Live data from Hacker News

Sam and Greg's response to OpenAI Safety researcher claims

twitter.com

281–290 of 508 posts

Re: Sam and Greg's response to OpenAI Safety researcher claims

#281

Earlier quoted context omitted.

If you look at compute scaling and model improvement on that compute, we’re going to get there pretty fast. Both compute, architecture and cost matter, and those have been improving like crazy. I don’t know about you, but I didn’t expect GPT-4o to come out at half the price one year later, with real-time voice, image and text. There is zero sign of slowing, Nvidia keeps building beefier GPUs specialized in LLMs, the…

> You can plot this exponential growth out over time and calculate when these models will have the complexity of the brain. Then you can assume some penalty for shitty architecture (that gets better over time), and you’ll have a ballpark estimate. The same thing could’ve been said for self driving cars, or the space program, or a lot of things that seemed to be progressing quickly at the time.

And if you did say it, you would have been a lot more correct than if you'd said they won't amount to anything. Robotaxis with nobody in the driver's seat are available in three major US cities; most people use a network of navigational satellites every time they want to figure out how to get somewhere new.

Re: Sam and Greg's response to OpenAI Safety researcher claims

#282

My thought about the demise of OAI’s AGI safety team is that they outlived their usefulness. It was a marketing campaign. “We’re super cool. We’re super advanced. We hope don’t make the terminator! wink wink ” “Oh no! GPT-2 is too dangerous! We better not release it! Oh wait. We did. ¯\_(ツ)_/¯ “) It was never a serious thing. Sure here and in the depths of weirdo forums people whispered about the AGI that can’t be na…

> It was a marketing campaign. > it was never a serious thing Yeppppp. This is what i was saying for weeks and weeks, if not months. The amount of people on Reddit who can not see the puzzle pieces of “OpenAI sells AI” , “AI is hyped” , “more hype = more attention = more money and media coverage” and put them together, is actually shocking. Every time i would point this out on Reddit, more often than not i would get…

I find there can be significant cargo culting [0] on HN too when any of these topics hit the front page: OpenAI / ChatGPT / LLMs; Zoom; E2EE; regulation / policy; crypto currency / NFTs [1] etc.

That said, I’m kind of glad to hear that Reddit is worse, as it puts the situation here on HN into some wider perspective (it’s not that bad here).

[0]: meant in the form of: closed minded absolutism without full contextual appreciation or understanding of the domain (what you seem to be describing as happening on Reddit)

[1]: less so these days, I wonder if that has something do with the current AI hype cycle cough opportunistic bandwagon jumping cough o_o

Re: Sam and Greg's response to OpenAI Safety researcher claims

#283
post #85

Earlier quoted context omitted.

The difference here is that these aren't standard-issue HR/legal issues. The technology they're working on poses the gravest of dangers. This is uncharted territory, not just for the tech sector, but period. Whether the recent public back-and-forth is just internal drama/politics spilling out, or there really is a lack of gravitas around the handling of these issues within the company-- neither is good. Others have s…

Text continuation poses "the greatest of dangers"?

I will assume you’re asking in good faith.

Text continuation to match what people around the internet are saying requires the model internally to “program itself”. It is a non-von-neumann architecture but it encodes durable structures in BILLIONS of bytes, that’s like a codebase of gigabytes. It can pass the Bar exam, various pre-med exams, knows more than any particular human on every subject, and it’s only been a couple years.

The text continuation was the first interface, now they are doing multimodal. As it learns to predict the latent space for those things, it will be like predicting things in the real world. Training physical robots will become faster and easier too.

In short, this is basically a system to search through the space of all possible alhorithms in order to “reason about” a huge model of the world, enough to understand it and predict it with small loss.

AlphaZero used MCTS and can beat every human in chess now, and even every program written by a human, with beautiful gameplay. And even without AI, just the fact that the best performing algorithms are downloaded onto millions of computers is already multiplying the power far more than the 1% of humans who might do better in some area. So it can be better than 95% of humans at most tasks !

Re: Sam and Greg's response to OpenAI Safety researcher claims

#284
post #81

Earlier quoted context omitted.

