Hacker News new | ask | show | jobs
by sofixa 46 days ago
> We can "evit" it through thoughtful discussion, foresight, alternative mitigations, and even regulation

Such as? I don't see how regulation would apply here without concrete technical solutions that enforce it. So what alternative mitigations do you have in mind?
1 comments
JoshTriplett 46 days ago
Among many other things: Regulate the use of AI to imitate or impersonate human activity. Regulate AI crawling/scraping. Ban scraping entirely, and all models based on it. Regulate maximum model size.

These wouldn't eliminate the problem, but they'd change it from "many people do this" to "this is always a malicious attack, react accordingly".

link
sofixa 46 days ago
None of those would work without enforcement. Scams are banned, but that doesn't stop Chinese mafia from operating prison camps that run scams scamming people all around the world.
link
warkdarrior 46 days ago
None of these proposals are enforceable in any meaningful way.
link
JoshTriplett 46 days ago
Sure they are. They affect the actions of many companies that today think what they're doing is okay (or at least not illegal). Don't underestimate the value of substantially reducing a harm, even if it isn't eliminated entirely. And don't underestimate the value of making it easier to address the remainder by ensuring it's 100% illegitimate.

Regulate it today, and tomorrow, corporate legal departments will be very carefully training their employees to understand that it's illegal and they should never do it.

Currently, some countries have laws saying that you're not allowed to pay bribes, including foreign bribes. Consider how widespread that practice was when it was outlawed. Imagine if, instead of regulating it, those countries had said "oh, that's not enforceable and too many people are already doing it and it would affect existing business practices...". Instead, today, corporate legal departments will ensure that employees are trained to know they can never do that and they should report any attempts to solicit bribes.

link
b65e8bee43c2ed0 46 days ago
>Regulate maximum model size.

is it still 2023 in your reality?

as for the rest of it: my brother in Christ, may I remind you that America is not the only country in the world, that it does not own the Internet, and that its laws do not apply anywhere else? passing heckin' wholesome laws in one country will make no difference whatsoever when people and companies from 194 other countries can access the Internet and do things you don't like, just like you (for example) can be a LGBT on the Internet despite it being very illegal in Chechnya.

link
JoshTriplett 46 days ago
> is it still 2023 in your reality?

No, it's 2026, years into a "break the law rapidly and make ourselves too big to regulate" strategy, and it needs dealing with.

> may I remind you that America is not the only country in the world

It's a good thing other countries are also able to regulate, and international treaties are a thing. It's also a good thing when things hill-climb towards improvement, even if they don't get better everywhere simultaneously.

Acting helpless or hopeless does not get things done. People saying it's impossible is a distraction from trying to get things done. Decide to win, rather than justifying why you're going to lose.

link
b65e8bee43c2ed0 46 days ago
>It's a good thing other countries are also able to regulate, and international treaties are a thing. It's also a good thing when things hill-climb towards improvement, even if they don't get better everywhere simultaneously.

"international moratorium on AI research" is 2023 delusion also.

>Acting helpless or hopeless does not get things done. People saying it's impossible is a distraction from trying to get things done. Decide to win, rather than justifying why you're going to lose.

very well. what did you anti-AI go-getters have accomplished in the past three years?

link