‘Worried’ Warren Buffet compares AI to the creation of the ‘atom bomb’ | Hacker News Reader

‘Worried’ Warren Buffet compares AI to the creation of the ‘atom bomb’

nypost.com

9 pointsmikece6 comments

Warren Buffett is a brilliant investor, not a domain expert in AI. He’s just out of his depth here and his opinion doesn’t mean much.

When people talk about AI they often mean general AI which doesn’t exist and I think is a lot further away than many people are suggesting. I’m a middle aged software engineer, but I think it’s quite possible I don’t live to see it. Maybe not the most likely outcome, but definitely possible that it’s much harder than we think.

But general AI would be a titanic sea change, a singularity beyond which we cannot see. It would at minimum be massively disruptive and at maximum be the end of the human race.

>Warren Buffett is a brilliant investor, not a domain expert in AI

You don't have to be an AI domain expert to be able to have a good insight on what the AI can be good and bad for, work, economy, or society wise.

Similar how you don't have to be an expert on nuclear physics to be able to understand the impact of the nuclear bomb on the world.

I'd say that the last people who would be able to good unbiased advice on the results of AI use, are the technical experts on AI.

This is just ignorant. You have to at least understand what it does, not some made up pop culture version. Bombs are easy to understand. Almost nobody (if anybody) has a clear handle on what "AI" does, as in what it's actual long run use or effect is. There are lots of dumb analogies (read "prediction machines" for a particularly shitty take) but nothing with any demonstrated predictive power.

People who understand the technology at least can descriptively tell you what it does, which is a lot more useful than what laypeople think based on whatever populist media they've consumed.

What people who understand a technology at the technical level expect it to be used for and how, has historically not been very relevant to what the technology ends up being used for.

What it does at a technical level (better described is "how it does what it does"), which is what the experts on AI are experts at, is not that important.

What you need to know to have a chance to see its future impact is a very high level understanding of what it does on a non-technical level, and how better it could do it in the next 5-10 years. Which doesn't require a PhD on AI at all. Not even a BSc.

See how many top computer scientists missed where the PC and the web revolution would end up and what it would allow. Even Gates missed the important of this internet thing. Where someone not that technical like Jobs could get it much better...

https://www.theverge.com/2014/12/15/7396569/Steve-jobs-inter...

https://www.businessinsider.com/nine-predictions-about-the-i...

In the long run, the most obvious effect will be an abdication of responsibility by the organizations that employ it. By superficially removing the humans from the decision making process, they also remove said humans from personal and professional liability. We can already see this with examples like Palantir.

This would also legitimize fraud, and other unfair business tactics which are currently illegal under the law but are not being enforced under the reasoning that if an algorithm does it its not what it is. So people can design algorithms to do it and claim it wasn't intentional (when it was), how can you prove a negative?. IMO, Anyone who approved the algorithm for such use constitutes approval of the practice, even when it wasn't known, they should have had constructive knowledge to know about it as part of their duty of care.

Things like bait and switch will be called something else like 'dynamic pricing'.

Oh wait that's already being done ... (see amazon).

Then there's the societal mechanics, what 'good' happens from stalling the economic engine (i.e. eliminating the bottom 2/3 of the factor markets)?

Also don't forget, violation of constitutional rights, under false positives, and the crazy maker algorithms that drive insanity (social credit like).

The requirements to use AI are negligible compared to an atom bomb. Comparing it thus is a poor comparison because its so much worse.

Reply on news.ycombinator.com