We need protection from people who control AI. Who will provide that…

Open source/science! If open models approach/maintain near-SotA capabilities, then there's hope for avoiding a Big Tech AI aristocracy. Currently, that means Meta releasing their 405B GPT-4 killer (and probably subsequent models). We'll need community unity, to implement research results, test/harden models, create/aggregate/clean massive new datasets, figure out distributed training, lower hardware requirements, enc…

In my opinion, regulations are more important. Open source models cannot get to the same mainstream adoption as closed source ones, simply because closed source models will get more funding.

No-one beyond the tech circle will ever hear about these models or use them. This is a good second step but the first step is regulations.

Re: Sam and Greg's response to OpenAI Safety researcher claims

#285

Earlier quoted context omitted.

While we're being overly reductionist: it's just moving bits around, what harm could that possibly do?

"AI" is one of two things depending on whether you think it's actually intelligent: * If "AI" is actually intelligent, then it's no worse a threat than any living being known to man. * If "AI" is not actually intelligent, then it's no worse a threat than any other computer program. Both threat models are very thoroughly known and understood. I second parent commenter's sentiment that calling "AI" the "gravest of dang…

If "AI" is actually intelligent, then it's no worse a threat than any living being known to man

How does the conclusion follow?

Re: Sam and Greg's response to OpenAI Safety researcher claims

#286
post #167

Earlier quoted context omitted.

What's next?

Are you capable of being manipulated through text? How about an AI trained to generate text that will extract exactly the desired response from you? An advertiser's wet dream.

That’s thinking too small. Most people are nudged or coerced into taking action or supporting an action by the people who surround them. Their friends. Their coworkers. Their community. Like why people were on facebook 20 years ago, or forwarding chain emails, for instance.

Unleashing an army of AI bots to infiltrate an online community and shift its discourse can be done at scale. While at the same time those humans who resist can be endlessly distracted by either arguing with bots or their own friends who have been affected by bots.

All that is possible to do with CURRENT TECHNOLOGY.

Re: Sam and Greg's response to OpenAI Safety researcher claims

#287
post #201

Earlier quoted context omitted.

What probability for extinction do you consider to be low enough that it effectively becomes a Pascalian wager? Pascal was also writing about a single person's fate (AFAICT), whereas this is about the fate of everyone.

The point is that the probability doesn’t matter because the outcome, total human extinction, is given infinite weight.

Maybe Pascal argued that way for the human soul, but I don't think the AI risk argument needs infinite value to be at stake. Would you say that preparing for the risk of a solar flare or an asteroid impact is also a Pascalian wager?

Re: Sam and Greg's response to OpenAI Safety researcher claims

#288

Earlier quoted context omitted.

Text continuation poses "the greatest of dangers"?

While we're being overly reductionist: it's just moving bits around, what harm could that possibly do?

That’s the thing though; reducing the risk rhetorically reduces existential threat.

If it is propagated as “just electrons” no one gets anxious

If it’s propagated as “end is nigh” everyone panics

Both are reasonable in their grammatical correctness but only one seems a reasonable philosophy

Re: Sam and Greg's response to OpenAI Safety researcher claims

#289

Earlier quoted context omitted.

Text continuation poses "the greatest of dangers"?

https://www.theguardian.com/technology/article/2024/may/17/a... "...I also went to a panel in Palantir’s booth titled Civilian Harm Mitigation. It was led by two “privacy and civil liberties engineers” – a young man and woman who spoke exclusively in monotone. They also used countless euphemisms for bombing and death. The woman described how Palantir’s Gaia map tool lets users “nominate targets of interest” for “the…

But the claim was "th gravest of dangers". While your scenario is a nightmare, I'm not sure it qualifies as "the gravest".

Re: Sam and Greg's response to OpenAI Safety researcher claims

#290

Earlier quoted context omitted.

There seem to be only three options here: 1) AGI is not around the corner, so no worries 2) He no longer cares about the possible negative effects, a departure from his past statements 3) I missed something, please help me learn

> 1) AGI is not around the corner, so no worries I agree with this. I no longer agree that the "safety" people want the same things we do. I suspect after the bard debacle a bit back that "safety" looks a lot more like 1984. > He no longer cares ... I missed something Sam was gone, none of these people spoke up, and now they are leaving. I think the story is plain as day but we're pretending it isnt. Sam isnt a leade…

Wasn't sama a YC CEO or something?
Post reply on HN