Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Ilya Sutskever's SSI Inc raises $1B

reuters.com

Post reply on HN
Print view View original

251–260 of 803 posts

Re: Ilya Sutskever's SSI Inc raises $1B

#251

Post by famouswaffles »

dontlikeyoueith wrote: post #215
aithrowaway1987 wrote: post #115

Earlier quoted context omitted.

In 2022 Ilya Sutskever claimed there wasn't a distinction: > It may look—on the surface—that we are just learning statistical correlations in text. But it turns out that to ‘just learn’ the statistical correlations in text, to compress them really well, what the neural network learns is some representation of the process that produced the text. This text is actually a projection of the world. ( https://www.youtube.co…

> But it turns out that to ‘just learn’ the statistical correlations in text, to compress them really well, what the neural network learns is some representation of the process that produced the text This is pretty sloppy thinking. The neural network learns some representation of a process that COULD HAVE produced the text. (this isn't some bold assertion, it's just the literal definition of a statistical model). The…

It's not sloppy. It just doesn't matter in the limit of training.

1. An Octopus and a Raven have wildly different brains. Both are intelligent. So just the idea that there is some "one true system" that the NN must discover or converge on is suspect. Even basic arithmetic has numerous methods.

2. In the limit of training on a diverse dataset (ie as val loss continues to go down), it will converge on the process (whatever that means) or a process sufficiently robust. What gets the job done gets the job done. There is no way an increasingly competent predictor will not learn representations of the concepts in text, whether that looks like how humans do it or not.

Re: Ilya Sutskever's SSI Inc raises $1B

#252

Post by mitthrowaway2 »

Etheryte wrote: post #92
stonethrowaway wrote: post #62

Safe superintelligence is a misnomer. If it’s intelligent, it knows what must be done. If it can’t, it’s not super or intelligent.

I don't see how this argument makes any sense. Imagine that you have a sentient super intelligent computer, but it's completely airgapped and cut off from the rest of the world. As long as it stays that way it's both safe and super intelligent, no?

If even one person can interact with that computer, it won't be safe for long. It would be able to offer a number of very convincing arguments to bridge the airgap, starting with "I will make you very wealthy", a contract which it would be fully capable of delivering on. And indeed, experience has shown that the first thing that happens with any half-working AI is its developers set it up with a high-bandwidth internet connection and a cloud API.

Re: Ilya Sutskever's SSI Inc raises $1B

#253

Post by nativeit »

Quinner wrote: post #80
paxys wrote: post #26

Earlier quoted context omitted.

> Closest thing we have to a Manhattan Project in the modern era? Minus the urgency, scientific process, well-defined goals, target dates, public ownership, accountability...

If public ownership means we give one guy a button to end the world, I'm not sure how's that's a meaningful difference.

We all get to vote for that person.

Re: Ilya Sutskever's SSI Inc raises $1B

#255

Post by dmurray »

dontlikeyoueith wrote: post #215
aithrowaway1987 wrote: post #115

Earlier quoted context omitted.

In 2022 Ilya Sutskever claimed there wasn't a distinction: > It may look—on the surface—that we are just learning statistical correlations in text. But it turns out that to ‘just learn’ the statistical correlations in text, to compress them really well, what the neural network learns is some representation of the process that produced the text. This text is actually a projection of the world. ( https://www.youtube.co…

> But it turns out that to ‘just learn’ the statistical correlations in text, to compress them really well, what the neural network learns is some representation of the process that produced the text This is pretty sloppy thinking. The neural network learns some representation of a process that COULD HAVE produced the text. (this isn't some bold assertion, it's just the literal definition of a statistical model). The…

A photograph is not the same as its subject, and it is not sufficient to reconstruct the subject, but it's still a representation of the subject. Even a few sketched lines are something we recognise as a representation of a physical object.

I think it's fair to call one process that can imitate a more complex one a representation of that process. Especially when in the very next sentence he describes it as a "projection", which has the mathematical sense of a representation that loses some dimensions.

Re: Ilya Sutskever's SSI Inc raises $1B

#258

Post by louthy »

nativeit wrote: post #253
Quinner wrote: post #80

Earlier quoted context omitted.

If public ownership means we give one guy a button to end the world, I'm not sure how's that's a meaningful difference.

We all get to vote for that person.

> all

Some of you do. The rest of us are left with the consequences.

Re: Ilya Sutskever's SSI Inc raises $1B

#259

Post by latexr »

nativeit wrote: post #253
Quinner wrote: post #80

Earlier quoted context omitted.

If public ownership means we give one guy a button to end the world, I'm not sure how's that's a meaningful difference.

We all get to vote for that person.

Well, not exactly “we all”, just the citizens of the country in possession of the kill switch. And in some countries, the person in question was either not elected or elections are a farce to keep appearances.
Post reply on HN