12 comments

[ 3.2 ms ] story [ 26.2 ms ] thread
Regulating away the competition is one way to win the market.
> Congress should make it illegal, the report recommends, to train AI models using more than a certain level of computing power.

This is the most lazy 'research' I've seen so far to bring answers to regulation.

> It was written by Gladstone AI, a four-person company.

Now that makes sense.

But seriously, how it the Government supposed to regulate how much computing power you use, its a ridiculous idea.

> The recommendation to outlaw the open-sourcing of advanced AI model weights, they expect, will not be popular. “Open source is generally a wonderful phenomenon and overall massively positive for the world,”

Honestly the proposed solutions sound like the first brown bag session of ideas in the lunch room. If I was the government who commissioned this, I would want my money back.

licensure
Hmm, well intentioned but not a valid solution of course to any global technology, given the level of lobbying that the AI front-runners already do.. I don't see that inspiring people to even try in the US.

Probably moving to other markets to get ahead without any interference from the US Fed. (They actually correctly do touch on these topics in the article)

To be clear, I am not advocating it, I just expect that as a resolution.
> But seriously, how it the Government supposed to regulate how much computing power you use, its a ridiculous idea.

Same way we saw runaway power consumption from Bitcoin miners. Every gold rush brings with it pressure on resources and both direct and indirect impact on others.

> federal contract worth $250,000, according to public records. It was written by Gladstone AI, a four-person company that runs technical briefings on AI for government employees

$250000 feels like way too little to get the incentives right. A tiny fraction of what $bigtech can pay them to draft up a report that specifically benefits them.

Also, I wonder how these companies are chosen by the government.

Reminder that the US government has a public RfC on widely available model weights closing at the end of the month. If you don't want unserious ideas like these dominating the conversation please remember to provide your input: https://www.regulations.gov/document/NTIA-2023-0009-0001
We've seen how it (didn't) with net neutrality. No reasons to think it'll work this time. Especially when current administration is looking in any way to shift attention from migrants packing in US. They will 'show strength and responsibility' on something else, like leading role in AI regulations.
The only extinction level threat I see with AI is corpo rats being the only ones with a say in how models are trained, how they are to function, and how they are to be used.
Replaced "AI" with "WMDs" and it reads like the 2002 National Intelligence Estimate that was the precursor to invading Iraq