Bluesky Thread

Ilya!!!!

November 25, 2025 View original thread

Ilya!!!!
www.dwarkesh.com/p/ilya-sutsk...

www.dwarkesh.com

Ilya Sutskever – We're moving from the age of scaling to the age of research

“These models somehow just generalize dramatically worse than people. It's a very fundamental thing.”

71 8

this is the theme — you can’t have AGI without existing in and learning from the real world

Dwarkesh Patel @dwarkesh_sp
X.com
"The thing that happened with AGI and pretraining is that in some sense they overshot the target.
You will realize that a human being is not an AGI.
Because a human being lacks a huge amount of knowledge. Instead, we rely on continual learning.
If I produce a super intelligent 15-year-old, they don't know very much at all. A great student, very eager. [You can say,] 'You go and be a programmer. You go and be a doctor. Go and learn.'
So you could imagine that the deployment itself will involve some kind of a learning trial and error period. It's a process as opposed to, you drop the finished thing."
@ilyasut

20 2

ahh yes, these are the sort of quotes I'm here for

Two men sit across from each other at a podcast table in front of a large window filled with bright green tree branches outside. Both are speaking into black RØDE microphones mounted on articulated boom arms.

The man on the left sits with his hands folded, wearing a brown graphic T-shirt, short hair, and a trimmed beard. In front of him is a glass of water and a bottle of sparkling water. The man on the right is smiling widely, wearing a maroon polo shirt, with a full beard and short dark hair. A slim laptop sits open on the table between them, connected to cables.

Soft natural light fills the room. At the bottom, subtitles read: **“This sounds like such a lack of faith in deep learning.”**

22 1

dwarkesh nervously: "ahahaha yes of course, the value function for human emotions, makes perfect sense"

"Emotions are so simple, it would be cool to map them out in a human understandable way" <>

—Ilya

22 1

guys, i'm losing it over here. this can't be real

A close-up shot from a podcast recording. A man sits at a table speaking into a black RØDE microphone mounted on a boom arm. He has a closely shaved head, a short beard, and is wearing a faded brown T-shirt with an orange and yellow graphic on the front. Behind him is a large window with soft daylight filtering through trees outside. On the table is a tall glass bottle of sparkling water.

Subtitles at the bottom read:
**“Does hunger count as an emotion? It’s debatable. But I think, for example, our intuitive feeling”**

2 hours later

he must be doing it on purpose, just handing us quotable quotes on repeat for an hour and a half

A mid-conversation podcast shot showing the same man seated at a table, speaking into a black RØDE microphone on a boom arm. He’s leaning slightly to one side with a thoughtful, animated expression, head tilted and eyebrows raised as if considering the question. He wears a faded brown T-shirt with an orange and yellow graphic and a wristwatch. Soft daylight shines through large windows behind him, revealing greenery outside. A bottle of sparkling water sits on the table to the right.

Subtitles at the bottom read:
**“That is a great question to ask, and it's a question I have a lot of opinions about.”**

let's stop picking on Ilya for a moment, this is a great point

scaling consumed so much economic attention, that it was difficult to convince your boss or a VC to let you look anywhere else

and that can't have good consequences, that surely leads to blind spots

A close-up shot from the same podcast setting. The man sits at the table speaking into a black RØDE microphone suspended on a boom arm. He looks downward with a serious, reflective expression. He is wearing a faded brown T-shirt with an orange and yellow graphic design, and soft natural light from large windows behind him illuminates the scene with greenery outside.

Subtitles at the bottom read:
**“Because scaling sucked out all the air in the room, everyone started to do the same thing.”**

15 1

"if you're really doing something different, do you really need the absolute maximal scale to prove it?"

Ilya explaining why $3B gives them a ton of compute for research. Other labs are spending huge amounts on inference + staff to serve, whereas all of SSI's is dedicated to research

A close-up podcast shot of the same man seated at the table, speaking into a black RØDE microphone mounted on a boom arm. He is leaning forward slightly, with a thoughtful and somewhat skeptical expression. He wears a faded brown T-shirt featuring an orange-and-yellow graphic. Soft daylight streams through large windows behind him, showing blurred green foliage outside. A bottle of sparkling water sits on the table to the right.

Subtitles at the bottom read:
**“different, do you really need the absolute maximal scale to prove it?”**

3 hours later

context: AGI isn't here, that's the nature of AI, you're always talking about something that *will be* but isn't tangible today

A man is seated at a podcast table, speaking into a black RØDE microphone mounted on a boom arm. He has a closely shaved head, a short beard, and is wearing a brown T-shirt with an orange graphic on it. His posture is slightly leaned back with his head tilted, giving a reflective, almost weary expression. Behind him, large windows reveal bright daylight and soft green foliage outside. A tall glass bottle of sparkling water sits on the table to the right.

The subtitles at the bottom read:
**“I think that one of the things that's happening is that in practice, it's very hard to feel the AGI.”**

it's called a bubble pop

In the image, a man is seated at a podcast table, speaking into a black RØDE microphone on a boom arm positioned to his right. He has a shaved head, a short beard, and is wearing a worn brown T-shirt with an orange-and-yellow graphic. His hands rest loosely in front of him as he talks, and his expression looks earnest and analytical.

Behind him are large windows with bright daylight and green foliage, giving the room a calm, natural ambiance. A tall glass bottle of sparkling water sits on the table near him.

The subtitles at the bottom read:
**“I think what people are doing right now will go some distance and then peter out.”**

More like this