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

211–220 of 803 posts

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

#211

Post by zeroonetwothree »

davedx wrote: post #70
Yizahi wrote: post #47

Earlier quoted context omitted.

There is significant possibility that true AI (what Ilia calls superintelligence) is impossible to build using neural networks. So it is closer to some tokenbro project than to nuclear research. Or he will simply shift goalposts, and call some LLM superintelligent.

> There is significant possibility that true AI (what Ilia calls superintelligence) is impossible to build using neural networks What evidence can you provide to back up the statement of this "significant possibility"? Human brains use neural networks...

For any technology we haven’t achieved yet there’s some probability we never achieve it (say, at least in the next 100 years). Why would AI be different?

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

#213

Post by petulla »

hn_throwaway_99 wrote: post #200

Lots of comments either defending this ("it's taking a chance on being the first to build AGI with a proven team") or saying "it's a crazy valuation for a 3 month old startup". But both of these "sides" feel like they miss the mark to me. On one hand, I think it's great that investors are willing to throw big chunks of money at hard (or at least expensive) problems. I'm pretty sure all the investors putting money in…

Ilya has discussed this question: https://www.youtube.com/watch?v=YEUclZdj_Sc

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

#214

Post by dartos »

jejeyyy77 wrote: post #198
dartos wrote: post #109

Earlier quoted context omitted.

All I’m saying is you used the word “if” a lot there. AGI assumes exponential, preferably infinite and continuous improvement, something unseen before in business or nature. Neither siri nor Alexa were sold as AGI and neither alone come close to a $1B product. gpt and other LLMs has quickly become a commodity, with AI companies racing to the bottom for inference costs. I don’t really see the plan, product wise. Moreo…

"if" is the name of the game in investing. you say you don't see it. fine. these investors do - thats why they are investing and you are not.

You should read the entire comment.

They also have the warchest to afford a $1B gamble.

If the math worked out for me too, I’d probably invest even if I didn’t personally believe in it.

Also investors aren’t super geniuses, they’re just people.

I mean look at SoftBank and Adam Neuman… investors can get swept up in hype and swindled too.

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

#215

Post by dontlikeyoueith »

aithrowaway1987 wrote: post #115
HarHarVeryFunny wrote: post #41

Earlier quoted context omitted.

I wonder if "Super Intelligence" means anything .. just LLMs, or maybe they are pursuing new architectures and shooting for AGI ?

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).

There is no guarantee it is the same as the actual process. A lot of the "bow down before machine God" crowd is guity of this same sloppy confusion.

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

#216

Post by aithrowaway1987 »

WithinReason wrote: post #203
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…

> newer LLMs appear to be great at arithmetic, but they still fail basic counting tests How does the performance of today's LLMs contradict Ilya's statement?

Because they can learn a bunch of symbolic formal arithmetic without learning anything about quantity. They can learn

  5 x 3 = 15
without learning

  *****    ****     *******
  ***** =  *****  = *******
  *****    ******   *
And this generalizes to almost every sentence an LLM can regurgitate.

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

#217

Post by zeroonetwothree »

liminvorous wrote: post #57
Yizahi wrote: post #47

Earlier quoted context omitted.

There is significant possibility that true AI (what Ilia calls superintelligence) is impossible to build using neural networks. So it is closer to some tokenbro project than to nuclear research. Or he will simply shift goalposts, and call some LLM superintelligent.

No one had built a nuclear bomb before the Manhattan project either.

this is not evidence in favor of your position. We could use this to argue in favor of anything such as “humans will eventually develop time travel” or “we will have cost effective fusion power”.

The fact is many things we’ve tried to develop for decades still don’t exist. Nothing is guaranteed

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

#218

Post by teqsun »

We don't even understand how the brain functions completely, not even close. Until we have a complete understanding of how our own GI works down to the exact bio-mechanical level, we can't achieve AGI.

That's the theoretical basis and path for achieving AGI (if it's even possible). I'm tired of all the "we stick enough data in the magic black box blender and ta-da! AGI!"

Every giant technological break-through throughout history has had a massive underpinning of understanding before ever achieving it. And yet, with the AI bubble somehow we're just about to secretly achieve it, but we can't tell you how.

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

#219

Post by zeroonetwothree »

logicchains wrote: post #112
Yizahi wrote: post #72

Earlier quoted context omitted.

Theoretical foundation was slowly built over decades before it started though. And correct me if I'm wrong, but calculations that it was feasible were present before the start too. They had to calculate how to do it, what will be the processes, how to construct it and so on, but theoretically scientists knew that this amount of material can start such process. On the other hand not only there is no clear path to AI t…

The theoretical foundation of transformers is well understood; they're able to approximate a very wide family of functions, particularly with chain of thought ( https://arxiv.org/abs/2310.07923 ). Training them on next-token-prediction is essentially training them to compress, and more optimal compression requires a more accurate model of the world, so they're being trained to model the world better and better. Howev…

The disagreement here seems merely to be about what we mean by “AGI”. I think there’s reasons to think current approaches will not achieve it, but also reason to think they will.

In any case anyone who is completely sure that we can/can’t achieve AGI is delusional.

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

#220

Post by dartos »

kwant_kiddo wrote: post #194
sirspacey wrote: post #85

Lots of dismissive comments here. Ilya proved himself as a leader, scientist, and engineer over the past decade with OpenAI for creating break-through after break-through that no one else had. He’s raised enough to compete at the level of Grok, Claude, et al. He’s offering investors a pure play AGI investment, possibly one of the only organizations available to do so. Who else would you give $1B to pursue that? That’…

I repeatedly keep seeing praise for Ilyas achievements as a scientist and engineer, but until ChatGPT OpenAI was in the shadow of DeepMind, and to my knowledge (I might be wrong) he has not been that much involved with ChatGPT? the whole LLM race seems deaccelerate, and all the hard problems about LLMs seems not do have had that much progress the last couple of years (?) In my naaive view I think a guy like David Sil…

I’m not optimistic about AGI, but it’s important to give credit where credit is due.

Even assuming the public breakthroughs are the only ones that happened, the fact that openai was able to make an llm pipeline from data to training to production at their scale before anyone else is a feat of research and engineering (and loads of cash)

Post reply on HN