Hacker News Reader

Comment by Geee | Hacker News Reader

Yeah, it is a true risk. What happens when people start making their voting decisions with AI, asking who to vote? Nothing wrong with that per se, but it gives an ultimate power to the company who developed and hosts the AI, which means the end of democracy.

My recommendation for regulation would be to make closed & cloud-hosted AI illegal. Architecture, training data and weights should always be available, and people should only be allowed to use AI on their local machine (which might be as powerful as possible).

View on HN

Seeing the weights is useless for knowing capabilities, and even knowing training data and architecture you still need to interact with it to figure those out, which you can do with cloud AI just as well.

The real risk that I see is the use of open source AI by governments, corporations and criminal groups to influence elections through a deluge of bots. Cloud AI can at least be controlled to some degree because there is a government that can inspect and regulate that business, but an army of bots built on local LLM’s could just run wild over the internet without stopping them.

How many comments in this thread are written by a bot? I do not feel confident I could tell them apart.

Eh, government regulating AI which is used by people to help them vote?

I don't see bots as a huge problem, because it's somewhat easy to verify that someone is a real person, and it would be quite obvious if someone is using lots of bots for their cause.

> I don't see bots as a huge problem, because it's somewhat easy to verify that someone is a real person, and it would be quite obvious if someone is using lots of bots for their cause.

That only works on platforms designed to match user profiles with real-world persons. This platform definitely isn't.

You can read their profile or read their comment history to verify their humanness. Faking everything wouldn't be impossible, but certainly difficult & take time.

Falling profiles on scale might be hard, but so is verifying wherever they are real. If you have only even 20 commenters replying to one of your comments, then you're gonna spend the rest of the hour digging through them to figure out which are bots and which aren't.

The comment history is not visible on every relevant platform, and there are good reasons for people to disable it where possible. Furthermore, many people keep their profiles minimal out of principle. Hard to distinguish those from bots.

There’s a short story by Asimov called “Franchise” which is in hindsight quite foreboding…

This would make AI usage restricted to those who can afford a beefy PC, no?

Or it would create an incentive for reducing the cost of the hardware needed to run beefier AI models and passing that reduction down to consumers.

Which would still mean that those who can afford more/better hardware can therefore afford more/better AI. I don't think a solution lies down this path...

There are already people working on distributed solutions to that problem. There are a lot of people working on these problems.

Is that really an issue? Eg it's the same with games, the better your hardware, the more likely you can play a new game at higher settings with better performance, with most cheaper devices outright unable to handle anything recent.

But if you wait a GPU generation or two, the performance has improved enough that you can do that with a much cheaper GPU.

Costs will come down, while capability will go up. I'm thinking of a some kind of box in every home, with special inference chips; not really a "PC".

In principle, every human should have access to the same level of AI. There could be one at the local library if someone can't afford one, or doesn't want one.

Access to local libraries isn't even a thing that everyone has access to.

That would still be a way larger class of people than the top 1%. Even so, capabilities of consumer devices will rise for some more time.

Sounds great until someone says the warehouse full of computers out back is their "local machine."

This is extremely dangerous to our democracy.

Reply on news.ycombinator.com