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

231–240 of 1001 posts

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

#231

Post by sweezyjeezy »

famouswaffles wrote: post #138
zugi wrote: post #114

Earlier quoted context omitted.

None of us have seen the letter so this may be off base, but I would expect people working at the world's most prominent AI research organization to have more skepticism about the ramifications of any one "breakthrough." Perhaps most did, but a couple didn't and wrote the letter? More than 3 decades ago when AI started beating humans at chess, some people feared AGI was right around the corner. They were wrong. Last…

>Last year a Google researcher thought his chat bot was sentient ( https://www.scientificamerican.com/article/google-engineer-c ... .) He was wrong. You've figured out how to test for sentience ?

I think the word "sentience" is a red herring. The more important point is that the researcher at Google thought that the AI had wants and needs 'like a human', e.g. that if it asked the AI if it wanted legal representation to protect its own rights, this was the same as asking a human the same question.

This needs much stronger evidence than the researcher presented, when slight variations or framing of the same questions could lead to very different outcomes from the LLM.

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

#232

Post by hnav »

dkjaudyeqooe wrote: post #133

You've got to hand it to OpenAI for their bleeding edge marketing strat: "WE'VE GOT AN AGI IN THIS CABINET RIGHT HERE AND WE DON'T WANT TO LET IT OUT BECAUSE IT'S SO POWERFUL, IT'S SO AI YOU HAVE NO IDEA, BEST YOU THROW MONEY AT US AND WE'LL KEEP EVERYONE SAFE BUT SORRY YOU"RE ALL REDUNDANT BECAUSE AGI, SORRY" You'd think that so soon after so many Crypto scams people would be just a wee bit less credulous.

Maybe they asked ChatGPT to come up with an unconventional marketing plan and it had a touch too much blogspam "that will SHOCK and PETRIFY you" in its training set so ChatGPT suggested that they feign a powerstruggle over keeping a humanity ending super-intelligence under wraps.

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

#233

Post by upwardbound »

atleastoptimal wrote: post #226
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

Altman has no stake in OpenAI, how could he make money licensing it?

By building quid pro quo or "revolving door" relationships. https://en.wikipedia.org/wiki/Revolving_door_(politics)

For example:

Sam spent the last 4 years making controversial moves that benefited Microsoft a lot https://stratechery.com/2023/openais-misalignment-and-micros... at the cost of losing a huge amount of top talent (Dario Amodei and all those who walked out with him to found Anthropic).

In November, Sam loses his job for unknown reasons, and is accused of having molested his younger sister Annie. https://www.themarysue.com/annie-altmans-abuse-allegations-a...

Despite this, his best buddy Satya Nadella immediately gives him a huge job offer without even putting him through an interview loop.

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

#234

Post by toomuchtodo »

ssnistfajen wrote: post #214
toomuchtodo wrote: post #10

Earlier quoted context omitted.

I trust AGI more than any of the humans so far who have tried to argue they can wrangle it.

Benevolent Gods do not exist. Never have, never will. So don't place your faith in them.

I never assumed it would be benevolent.

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

#235

Post by bhy »

bcherry wrote: post #115
vjerancrnjak wrote: post #81

Earlier quoted context omitted.

OpenAI already benchmarks their GPTs on leetcode problems and even includes a Codeforces rating. It is not impressive at all and there's almost no progress from GPT 2 to 4. I agree, why does this grade school math problem matter if the model can't solve problems that are very precisely stated and have a very narrow solution space (at least more narrow than some vague natural language instruction)?

I suspect the truth (if this claim is true) is a lot more nuanced than "it did grade-school math", and there's more context surrounding this claim which insiders have and makes it much more interesting.

Like most research they likely started with a smaller model like GPT 2 or 3 and shown that they can significantly boost the performance to the extent of solving grade school math.

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

#236

Post by swatcoder »

Waterluvian wrote: post #224
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

The moment I read about that clause I was shocked that Microsoft lawyers would agree to such subjectivity that can have incredibly expensive implications. Were they expecting it to never happen and were just ready to throw a mountain of lawyers at any claim?

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 blindspots and believes that a court-acceptable qualifying AGI isn't a real enough possibility to jeopardize their contract at all.

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

#237

Post by skepticATX »

dkjaudyeqooe wrote: post #133

You've got to hand it to OpenAI for their bleeding edge marketing strat: "WE'VE GOT AN AGI IN THIS CABINET RIGHT HERE AND WE DON'T WANT TO LET IT OUT BECAUSE IT'S SO POWERFUL, IT'S SO AI YOU HAVE NO IDEA, BEST YOU THROW MONEY AT US AND WE'LL KEEP EVERYONE SAFE BUT SORRY YOU"RE ALL REDUNDANT BECAUSE AGI, SORRY" You'd think that so soon after so many Crypto scams people would be just a wee bit less credulous.

The report is already starting to get rolled back. We now have a direct denial from OpenAI, and The Verge is reporting that their sources have also denied this happened.

The lack of skepticism that is demonstrated whenever OpenAI/AGI comes up, especially among technologists, is concerning.

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

#239

Post by mvdtnz »

Andrex wrote: post #215
ikesau wrote: post #195

Earlier quoted context omitted.

This is exactly how confirmation bias fuels conspiracy theories. No one believes anything that they think sounds implausible. As a general rule, you should give very little thought to anonymous 4chan posts.

That's absolutely true. But They have leaked real things in the past, in exactly the same way. It may be 5% or less that turn out to be true, but there's the rub. That's why no one can completely dismiss it out of hand (and why were even discussing it on an HN comment thread in the first place).

I assure you, I can (and did) dismiss it out of hand. 4chan shitposting is not credible.

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

#240

Post by pk-protect-ai »

hn_throwaway_99 wrote: post #165
pk-protect-ai wrote: post #78

Earlier quoted context omitted.

I compared GPT-4 Turbo with my previous tests on GPT-4, and the results are quite interesting. GPT-4 Turbo is better at arithmetic and makes fewer errors in multiplying four-digit numbers. In fact, it makes significantly fewer errors with five-digit numbers. The level of errors on five-digit numbers is high but much lower than with four-digit numbers in GPT-4. multiplication of floats XX.MMMM and YYY.ZZZZ produces er…

But the point about how it just "improves" with slightly larger numbers, but still fails at really big numbers, shows that it's not really "reasoning" about math in a logical way - that's the point I was getting at. For example, once you teach a grade schooler the basic process for addition, they can add 2 30 digit numbers correctly fairly easily (whether they want to do it or not is a different story). The fact that…

Of course, it isn't. It approximates. I bet you'll get better results by increasing the depth of the network, as with each layer, you'll achieve a more accurate approximation. I have an idea for achieving this without significantly increasing the number of layers, and I'm currently working on it as a side project. However, this idea might prove to be useless after all, as it requires training the model from scratch with a lot of synthetic data mixed in. Experiments on small models look promising, but they are negligible, and I can't afford to train a larger model from scratch for a side project.
Post reply on HN