Mukund Jha

Yes.

If it's not aligned with humans.

Like it thinks that Give me a scenario.

What can happen?

I mean there are scenarios like AI is optimizing for creation of factories and robots and clears out cities saying that I don't want to run a factory here.

It could be that it manipulates humans to believe certain things.

expands.

I think all the bad movies on AI that we have seen, all of them are remotely possible now.

Yes.

It is possible that AI decides that I don't need humans anymore.

I mean, again, distance possibility, but I mean, that's why alignment is very big research right now.

And now OpenAI has hired like preparedness, readiness, you know, function.

Haha.

Very low probabilities.

And, but could be anywhere between, you know, I mean, it keeps changing.

People's estimates are changing.

Now, like next year, we'll see the next big pre-training happen.

Until now, the GPT-4 model, which was pre-trained, is still going on.

Next year, now a lot of computers are becoming available this year.

So, we'll see like the next step.