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

Yann LeCun: AI one-percenters seizing power forever is real doomsday scenario

businessinsider.com

Post reply on HN
Print view View original

151–160 of 873 posts

Re: Yann LeCun: AI one-percenters seizing power forever is real doomsday scenario

#151

Post by neodypsis »

jszymborski wrote: post #5
Animats wrote: post #4

It's too late for monopolizing large language models. The technology for making them is known and the cost keeps coming down. There are things to worry about, but this doesn't seem to be one of them.

Two thoughts on that: 1. GPU prices aren't as high as they were during the mining peak, but you can still buy a car with the money it takes to buy a pod of A100s. Further to that point, Google et al. are making the next generation of ML accelerators and they won't even sell the best ones. 2. Compute is only half the problem. Access to data (e.g. in the form of email and smart assistant queries and ring door bells) is…

The biggest cost, other than capital costs, is energy. Some geographies have competitive advantages w.r.t. energy.

Re: Yann LeCun: AI one-percenters seizing power forever is real doomsday scenario

#152

Post by jandrewrogers »

PeterStuer wrote: post #48

People focus on LLM's and diffusion models because they are so omnipresent now. But an AI for stock picking and prediction that would be realy next level and outcompete the current ones cosistently would siphon of so much wealth it would basically own society if the operators were clever enough not to get so gready short term that the system would litterally collapse overnight.

There are fundamental limits to prediction. Characterizing these limits is an essential part of algorithmic information theory[0] (AIT) and is an operative bound on AI improvement. One of the strongest arguments against "hard take-off" scenarios is that AIT constrained by physics doesn't really allow it.

[0] https://en.wikipedia.org/wiki/Algorithmic_information_theory

Re: Yann LeCun: AI one-percenters seizing power forever is real doomsday scenario

#153

Post by troupo »

RandomLensman wrote: post #139
troupo wrote: post #123

Earlier quoted context omitted.

> The laws we have already cover malevolent and abusive uses of software applications. They don't. Or not nearly enough. Otherwise you wouldn't have automated racial profiling, en masse face recognition, credit and social scoring etc. And it's going to get worse. Because AI is infallible, right? Right?! That's why, among other thing, EU AI regulation has the following: - Foundation models have to be thoroughly docume…

AI not being allowed in high-risk applications is a bad idea, same for explicable in human terms (cannot do that now for human decisions anyway). The latter would kind of rule out any quantum computing, too, as the only real explanation there is through maths, not words.

> AI not being allowed in high-risk applications is a bad idea

wat

> same for explicable in human terms (cannot do that now for human decisions anyway).

What? Human decisions can be explained and appealed.

Good luck proving something when a blackbox finds you guilty of anything.

Re: Yann LeCun: AI one-percenters seizing power forever is real doomsday scenario

#154

Post by FeepingCreature »

candiodari wrote: post #143
wood_spirit wrote: post #122

Earlier quoted context omitted.

You can’t _stop_ it. You can make it illegal in some countries. But outlawing it just means some obvious candidate countries or private actors develop it first instead.

I think these attempts will simply point out that math is a higher power than law. If something becomes possible, and becomes easy, there is nothing law can do to stop it. Just look at the attempts to stop drugs in North West Europe. Huge sentences (so huge they are in fact never fully carried out), tens of dead police officers on a yearly basis, triple that in civilian victims and drugs are easier to get than ever b…

The aim is to stop it from becoming easy.

The drug war would look very different if good cocaine was only produced in one country with extremely costly chemistry machines from one company.

Re: Yann LeCun: AI one-percenters seizing power forever is real doomsday scenario

#155

Post by logicchains »

mg wrote: post #141
logicchains wrote: post #38

Earlier quoted context omitted.

>Is that possible? Has that happened in any other industry? It's absolutely happened to the nuclear weapons industry. And Big AI is lobbying for the government to treat AI development the same way it treats nuclear proliferation.

There are more nuclear warheads in Russia than in the USA.

Would you like to see a world where the only countries with their own AIs are Russia, the United States, China, France, the United Kingdom, Pakistan, India, Israel and North Korea, and all use of AI is controlled by the governments of those countries?

Re: Yann LeCun: AI one-percenters seizing power forever is real doomsday scenario

#156

Post by maeil »

topspin wrote: post #142
JoshTriplett wrote: post #52

Earlier quoted context omitted.

> seize power via regulatory capture Not even a little bit. "Stop" is not regulatory capture. Some large AI companies are attempting to twist "stop" into "be careful, as only we can". The actual way to stop the existential risk is to stop. https://twitter.com/ESYudkowsky/status/1719777049576128542 > the push to close off AI with farcical licensing and reporting requirements "Stop" is not "licensing and reporting requ…

> "Stop" is not regulatory capture. "Stop" is a fiction that exists exclusively inside your head. Unless your solution anticipates a planet wide GULAG system run by a global government your "stop" is infeasible.

If tomorrow TSMC, NVIDIA and ASML get destroyed in the next 9/11, it's "stopped". It's completely possible.

Re: Yann LeCun: AI one-percenters seizing power forever is real doomsday scenario

#157

Post by troupo »

tluyben2 wrote: post #128
troupo wrote: post #123

Earlier quoted context omitted.

> The laws we have already cover malevolent and abusive uses of software applications. They don't. Or not nearly enough. Otherwise you wouldn't have automated racial profiling, en masse face recognition, credit and social scoring etc. And it's going to get worse. Because AI is infallible, right? Right?! That's why, among other thing, EU AI regulation has the following: - Foundation models have to be thoroughly docume…

This type of regulatory would be better than 'stop' or making all kinds of expensive rules that only big corps can follow.

There are still quite a few rules, but they seem common sense: https://softwarecrisis.dev/letters/the-truth-about-the-eu-ac...

Re: Yann LeCun: AI one-percenters seizing power forever is real doomsday scenario

#158

Post by VaxWithSex »

logicchains wrote: post #114
PeterStuer wrote: post #100

Earlier quoted context omitted.

Would that still not in the above toy example lead to 0$ vs 2.000.000$/(some factor)?

Yes, but it's not going to have any meaningful impact on the overall gdp because it couldn't happen at scale (requires one party to have 1.000.000$ to spend in the first place).

All money is fiat.

Re: Yann LeCun: AI one-percenters seizing power forever is real doomsday scenario

#160

Post by RandomLensman »

troupo wrote: post #153
RandomLensman wrote: post #139

Earlier quoted context omitted.

AI not being allowed in high-risk applications is a bad idea, same for explicable in human terms (cannot do that now for human decisions anyway). The latter would kind of rule out any quantum computing, too, as the only real explanation there is through maths, not words.

> AI not being allowed in high-risk applications is a bad idea wat > same for explicable in human terms (cannot do that now for human decisions anyway). What? Human decisions can be explained and appealed. Good luck proving something when a blackbox finds you guilty of anything.

You can have ex-post narratives for what humans do, but that is not the same as an actual explanation. Actual human cognition is a black box as of now. The issue of appeals processes is totally separate. That doesn't mean we handover everything to AI, but the idea that we understand human decisions cannot be the reason for that.

For some high risk things like controlling certain complex machinery or processes AI might indeed be needed because control is beyond human understanding (other than in the abstract).

Post reply on HN