Dwarkesh: When do we get AI that beats top human experts at all computer-based cognitive work? Beren: I would say 3–4 years. Dwarkesh: THE FUCK? 😭

the mechanical engineers will be graced with an additional few years of relevancy.

that's John Schulman, not Beren!

yes thanks for the correction

I can't believe people are still this pessimistic when people have been saying "AGI by 2030" in 2025 and now we will get it in 2026...

Bs, no more than 6-9 months at this rate. Just look at Astra, go figure with Bel

that long???

The annoying thing with such predictions is that we are not sure what is right anymore. Is the event horizon for ASI nearer or farther? Does it even exist? The argument from both sides sounds equally plausible . Even a random guess have >0.5 chance of being true

yes, but still needs humans driving it

That’s Schulman

Well it's quite closer than expected , but at what cost and efficiency can be a real question to answer.

that was john

This is what happens when you no longer work at a frontier lab, you are just as in the dark as any of us. Clearly they are not updated to what is happening, I also imagine this was filmed before the Navier Stokes breakthrough.

I’d give it by the end of 2027! People really don’t understand exponentials do they🤭🤫😇

Not beren.

I just don’t believe these predictions. They have been saying that AI is going to destroy the entire world for some time. What I try to put in my mind is we will have great advancements, we will need to adapt, and it will be less than what these AI gurus are predicting.

I'd say 5 to 10 years