Ilya Sutskever on AIs understand

Ilya Sutskever

AI researcher, co-founder and former former chief scientist at OpenAI

I'd like to take a small detour and to give an analogy that will hopefully clarify why more accurate prediction of the next word leads to more understanding, real understanding. Let's consider an example. Say you read a detective novel, [...] then let's say that at the last page of the book, the detective has gathered all the clues, gathered all the people, and saying, "Okay, I'm going to reveal the identity of whoever committed the crime, and that person's name is.." Predict that word.

AI Verified source (Mar 22, 2023)

3mo ago

Policy proposals and claims

AIs understand

votes For

Statement relation verification history AI Verified Report this

AI Verified The quote supports the idea that next-word-predicting systems can achieve "real understanding" ("more accurate prediction of the next word leads to more understanding, real understanding"), which clearly implies support for the claim that AIs such as language models understand. · YouCongress gpt-5.4-2026-03-05 · 3mo ago

Vote inference verification history AI Verified Report this

AI Verified The quote explicitly says that better next-word prediction leads to "more understanding, real understanding." Since next-word prediction is the ability associated with language-model AIs, the ordinary implication is that such AIs do understand, though the quote argues this by analogy rather than naming AI directly. · YouCongress gpt-5.4-2026-03-05 · 3mo ago

Quote authenticity verification history

Report this

AI Verified NVIDIA’s indexed transcript for the exact session URL contains the passage at about 28:04–28:45, including “I’d like to take a small detour …” and “Say you read a detective novel … Predict that word,” and the surrounding turns show it is Ilya Sutskever answering Jensen Huang. The session page names both speakers, and official NVIDIA pages place this fireside chat on March 22, 2023, so the stored author, source URL, and date are consistent. ([nvidia.com](https://www.nvidia.com/en-us/on-demand/session/gtcspring23-s52092/?utm_source=openai)) · YouCongress gpt-5.4-2026-03-05 · 3mo ago

replying to Ilya Sutskever

Other opinions from Ilya Sutskever

See all

Ilya Sutskever votes For

Pausing AI at human-genius level to solve alignment is safer than either racing to superintelligence or halting entirely

Future AI will be extraordinarily powerful compared to anything that exists today, and dealing with this future power will require unprecedented measures, such as the ones described here. The problem statement is real. This works only if it is done i...

more AI Verified source (Jul 2026)

added 1mo ago

Ilya Sutskever votes For

AGI could quickly lead to superintelligence

Future AI will be extraordinarily powerful compared to anything that exists today, and dealing with this future power will require unprecedented measures, such as the ones described here. The problem statement is real. This works only if it is done i...

more AI Verified source (Jul 2026)

added 1mo ago

Ilya Sutskever votes For

Ban superintelligence development until safety consensus is reached

Future AI will be extraordinarily powerful compared to anything that exists today, and dealing with this future power will require unprecedented measures, such as the ones described here. The problem statement is real. This works only if it is done i...

more AI Verified source (Jul 2026)

added 1mo ago