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

761–770 of 803 posts

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

#761

Post by czl »

trashtester wrote: post #756
czl wrote: post #742

Earlier quoted context omitted.

> Evolutionary history points towards brain size being limited by a combination of what is safe for the female hip width and also what amount of energy cost can be justified by increasing the size of the brain. If bigger brains lead to higher intelligence, why do many highly intelligent people have average-sized heads? And do they need to eat much more to fuel their high IQs? If larger brains were always better, woul…

> If bigger brains lead to higher intelligence, why do many highly intelligent people have average-sized heads? There IS a correlation between intelligence and brain size (of about 0.3). But the human brain does a lot of things apart from what we measure as "IQ". What shows up in IQ tests tend to be mostly related to variation of the thickness of certain areas of the cortex [1]. The rest of the brain is, however, res…

> There IS a correlation between intelligence and brain size (of about 0.3).

Here is a page that shows what such a weak correlation looks like visually:

https://resources.nu.edu/statsresources/correlation

It also explains that a correlation of r = 0.3 means only about 9% of the variability in one variable is explained by the other. This makes me wonder: can intelligence really be estimated within 10% accuracy? I doubt it, especially considering how IQ test results can vary even for the same person over time.

Kids and teens have smaller brains, but their intelligence increases as they experience more mental stimulation. It’s not brain size that limits them but how their brains develop with use, much like how muscles grow with exercise.

> a chimp brain is not likely to produce (adult, healthy) human-level intelligence. > Human brains consume significantly more calories than chimp brains.

If brain size and calorie consumption directly drove intelligence, we’d expect whales, with brains five times larger than humans, to be vastly more intelligent. Yet, they aren’t. Whales’ large brains are likely tied to their large bodies, which evolved to cover great distances in water.

Brains can large like arms can be large but big arms do not necessarily make you strong -- they may be large due to fat.

> But generally, over the last several million years, human intelligence has been rising sharply.

Yes, smaller-brained animals are generally less intelligent, but exceptions like whales and crows suggest that intelligence evolves alongside an animal’s ecological niche. Predators often need more intelligence to outsmart their prey, and this arms race likely shaped human intelligence.

As humans began living in larger communities, competing and cooperating with each other, intelligence became more important for survival and reproduction. But this has limits. High intelligence can lead to emotional challenges like overthinking, isolation, or an awareness of life’s difficulties. Highly intelligent individuals can also be unpredictable and harder to control, which may not always align with societal or biological goals.

As I see it, ecological niche drives intelligence, and factors like brain size follow from that. The relationship is dynamic, with feedback loops as the environment changes.

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

#762

Post by HarHarVeryFunny »

famouswaffles wrote: post #754
dontlikeyoueith wrote: post #725

Earlier quoted context omitted.

> whether that looks like how humans do it or not. So you agree with me that there is no guarantee it learns any representation of the actual process that produced the training data.

Sure I agree. But if that's what you're getting hung up on, i think you've missed his point entirely. Whether the machines becomes a human brain clone or something entirely alien is irrelevant. The point is, you can't cheat reality. Statistics is not magic. You can't predict text that understands without understanding.

Sure you can, and if your predictive engine doesn't have the generality and power of the original generative one, then you have no choice.

Machine learning isn't magic - the model will learn what it can to minimize the error over the specific provided loss function, and no more. Change the loss function and you change what the model learns.

In the case of an LLM trained with a predict next word loss function, what you are asking/causing the model to learn is NOT the generative process - you are asking it to learn the surface statistics of the training set, and the model will only learn what it needs to (and is able to, per the model architecture being trained) in order to do this.

Now of course learning the surface statistics well does necessitate some level of "understanding" - are we dealing with a fairy tale or a scientific paper for example, but there is only so much the model can do. Chess is a good example, since it's easy to understand. The generative process for world class chess (whether human, or for an engine) involves way more DEPTH (cf layers) of computation than the transformer has available to model it, so the best it can do is to learn the surface statistics via much shallower pattern recognition of the state of the board. Now, given the size of these LLMs, if trained on enough games they will be able to play pretty well even using this pattern matching technique, but one doesn't need to get too far into a chess game to reach a position that has never been seen before in recorded games (e.g. watch agadmator's YouTube chess channel - he will often comment when this point has been reached), and the model therefore has no choice but to play moves that were seen in the training set in similar, but not identical positions... This is basically cargo-cult chess! It's interesting that LLMs can reach the ELO level that they do (says more about chess than about LLMs), but this same "cargo-cult" (follow surface statistics) generation process when out of training set applies to all inputs, not just chess...

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

#763

Post by HarHarVeryFunny »

famouswaffles wrote: post #750
HarHarVeryFunny wrote: post #553

Earlier quoted context omitted.

