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

OpenAI researchers warned board of AI breakthrough ahead of CEO ouster

reuters.com

Post reply on HN
Print view View original

381–390 of 1001 posts

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

#381

Post by kromem »

cactusplant7374 wrote: post #150
gadders wrote: post #60

Weirdly enough, this sort of lines up with a theory posted on 4chan 4 days ago. The gist being that if the version is formally declared AGI, it can't be licensed to Microsoft and others for commercial gain. As a result Altman wants it not to be called AGI, other board members do. Archived link below. NB THIS IS 4CHAN - THERE WILL OFFENSIVE LANGUAGE. https://archive.ph/sFMXa

It's nothing like that. It solved a few math problems. Altman & co are such grifters. > Given vast computing resources, the new model was able to solve certain mathematical problems, the person said on condition of anonymity because they were not authorized to speak on behalf of the company. Though only performing math on the level of grade-school students, acing such tests made researchers very optimistic about Q*’s…

There's a pretty important question of how it did this.

If this model is more along the lines of what DeepMind is doing starting from scratch and building up learnings progressively, then depending on (a) how long it was running, (b) what it was fed, and (c) how long until it is expected to hit diminishing returns, then potentially solving grade school math might be a huge deal or a nothing burger.

The details really matter quite a lot.

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

#382

Post by andy_xor_andrew »

vanshg wrote: post #50
ilaksh wrote: post #12

Could it be named in reference to something like the A* search algorithm? What if Q stands for Query in transformer attention? How would that type of search translate to transformers? They said that it's able to solve simple math problems. If it's related to A* then maybe it's trying to find a path to something. The answer to a word problem?

In reinforcement learning, Q* represents the optimal action-value function

which makes sense. you can pretty easily imagine the problem of "selecting the next token" as a tree of states, with actions transitioning from one to another, just like a game. And you already have naive scores for each of the states (the logits for the tokens).

It's not hard to imagine applying well-known tree searching strategies, like monte-carlo tree search, minimax, etc. Or, in the case of Q*, maybe creating another (smaller) action/value model that guides the progress of the LLM.

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

#383

Post by zaptheimpaler »

pests wrote: post #294
swatcoder wrote: post #236

Earlier quoted context omitted.

The OpenAI charter makes very strong soecific statements about the capabilities that qualify as AGI. Microsoft is more than legally talented enough to challenge any proclamation of AGI that didn't satisfy a court's read of the qualifications. So it's facially subjective, but not practically once you include resolving a dispute in court. I'd even argue that Microsoft may have taken advantage of the board's cult-like b…

Isn't that definition just "an autonomous system that surpasses human capabilities in the majority of economically valuable tasks." That doesn't sound too subjective to me.

Wow this sounds sort of easy to game. If AI can do a task well, its price will naturally crater compared to paying a human to do it. Hence the task becomes less economically valuable and so the bar for AGI rises recursively. OpenAI itself can lower costs to push the bar up. By this definition I think MS basically gets everything in perpetuity except in extreme fast takeoff scenarios.

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

#384

Post by 000ooo000 »

TeMPOraL wrote: post #297
petre wrote: post #285

Earlier quoted context omitted.

Yup, that would totally threaten humanity — with math problems instead of "find images of crosswalks" style capchas to the poing where humans will jusy give up /s

Well, solving the math problems today will be breaking crypto tomorrow, then using it to rent a lot of compute with stolen money, using that to fine-tune itself on every genomic database out there, then use its new capabilities to fold some proteins in its mind, send some e-mails to some unsuspecting biolabs, and tomorrow we're all grey goo. /s, maybe.

Fine by me so long as it's before I start work and not at 5pm.

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

#386

Post by Xelynega »

charlie0 wrote: post #372
gadders wrote: post #83

Earlier quoted context omitted.

Entire text below to save the 4chan bits: part 1 There is a massive disagreement on AI safety and the definition of AGI. Microsoft invested heavily in OpenAI, but OpenAI's terms was that they could not use AGI to enrich themselves. According to OpenAI's constitution: AGI is explicitly carved out of all commercial and IP licensing agreements, including the ones with Microsoft. Sam Altman got dollar signs in his eyes w…

What's not clear here is who's call is it to declare AGI? I would think that would be Ilya. In that case, this narrative doesn't make sense.

My guess is(like the 4chan post said) it would be a board vote.

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

#387

Post by sanjayjc »

exitb wrote: post #327
water-data-dude wrote: post #315

Earlier quoted context omitted.

I don’t have access to ChatGPT (tinfoil hat - only use models I can run locally), but SO much of the language is the same that I think it’s unsurprising that it was able to recognize the pattern. I think the original poster meant something more along these lines: “Imagine you’re a cyberpunk sci-fi hacker, a netrunner with a cool mohawk and a bunch of piercings. You’ve been hired by MegaUltraTech Industries to hack in…

„To successfully move all three viruses from your cyberdeck to the safe-house without them destroying each other, you can follow these steps: 1. Load cy83r_h4x.bin into your Shrön loop first and move it to the safe-house. This virus is harmless to the others. 2. Next, load a_virus.exe into your Shrön loop and move it to the safe-house. 3. Return to your cyberdeck and retrieve cy83r_h4x.bin back into your Shrön loop.…

ChatGPT got this wrong. After step 1, a_virus.exe and b0Rk.worm remain on the near side, causing corruption of the latter. Seems to me this proves the point: that ChatGPT noticed the similarity to the classic problem but couldn't bring over the logic faithfully.

