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

Sam Altman goes before US Congress to propose licenses for building AI

reuters.com

Post reply on HN
Print view View original

891–900 of 1001 posts

Re: Sam Altman goes before US Congress to propose licenses for building AI

#891

Post by srslack »

adamsmith143 wrote: post #168
tgv wrote: post #36

Earlier quoted context omitted.

I'm squarely in the "stochastic parrot" camp (I know it's not a simple markov model, but still, ChatGPT doesn't think), and it's clearly possible to interpret this as a grifting, but your argumentation is too simple. You're leaving out the essentials. These models do more than fitting the data given. They can output it in a variety of ways, and through their approximation, can synthesize data as well. They can output…

Please explain how Stochastic Parrots can perform chain of reasoning and answer out of distribution questions from exams like the GRE or Bar.

Probably because it fits the data. CoT and out of order questions from exams says nothing about whether it can generalize and adapt to things outside of its corpus.

Re: Sam Altman goes before US Congress to propose licenses for building AI

#892

Post by burtonator »

happytiger wrote: post #497

We need to MAKE SURE that AI as a technology ISN'T controlled by a small number of powerful corporations with connections to governments. To expound, this just seems like a power grab to me, to "lock in" the lead and keep AI controlled by a small number of corporations that can afford to license and operate the technologies. Obviously, this will create a critical nexus of control for a small number of well connected…

I'd go further and say if only corporations with government license are allowed to create LLMs, then we should NOT have LLMs.

Let the market develop organically.

Re: Sam Altman goes before US Congress to propose licenses for building AI

#893

Post by srslack »

adamsmith143 wrote: post #173
jazzyjackson wrote: post #152

Earlier quoted context omitted.

your comment reads to me as totally disconnected to the OP, whose concern relates to using the appearance of intelligence as a scare tactic to build a regulatory moat.

Actually OP is clearly, ironically, parroting the stochastic parrot idea that LLMs are incapable of anything other than basic token prediction and dismissing any of their other emergent abilities.

It can't generalize and adapt outside of its corpus, not in a way that's correct anyhow, and there's nothing "emergent." They are incapable of anything other than token prediction on its corpus and context. it just produces really good predictions. Funny how everyone keeps citing that Microsoft paper, when Microsoft is who is lobbying for this regulatory capture, and it's already been shown that such emergence on the tasks they chose when you scale up was a "mirage."

Re: Sam Altman goes before US Congress to propose licenses for building AI

#894

Post by ChatGTP »

ur-whale wrote: post #54
ChatGTP wrote: post #5

Earlier quoted context omitted.

Where is the regulatory capture? You might just need to apply for a license. Why is that so horrible? Most travel agents need a license, taxi drivers etc. Not sure why the same shouldn't apply for "AI"?

> Most travel agents need a license, taxi drivers etc. You say it like it's a good thing.

I'm saying that a license might be a simple and cheap thing to get, there's no need for 930 comments freaking out just yet.

Re: Sam Altman goes before US Congress to propose licenses for building AI

#895

Post by zamnos »

api wrote: post #831
SkyMarshal wrote: post #695

Earlier quoted context omitted.

> To expound, this just seems like a power grab to me, to "lock in" the lead and keep AI controlled by a small number of corporations that can afford to license and operate the technologies. If you actually watch the entire session, Altman does address that and recommend to Congress that regulations 1) not be applied to small startups, individual researchers, or open source, and 2) that they not be done in such a way…

That would make the regulations fairly pointless unless you think only mega-corps will ever be able to afford the compute for these things. Compute continues to get cheaper and cheaper. We have not hit the physics wall yet on that. That and if someone cracks efficient distributed training in a swarm type configuration then you could train models Seti@Home style. Lots of people would be happy to leave a gaming PC on t…

Shit, vast.ai will pay you right now for access to your gaming PC's GPU

Re: Sam Altman goes before US Congress to propose licenses for building AI

#897

Post by srslack »

adamsmith143 wrote: post #160
srslack wrote: post #9

Imagine thinking that regression based function approximators are capable of anything other than fitting the data you give it. Then imagine willfully hyping up and scaring people who don't understand, and because it can predict words you take advantage of the human tendency to anthropomorphize, so it follows that it is something capable of generalized and adaptable intelligence. Shame on all of the people involved in…

"It's just a stochastic parrot" is one of the dumbest takes on LLM's of all time.

That's not what I said. But, feel free to find some emergent properties or capability of total abstract reasoning and generalizing to data outside of its corpus that doesn't turn out to be a mirage of the wishes of the researchers.

Re: Sam Altman goes before US Congress to propose licenses for building AI

#898

Post by jart »

antiloper wrote: post #150
generalizations wrote: post #134

This is going to be RSA export restrictions all over again. I wish the regulators the best of luck in actually enforcing this. I'm tempted to think that whatever regulations they put in place won't really matter that much, and progress will march on regardless. Give it a year and a 10x more efficient algorithm, and we'll have GPT4 on our personal devices and there's nothing that any government regulator will be able…

Enforcing this is easy. The top high performance GPU manufacturers (Nvidia, AMD, Intel) are all incorporated in the U.S.

Back when crypto was a munition, common people didn't care about being able to invent their own crypto algorithms. They just wanted to use the existing ones, and all they needed to do that was some C code that Phil Zimmermann famously published in a book, to get around export controls. Controlling GPUs won't control the use of AI because GPUs are only needed to train new models. If you just want to use a large language model then CPU works great and you just need one guy to publish the weights.

That happened a few months ago with LLaMA. Since then, the open source community has exceeded all expectations democratizing the hardware angle. AI regulators are already checkmated and they don't know it yet. If their goal is to control the use of AI (rather than simply controlling the people building it) then they'd need to go all the way when it comes to tyranny in order to accomplish their goals. Intel would need to execute Order 66 with their Management Engine and operating systems would need to modify their virus scanners to monitor and report the use of linear algebra. It'd be outrageous.

Re: Sam Altman goes before US Congress to propose licenses for building AI

#899

Post by robomartin »

I get the feeling that a lot of commenters on this thread did not bother to watch the congressional hearing at all. Comments seem to be painting extremes that were not part of this hearing at all.

It was almost refreshing in that discussions these days --of any kind, not just hearings-- seem to devolve into people tossing feces at each other rather than having constructive engagement.

Well worth watching. Likely one of many to come.

Post reply on HN