LLMs are here to stay until something better replaces them, and will be used for those things they are capable of. It shouldn't be surprising they are not great at reasoning, or everything one would hope for from an AGI, since they simply were not built for that. If you look at the development history, the transformer was a successor to LSTM-based seq-2-seq models using Bahdanau attention, whose main goal was to more…

Reasoning is reasoning. "Look as if it is reasoning" is an imaginary distinction you've made up. One that is very clear because everybody touting this "fake reasoning" rhetoric is still somehow unable to define a testable version of reasoning that disqualifies LLMs without also disqualifying some chunk of humans. >As humans we routinely think for seconds/minutes or even hours before speaking or acting No human is ite…

> Reasoning is reasoning. "Look as if it is reasoning" is an imaginary distinction you've made up.

No - just because something has the surface appearance of reasoning doesn't mean that the generative process was reasoning, anymore than a cargo cult wooden aircraft reflects any understanding of aerodynamics and would be able to fly.

We've already touched on it, but the "farmer crossing river" problems is a great example. When the LLM sometimes degenerates into "cross bank A to B with chicken, cross band B to A with chicken, cross bank A to B with chicken.. that is the fewest trips possible", this is an example of "looks as if it is reasoning" aka cargo-cult surface-level copying of what a solution looks like. Real reasoning would never repeat a crossing without loading/unloading something since that conflicts with the goal of fewest trips possible.

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

#764

Post by famouswaffles »

HarHarVeryFunny wrote: post #762
famouswaffles wrote: post #754

Earlier quoted context omitted.

Sure I agree. But if that's what you're getting hung up on, i think you've missed his point entirely. Whether the machines becomes a human brain clone or something entirely alien is irrelevant. The point is, you can't cheat reality. Statistics is not magic. You can't predict text that understands without understanding.

Sure you can, and if your predictive engine doesn't have the generality and power of the original generative one, then you have no choice. Machine learning isn't magic - the model will learn what it can to minimize the error over the specific provided loss function, and no more. Change the loss function and you change what the model learns. In the case of an LLM trained with a predict next word loss function, what yo…

>the model will learn what it can to minimize the error over the specific provided loss function, and no more. Change the loss function and you change what the model learns.

You clearly do not really understand what it means to predict internet scale text with increasing accuracy. No more than that ? Fantastic

LLMs do not just learn surface statistics. So many papers have thoroughly disabused this that i'm just not going to bother. This is just straight up denial.

This havs been evidently shown in chess as well. https://arxiv.org/abs/2403.15498v2

You have no idea what you are talkin about. You've probably never even played with 3.5-turbo-instruct. That's how you can say this nonsense. You have your conclusion and keep working backwards to get a justification.

>It's interesting that LLMs can reach the ELO level that they do (says more about chess than about LLMs)

When you say this for everything LLMs can do then it just becomes a meaningless cope statement.

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

#765

Post by mvkel »

vasilipupkin wrote: post #546
mvkel wrote: post #540

Earlier quoted context omitted.

Current investors just need the co to be valued at $50B on the next round (likely, given fomo and hype) to make a 10X gain. Actually converting it to cash? That doesn't happen anymore. Everyone just focuses on IRR and starts the campaign for Fund II.

you are missing the point. SSI believes that it can build a super intelligence. Regardless of whether you personally buy into that or not, the expected value of such an investment is infinity effectively. 5 billion dollar valuation is a steal

OpenAI and Anthropic also believe that, too. The thing they have is products, customers, and traction.

$5B pre-product, betting on the team is fine. $50B needs to be a lot more than that.

Many examples of industries collapsing under the weight of belief. See: crypto.

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

#766

Post by famouswaffles »

HarHarVeryFunny wrote: post #763
famouswaffles wrote: post #750

Earlier quoted context omitted.

Reasoning is reasoning. "Look as if it is reasoning" is an imaginary distinction you've made up. One that is very clear because everybody touting this "fake reasoning" rhetoric is still somehow unable to define a testable version of reasoning that disqualifies LLMs without also disqualifying some chunk of humans. >As humans we routinely think for seconds/minutes or even hours before speaking or acting No human is ite…

> Reasoning is reasoning. "Look as if it is reasoning" is an imaginary distinction you've made up. No - just because something has the surface appearance of reasoning doesn't mean that the generative process was reasoning, anymore than a cargo cult wooden aircraft reflects any understanding of aerodynamics and would be able to fly. We've already touched on it, but the "farmer crossing river" problems is a great examp…

I never said anything about the surface appearance of reasoning. Either the model demonstrates some understanding or reasoning in the text it generates as it is perfectly capable of or it reasons faultily or lacks understanding in that area. This does not mean LLMs don't reason anymore than it means you don't reason.

The idea that LLMs "fake reason" and Humans "really reason" is an imaginary distinction. If you cannot create any test that can distinguish the two then you are literally making things up.

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

#767

Post by HarHarVeryFunny »

famouswaffles wrote: post #764
HarHarVeryFunny wrote: post #762

Earlier quoted context omitted.