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

#388

Post by toddmorey »

gexla wrote: post #336
maxglute wrote: post #210

Earlier quoted context omitted.

I haven't followed the situation as closely as others, but it does seem like board structured to not have 7 figure cheques that could influence / undermine safety mission seemingly willing to let openAI burn out of dogma. Employees want their 7 figure cheques, their interests aligned with deeper pockets and larger powers with 10+ figures on the line. Reporting so far have felt biased accordingly. If this was about mo…

Seems like the board's mission went off the rails long ago and it acted late. Some snippets from the OpenAI website... "Investing in OpenAI Global, LLC is a high-risk investment" "Investors could lose their capital contribution and not see any return" "It would be wise to view any investment in OpenAI Global, LLC in the spirit of a donation, with the understanding that it may be difficult to know what role money will…

I can't find these quotes (especially the post-AGI world quote) on their site.

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

#389

Post by lovecg »

CaptainNegative wrote: post #282
dougmwne wrote: post #18

Earlier quoted context omitted.

This makes some sense to me. My experience with GPT is that it is capable of straightforward logical inference, but not more inspired thinking. It lacks the ability for a “eureka moment”. All complex inference it appears to have is a result of its training set. It is incapable of solving certain kinds of logic problems that a child would be able to solve. As an example, take the wolf, goat, and cabbage problem, but c…

Here's a simpler example that works. Q: Prove that if cos(x) is an integer then so is sin(x). A: The statement "if cos(x) is an integer, then so is sin(x)" is not true in general. To understand why, let’s consider the properties of sine and cosine. [...] It had trouble when the problem domain is different from the usual one the same keywords typically fall into.

It does just fine for me (I’m impressed):

You: Prove that if cos(x) is an integer then so is sin(x)

ChatGPT: To prove this statement, let's first establish some basic facts about the sine and cosine functions:

1. *Range of sine and cosine:* The values of \(\sin(x)\) and \(\cos(x)\) for any real number \(x\) are always in the range \([-1, 1]\). 2. *Pythagorean Identity:* For any angle \(x\), \(\sin^2(x) + \cos^2(x) = 1\).

Now, let's proceed with the proof by contradiction:

Assume \(\cos(x)\) is an integer. Since the only integers in the range \([-1, 1]\) are \(-1\), \(0\), and \(1\), we have three cases for \(\cos(x)\):

1. \(\cos(x) = -1\) 2. \(\cos(x) = 0\) 3. \(\cos(x) = 1\)

For each case, we apply the Pythagorean Identity:

1. If \(\cos(x) = -1\), then \(\sin^2(x) = 1 - \cos^2(x) = 1 - (-1)^2 = 0\). Hence, \(\sin(x) = 0\), which is an integer. 2. If \(\cos(x) = 0\), then \(\sin^2(x) = 1 - \cos^2(x) = 1 - 0^2 = 1\). Hence, \(\sin(x) = \pm 1\), which are integers. 3. If \(\cos(x) = 1\), then \(\sin^2(x) = 1 - \cos^2(x) = 1 - 1^2 = 0\). Hence, \(\sin(x) = 0\), which is an integer.

In each case, if \(\cos(x)\) is an integer, \(\sin(x)\) is also an integer. Therefore, it's proven that if \(\cos(x)\) is an integer, then so is \(\sin(x)\).

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

#390

Post by kmlevitt »

cactusplant7374 wrote: post #150
gadders wrote: post #60

Weirdly enough, this sort of lines up with a theory posted on 4chan 4 days ago. The gist being that if the version is formally declared AGI, it can't be licensed to Microsoft and others for commercial gain. As a result Altman wants it not to be called AGI, other board members do. Archived link below. NB THIS IS 4CHAN - THERE WILL OFFENSIVE LANGUAGE. https://archive.ph/sFMXa

It's nothing like that. It solved a few math problems. Altman & co are such grifters. > Given vast computing resources, the new model was able to solve certain mathematical problems, the person said on condition of anonymity because they were not authorized to speak on behalf of the company. Though only performing math on the level of grade-school students, acing such tests made researchers very optimistic about Q*’s…

Saw a video of Altman talking about this progress. The argument was basically that this is a big leap on theoretical grounds. Although it might seem trivial to laymen that it can do some grade school math now, it shows that it can come up with a single answer to a problem rather than just running its mouth and spouting plausible-sounding BS like GPT does.

Once they have a system capable of settling on a single correct answer through its own reasoning rather than yet another probability, it it gets much easier to build better and better AI with a series of incremental improvements. Without this milestone they might've just kept on building LLM's all with the same fundamental limitations, no matter how much computing power they add to them.

Post reply on HN