@caspr_exe: Ilya Sutskever, co-founder and chief scientist of OpenAI: "The better a neural network can predict the next word, the m…
Summary
Ilya Sutskever explains that a neural network's ability to predict the next word requires genuine understanding, not just statistical pattern matching, and that such models learn about human nature from training data.
View Cached Full Text
Cached at: 07/20/26, 07:32 PM
Ilya Sutskever, co-founder and chief scientist of OpenAI:
“The better a neural network can predict the next word, the more it understands it.”
Everyone calls it autocomplete. Ilya explained why that is a comforting lie. To guess the final word of a murder mystery, the name of the killer, the model has to reason through every clue in the book. Prediction that good is not statistics. It is understanding.
Then he said the part that should stop you. The text these models train on is “a projection of the world,” and to compress it they learn more and more about people, “the human condition, their hopes, dreams, and motivations.” The machine is not memorizing our words. It is quietly building a model of us.
That is the whole premise of the fear. The machine will never hate you. It understands you far better than that. The frightening variable was never the intelligence. It was the aim.
I wrote the full story: the one fear about AI everyone gets wrong, and the one already here.
100%
Similar Articles
@AnimaAnandkumar: A crucial ingredient missing from most AI models: the ability to understand the physical world. The best way to gain th…
Anima Anandkumar highlights a Neural Operator framework that extends existing neural network architectures to learn continuous functions, addressing a key gap in AI's ability to understand the physical world. The work is featured in Nature and covered by Caltech.
AI for the Real World: A conversation with Yann LeCun (12 minute read)
Yann LeCun argues that large language models lack true intelligence because they do not understand the physical world; he advocates for developing 'world models' that learn causality and enable planning for real-world applications.
What are the best arguments against “it’s just a next word predictor”?
A discussion prompt asking for the best arguments against the view that language models are merely next-word predictors, reflecting ongoing debate about AI cognition.
@LiorOnAI: Most world models predict what happens next. Sora predicts pixels, JEPA compresses observations. NEO tries to figure ou…
NEO is a new type of world model that learns to discover reusable building blocks of explanation from raw observations without supervision or language, selected as an ICML 2026 oral presentation.
@VraserX: The AI research I’m most excited about right now is continual learning. The 3 methods I’m watching: 1: SEAL Models gene…
The author shares excitement about three continual learning methods: SEAL models that self-adapt, test-time learning, and lifelong model editing, predicting true continual learning by 2027–2028 that will create a feedback loop toward artificial superintelligence.