OpenAI researchers warned board of AI breakthrough ahead of CEO ouster • Hacker News Board

Re: OpenAI researchers warned board of AI breakthrough ahead of CEO ouster

#971

Post by SpaceNoodled » Thu, Nov 23, 2023, 8:27 PM UTC

konschubert wrote: post #922

nopromisessir wrote: post #814

Earlier quoted context omitted.

All those things sound very boring to me. I can offer no concrete solutions. However, I have a friend who graduated from high school #1 of a big class and 2 years early. His mom explained that if he made at least a 1400(of 1600) on his SAT, she would buy him a new gaming computer. He then proceeded to make exactly a 1400. No more. No less. I recommend if you haven't tried already, an iteration to this approach using…

Be careful with reward systems, as it can destroy internal motivation. Additionally, one very important thing to learn as a young adult is how to motivate yourself to do things that have only long term payoffs. Of course I also understand that you can take SAT only once, so as bad as that is, it’s maybe not the best time to learn a life lesson.

The problem is that there are rarely guarantees for long-term playoffs.

Re: OpenAI researchers warned board of AI breakthrough ahead of CEO ouster

#972

Post by da39a3ee » Thu, Nov 23, 2023, 8:50 PM UTC

nostrademons wrote: post #593

wbhart wrote: post #491

I feel very comfortable saying, as a mathematician, that the ability to solve grade school maths problems would not be at all a predictor of ability to solve real mathematical problems at a research level. The reason LLMs fail at solving mathematical problems is because: 1) they are terrible at arithmetic, 2) they are terrible at algebra, but most importantly, 3) they are terrible at complex reasoning (more specifica…

What I wonder, as a computer scientist: If you want to solve grade school math problems, why not use an 'add' instruction? It's been around since the 50s, runs a billion times faster than an LLM, every assembly-language programmer knows how to use it, every high-level language has a one-token equivalent, and doesn't hallucinate answers (other than integer overflow). We also know how to solve complex reasoning chains…

What an absolutely idiotic comment.

> If you want to solve grade school math problems

That's not the aim here. Very obviously what we are talking about here is _complementing_ AI language models with improved mathematical abilities, and whether that leads to anything interesting. Surely you understand that? Aren't you one of the highest rated commenters on this site?

Re: OpenAI researchers warned board of AI breakthrough ahead of CEO ouster

#973

Post by patrick451 » Thu, Nov 23, 2023, 9:08 PM UTC

seanhunter wrote: post #735

nostrademons wrote: post #593

Earlier quoted context omitted.

What I wonder, as a computer scientist: If you want to solve grade school math problems, why not use an 'add' instruction? It's been around since the 50s, runs a billion times faster than an LLM, every assembly-language programmer knows how to use it, every high-level language has a one-token equivalent, and doesn't hallucinate answers (other than integer overflow). We also know how to solve complex reasoning chains…

There is a general result in machine learning known as "the bitter lesson"[1], which is that methods which come from specialist knowledge tend to be beaten by methods which rely on brute force computation in the long run because of Moore's law and the ability to scale things by distributed computing. So the reason people don't use the "add instruction"[2] for example is that over the last 70 years of attempting to bu…

When you say "a general result", what does that mean? In my world, a general result is something which is rigorously proved, e.g., the fundamental theorem of algebra. But this seems to be more along the lines of "we have lots of examples of this happening".

I'm certainly no expert, but it seems to me that Wolfram Alpha provides a counterexample to some extent, since they claim to fuse expert knowledge and "AI" (not sure what that means exactly). Wolfram Alpha certainly seems to do much better at solving math problems than an LLM.

Re: OpenAI researchers warned board of AI breakthrough ahead of CEO ouster

#974

Post by ugh123 » Thu, Nov 23, 2023, 9:11 PM UTC

cduzz wrote: post #572

ugh123 wrote: post #554

Earlier quoted context omitted.

Maybe I'm unsure what we're arguing here. Did the guys kid drum that up himself or did he learn it from yt? Knowledge can be inferred or extracted. If it comes up with a correct answer and shows it's work, who cares how the knowledge was obtained?

Yeah, my son only knows about imaginary numbers as far as the veritasium "epic math dual" video. As far as I can tell he inferred that |i+1| needs the Pythagorean theorem and that i and 1 are legs of the right triangle. I don't think anyone ever suggested that "absolute value" is "length". I asked him what |2i+2| would be an his answer of "square root of 8" suggests that he doesn't have it memorized as an answer beca…

Ah, makes sense. I think that is indeed impressive to fill in the paths of reasoning between unknown to known based on even small nuggets of information combined possibly with "learned intuition" which your son seems to have obtained an ability for to understand how to use both. Very cool

Re: OpenAI researchers warned board of AI breakthrough ahead of CEO ouster

#975

Post by Streiber » Thu, Nov 23, 2023, 9:49 PM UTC

I believe that the controversy is really about control. Control of the motivation for growth (capitalism vs. altruism) and control of the evolution of the technology (human control of the data to which GPT is exposed). The freak out by the board was the realization that once ChatGPT had 100 million or more users the algorithms would start learning from the queries in unpredictable ways. If you've used it you know that you can progressively tune your queries to home closer and closer to a usable result. If the algorithm can learn from that series of interactions then the genie is out of the bottle. I believe that when the board realized this and further knew that Sam had known all along that this would happen they fired him because they had lost control of the algorithms and Sam. Now what? Can Skynet become self-aware? A lot of my colleagues think that it's possible. From my 20+ years experience of working with neural nets I do not believe that Neural nets can behave like human geniuses because they cannot extrapolate outside of their experience. However, much of what we ascribe as intelligent behavior of human beings is definitely within the realm of neural nets. Humans learning from experience sometimes seems as uncommon as common sense.

