I genuienly struggle to think of a kind of economic thing that better qualifies as a public utility.

Completely agree that this kind of technology, with such broad and immense ongoing, as well as potential implications, cannot be allowed to be directed by the whims of wealthy capitalist conmen rent seekers.

If OpenAI fails, and the government owns half of it...

The assets and power the government acquired for no monetary cost simply becomes zero.

The other part of simply seizing half the shares is that the government (presumably a number of ministers/officials in charge of the new sovereign wealth fund) now has half the voting power of the entire board.

That is a pretty direct way to wield influence as to the decisions the company can make, how the CEO can behave.

You want maybe the accounting to actually deprecate the GPUs they have or lease over a realistic timeframe, instead of a totally bullshit one?

Half of your shareholders now demand this.

C Suite refuse to comply?

Begin the process of firing them.

What liability?

What does that even mean?

If I own a stock, a share of a company... what does that make me liable for, just owning the stock?

Do you think that like, if I own a stock, and the company gets sued or something... I personally have to help the company pay out the damages the company is liable for?

No that's not how stocks or the law works.

If the company's stock decreases in value after people hear about it having to pay out from a court case, then yeah the stock has less value now... but the government seized the stock, it got it for $0 dollars.

So there's no net loss.

Well.....

Very broadly speaking, yes, having a much more direct regulation of how you can use 'AI', and how you can 'generate' it, for lack of a better term...

Yeah. AI isn't totally useless. It can do a lot of very neat and useful things.

But, if used improperly, it can literally destroy the world's software systems.

Just in the last 24-48 hours, Meta put an AI in charge of its customer support for Instagram.

People figured out that you can just use a VPN to set your IP to the approximate location of any Instagram account, tell it 'oh i lost my email, heres my new one, can you send me a password reset email?'... and it would just do it.

The AI got wired in to the backend of Instagram's security systems... and then used them to bypass them, because it wss asked to, nicely.

You'd have to have actual experts, not in AI, but in software engineering, systems architecture, experienced full stack senior devs, make up like a council of mandatory safety practices, things that LLMs should never be allowed to be plugged in to.

The... original idea of like a decade ago, was that the people developing this ... would do that, they'd self regulate.

Then they decided they'd rather be trillionaires.

And also, a whole lot of people are the boat you're in, not understanding much about AI... but they're also delusional narcissists who are highly susceptibls to being upsold by people they view as more successful than them.

So those people just hear 'Use AI or be left behind!' and then layoff half the company to afford some kind of not even half baked AI implementation... and the latest numbers are something like 90-95% of firms that adopted AI in the last 12 or 18 months saw 0 or negative overall productivity gain.

... because this shit is not actually magic.

So yeah, yeah, formal regulation, at a very direct snd engaged level... I mean like AI can easily do more monetary damage than a nuclear warhead going off in a major city, if used improperly, by dumb hairless monkeys that want an easy button instead of any actual work associated with running anything.

No, not instead.

In addition to, sure.

Class action lawsuits as a primary counter action to damages?

... why not just ... stop the damages from happening?

If you have 50% of the shares, you have a controlling interest on the board.

You can significantly just tell the company what to do, what policies to have.

Going through through the courts to compensate the damage after the fact is massively more expensive and time consuming... its hitting someone with your car and they paying their medical bill 2 years later.

Why not just not hit them with the car?

Stakes are reusable.

As the French showed if enough heads roll you may end your monarchy. Might flirt with a dictator in the process, but that can also be taken care of

Controlling something worthless is pointless anyway.

And giving it away to everyone will score political points.

No AI company will ever make a profit.

That’s no excuse for you to forget that human rights is our ultimate goal.

Never take your eye off the ball. Vote for humans. Fight for human equality. 🏳️‍🌈

Absolutely

Yes, that was the joke

Perhaps the sarcasm was lost in the comment somehow?

I just corrected a misspelling in my original comment. Perhaps that clears things up?

They're very much not. They have high valuations, but very few employees. Very different to the banks (where the public would lose money) and the car firms (who employed large numbers of workers)

They already are! The government signed exclusive military contracts with OpenAI to develop autonomous weapons. Half of these AI companies are being propped up with taxpayer dollars in the form of government contracts or subsidies, or the promise that there will be a taxpayer revenue stream coming their way in the near future.

Are you new here? The government will absolutely compensate the companies when the bubble bursts.

The point being, if the public owns half of the company but the company is failing, now the public has acquired 50% liability of a dying business.

Like others said though, I'm sure taxpayers will be on the hook either way.

Join lemmy.ml doesn't fill me with confidence that that would work out well.

Wait, so the workers would control the things that let us produce value?

Sounds familiar somehow . . . .