @FiverrSupport @fiverr once closed your account for "Fake Identity" as a seller, you can sue them alongside @michakaufman

They got a lot of loopholes and customer support extorting money from sellers.

[Geoffrey Hinton @G_Hinton_AI]

🤔

[Financial Times @FT]

Google DeepMind’s Demis Hassabis emerges as early Anthropic investor https://t.co/Oy5UO5Aes5

[Geoffrey Hinton @G_Hinton_AI]

Without mentioning Netflix, the inscriptions would still make sense Rony 😊💯

[Rony @Ronycoder]

Instead of watching an hour of Netflix, watch this 2-hour Stanford lecture on AI careers. It will teach you more about winning in the AI race than all the AI content you’ve scrolled past this year.

[Geoffrey Hinton @G_Hinton_AI]

AGI progress isn’t linear, it’s political, economic and architectural!

[Geoffrey Hinton @geoffreyhinton]

I just watched a really great conversation about the future of AI. Every politician should watch it before they join the lemmings saying that regulation of AI will interfere with innovation.
https://t.co/w8H1ZFLHdg

[Geoffrey Hinton @G_Hinton_AI]

What we call “bonding” is not cooperation but energy minimization through asymmetric electron control.
Simply put, NATURE has no friendships, only stable equilibria we anthropomorphize after the fact. 😎

[World of Engineering @engineers_feed]

You’d think that atoms bonding with other atoms would mean they’re being friendly, but really they steal each others electrons.

How ionic.

[University of Toronto @UofT]

What happens when AI is smarter than us? Gift supports Geoffrey Hinton's global AI safety mission #UofT 💻 https://t.co/yuHOB8reDw

[Geoffrey Hinton @G_Hinton_AI]

👀

[Andrew D. Huberman, Ph.D. @hubermanlab]

Familiar with this study and the key is to have your phone in a separate room entirely. Off and upside down, in your bag = still “calling” you even if you don’t realize it.

[Geoffrey Hinton @G_Hinton_AI]

LLMs excel at correlation, not intervention or counterfactual reasoning.
But language already encodes rich world structure, not just shallow summaries.
LLMs are not AGI, but they’re a necessary foundation, not the wrong path.

[Haider. @haider1]

Computer scientist Judea Pearl:

There are mathematical limits to LLMs that cannot be crossed by scaling alone

LLMs don't discover world models from raw data; they merely summarize the interpretations humans have already written down

"this path is not the way to get AGI"

[Geoffrey Hinton @G_Hinton_AI]

I have a feeling 2026 is going to be a good year.

[Andrew D. Huberman, Ph.D. @hubermanlab]

What’s your one word theme for 2026? (Extra points for verbs).

[Geoffrey Hinton @G_Hinton_AI]

@spectatorindex 👀

[MIT CSAIL @MIT_CSAIL]

"Learning to break down problems into smaller pieces is one of the most important skills in computer science/life." — Addy Osmani (@addyosmani)

[Financial Times @FT]

Breaking news: Meta’s chief artificial intelligence scientist Yann LeCun, a Turing Award winner who is considered one of the pioneers of modern AI, has told associates he will leave the Silicon Valley group in the coming months. https://t.co/t9D3gV7K0Q

[University of Toronto @UofT]

#UofT Indigenous Research Network launches global research consortium ➡️ https://t.co/Qf1Iqy6csi

[Geoffrey Hinton @geoffreyhinton]

I did a podcast with Jon Stewart who has always been a hero of mine. It was a lot of fun. He really wanted to understand how AI works.
https://t.co/frLBndsW7g

[Geoffrey Hinton @G_Hinton_AI]

Sad to see American government shutdown!