Re: OpenAI researchers warned board of AI breakthrough ahead of CEO ouster

#976

Post by qgin » Thu, Nov 23, 2023, 9:55 PM UTC

riku_iki wrote: post #695

qgin wrote: post #64

There will come a day when 50% of jobs are being done by AI, major decisions are being made by AI, we're all riding around in cars driven by AI, people are having romantic relationships with AI... and we'll STILL be debating whether what has been created is really AGI. AGI will forever be the next threshold, then the next, then the next until one day we'll realize that we passed the line years before.

> day when 50% of jobs are being done by AI By OpenAI definition 50% is not enough to qualify for AGI, it has to be "nearly any economically valuable work"

It would be hard to find a single human who could handle nearly any/all economically valuable work. Getting good enough to get paid in one field is an achievement.

Re: OpenAI researchers warned board of AI breakthrough ahead of CEO ouster

#977

Post by Tlstrw » Thu, Nov 23, 2023, 10:25 PM UTC

BenoitP wrote: post #869

passwordoops wrote: post #860

Earlier quoted context omitted.

These guys smell so much like Tesla it's not even funny. Very impressive core tech, and genuinely advancing knowledge. But the hype train is just so powerful that the insane (to put it mildly) claims are picked up without any sense of critical thinking by seemingly intelligent people. They're both essentially cults at this point

Agreed, but IMHO it is sort of justified for Tesla. The size of its hype matches the size of the ICEs entrenchment in its moat. These have outsize influence on our economy, but climate change (and oil depletion) is quite inevitable. It takes irrational market cap to unseat that part of the economy being prisoner of its rent. And some allocators of capital have understood that.

It really wasn't. Musk kept making promises on insane deadline and missing them. Anyone with a lick of sence could see that EVs would take decades to reach the main stream and would have huge issues with the raw materials required to make it the transition possible. E.g. Building four more electricity grids worth of capacity to replace gas stations.

Re: OpenAI researchers warned board of AI breakthrough ahead of CEO ouster

#978

Post by valzam » Thu, Nov 23, 2023, 10:27 PM UTC

rsync wrote: post #481

valzam wrote: post #448

Earlier quoted context omitted.

Apparently Ilya has been leading people in "feel the AGI" chants and "I can feel the AGI" is his catchphrase within OA. So yes, some people might have gone of the rocker a little bit.

No … no no … That can’t be true. Right ?

https://twitter.com/ilyasut/status/1578238338288402432

Re: OpenAI researchers warned board of AI breakthrough ahead of CEO ouster

#979

Post by wegfawefgawefg » Thu, Nov 23, 2023, 11:51 PM UTC

Jeff_Brown wrote: post #964

wegfawefgawefg wrote: post #956

Earlier quoted context omitted.

Skull has limited volume. What room is unused by one capacity may be used by another.

Is this intuition scientifically supported? I've read that people who remember every detail of their lives tend not to have spectacular intelligence, but outside of that extreme I'm unaware of having seen the tradeoff actually bite. And there are certainly complementarities in knowledge -- knowing physics helps with chemistry, knowing math and drama both help with music, etc.

Chimps have a much better working memory than humans. They can also count 100 times faster than humans. However, the area of their brain responsible for this faculty is used for language in humans... The theory is that the prior working memory and counting ability may have been optimized out at some point to make physical room, assuming human ancestors could do it too.

Lookup the chimp test. the videos of the best chimp are really quite incredible.

There is also the measured inflation of map traversing parts of the brain in pro tetris players and taxi drivers. I vaguely remember an explanation about atrophy in nearby areas of the brain, potentially to make room.

Re: OpenAI researchers warned board of AI breakthrough ahead of CEO ouster

#980

Post by AgentMatt » Fri, Nov 24, 2023, 12:30 AM UTC

_emna wrote: post #781

AgentMatt wrote: post #758

Earlier quoted context omitted.

> To me, it makes sense to define abs(x) = sqrt(x^2) i.e. ignoring the negative solution enforces the positive result. Why does this make sense to you? You have some notion of what an absolute value should be, on an intuitive or conceptual level, and the mathematical definition you give is consistent with that (in the one dimensional case). Now taking this valid definition for the 1-d case and generalizing that to hi…

> Why does this make sense to you? You have some notion of what an absolute value should be, on an intuitive or conceptual level, and the mathematical definition you give is consistent with that (in the one dimensional case). In my mind abs(x) = x*sign(x) which is why the above formulation seems correct. This formulation is useful, for example, in formulating reflections. > Instead, you can go back to the conceptual…

> In my mind abs(x) = x * sign(x) which is why the above formulation seems correct.

> This is an interesting idea...how would you define sign(x) in a higher dimension?

You could think of the sign as the direction. In the 1-d case, you only have two directions. Positive sign means to the right of the origin, negative sign means to the left of the origin. But in higher dimensional case, you don't get a higher count of directions, instead direction becomes a space.

To see this analogy we can rewrite your abs(x) = x * sign(x) as x = abs(x) * sign(x). (Because 1/sign(x) = sign(x) except at 0, where the two equations agree anyway.)

Now consider that in higher dimensions, we can write x = ||x||*(x/||x||) for any vector x, where ||x|| denotes the magnitude and the term x/||x|| is the unit vector in direction of x. This term then plays the role of the sign.

A simple reflection can then still be done by multiplying this direction term with -1, which in the 2d case reflects at a line through the origin and perpendicular to the vector.

I can't comment on the Laplace transform, it's been too long since I used that.