Sure you can, and if your predictive engine doesn't have the generality and power of the original generative one, then you have no choice. Machine learning isn't magic - the model will learn what it can to minimize the error over the specific provided loss function, and no more. Change the loss function and you change what the model learns. In the case of an LLM trained with a predict next word loss function, what yo…

>the model will learn what it can to minimize the error over the specific provided loss function, and no more. Change the loss function and you change what the model learns. You clearly do not really understand what it means to predict internet scale text with increasing accuracy. No more than that ? Fantastic LLMs do not just learn surface statistics. So many papers have thoroughly disabused this that i'm just not g…

No of course not - they also learn whatever is necessary, and possible, in order to replicate those surface statistics (e.g. understanding of fairy tales, etc, as I noted).

However, you seem to be engaged in magical thinking and believe these models are learning things beyond their architectural limits. You appear to be star struck by what these models can do, and blind to what one can deduce - and SEE - they they are unable to do.

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

#768

Post by robertlagrant »

czl wrote: post #741
robertlagrant wrote: post #683

Earlier quoted context omitted.

> Assumption 3 is mostly techno-optimism, or an extrapolation of the trend that we are able to build ever more advanced devices. This is your weak link. I don't see why progress will be a straight line and not a sloping-off curve. You shouldn't see the progress we've made in vehicle speed and assume we can hit the speed of light.

> I don't see why progress will be a straight line and not a sloping-off curve. Technological progress often appears linear in the short term, but zooming out reveals an exponential curve, similar to compound interest. > You shouldn't see the progress we've made in vehicle speed and assume we can hit the speed of light. Consider the trajectory of maximum speeds over millennia, not just recent history. We've achieved…

> While reaching light speed is challenging, we're exploring novel concepts like light-propelled nano-vehicles

Do you think hitting light speed is an engineering problem or a fundamental constraints problem?

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

#769

Post by famouswaffles »

HarHarVeryFunny wrote: post #767
famouswaffles wrote: post #764

Earlier quoted context omitted.

>the model will learn what it can to minimize the error over the specific provided loss function, and no more. Change the loss function and you change what the model learns. You clearly do not really understand what it means to predict internet scale text with increasing accuracy. No more than that ? Fantastic LLMs do not just learn surface statistics. So many papers have thoroughly disabused this that i'm just not g…

No of course not - they also learn whatever is necessary, and possible, in order to replicate those surface statistics (e.g. understanding of fairy tales, etc, as I noted). However, you seem to be engaged in magical thinking and believe these models are learning things beyond their architectural limits. You appear to be star struck by what these models can do, and blind to what one can deduce - and SEE - they they ar…

You've said a lot of things about LLM chess performance that is not true and can be easily shown to be not true. Literally evidence right there that shows the model learning the board state, rules, player skills etc.

And then you've tried to paper over being shown that with a conveniently vague and nonsensical, "says more about bla bla bla". No, you were wrong. Your model about this is wrong. It's that simple.

You start from your conclusions and work your way down from it. "pattern matching technique" ? Please. By all means, explain to all of us what this actually entails in a way we can test for it. Not just vague words.

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

#770

Post by HarHarVeryFunny »

famouswaffles wrote: post #769
HarHarVeryFunny wrote: post #767

Earlier quoted context omitted.

No of course not - they also learn whatever is necessary, and possible, in order to replicate those surface statistics (e.g. understanding of fairy tales, etc, as I noted). However, you seem to be engaged in magical thinking and believe these models are learning things beyond their architectural limits. You appear to be star struck by what these models can do, and blind to what one can deduce - and SEE - they they ar…

You've said a lot of things about LLM chess performance that is not true and can be easily shown to be not true. Literally evidence right there that shows the model learning the board state, rules, player skills etc. And then you've tried to paper over being shown that with a conveniently vague and nonsensical, "says more about bla bla bla". No, you were wrong. Your model about this is wrong. It's that simple. You st…

An LLM will learn what it CAN (and needs to, to reduce the loss), but not what it CAN'T. How difficult is that to understand?!

Tracking probable board state given a sequence of moves (which don't even need to go all the way back to the start of the game!) is relatively simple to do, and doesn't require hundreds of sequential steps that are beyond the architecture of the model. It's just a matter of incrementally updating the current board state "hypothesis" per each new move (essentially: "a knight just moved to square X, so it must have moved away from some square a knight's move away from X that we believe currently contains a knight").

Ditto for estimating player ELO rating in order to predict appropriately good or bad moves. It's basically just a matter of how often the player makes the same move as other players of a given ELO rating in the training data. No need for hundreds of steps of sequential computation that are beyond the architecture of the model.

Doing an N-ply lookahead to reason about potential moves is a different story, but you want to ignore that and instead throw out a straw man "counter argument" about maintaining board state as if that somehow proves that the LLM can magically apply > N=layers of sequential reasoning to derive moves. Sorry, but this is precisely magical faith-based thinking "it can do X, so it can do Y" without any analysis of what it takes to do X and Y and why one is possible, and the other is not.

Post reply on HN