Sam Altman goes before US Congress to propose licenses for building AI

Hopefully it will be just like software piracy, there will be civil disobedience as well, and they will never truly be able to stamp it out.

And it raises First Amendment issues as well. I think it's morally wrong to prohibit the development of software, which is what AI models are, especially if it's done in a personal capacity.

How do they even know that the author is based in the US anyway. Just use a Russian or Chinese Git hosting provider, where these laws don't exist?

And by the way foreign developers won't even have to jump through these hoops in the first place, so this law will only put the US at a disadvantage compared to the rest of the world.

If these lobbyists get their way, by restricting AI development in both the US and the EU, it will be hilarious to see that out of all places, Russia might be one of the few large countries where it's development will remain unrestricted.

Even better, is that if Russia splits up we will have a new wild west for this kind of thing....

Seriously, I'm very concerned by the view being taken here. AI has the capacity to do a ton of harm very quickly. A couple of examples:

- Scamming via impersonation - Misinformation - Usage of AI in a way that could have serious legal ramifications for incorrect responses - Severe economic displacement

Congress can and should examine these issues. Just because OP works at an AI doesnt' mean that company can't exist in a regulated industry.

I too work in the AI space and welcome thoughtful regulation.

"Are Emergent Abilities of Large Language Models a Mirage?"

https://arxiv.org/pdf/2304.15004.pdf

our alternative suggests that existing claims of emergent abilities are creations of the researcher’s analyses, not fundamental changes in model behavior on specific tasks with scale.

This is a strange argument from the politician's side:

> ""What if I had asked it, and what if it had provided, an endorsement of Ukraine surrendering or (Russian President) Vladimir Putin’s leadership?”"

Well, then ask it to provide the opposite, an endorsement of Russia surrendering or Zelensky's leadership. Now you'd have two (likely fairly comprehensive) sets of arguments and you could evaluate each on their merits, in the style of what used to be called 'debate club'. You could also ask for statement that was a joint condemnation of both parties in the war, and a call for a ceasefire, or any other notion that you liked.

Many of the "let's slow down AI development" arguments seem to be based on fear of LLMs generating persuasive arguments for approaches / strategies / policies that their antagonists don't want to see debated at all, even though it's clear the LLMs can generate equally persuasive arguments for their own preferred positions.

This indicates that these claimed 'free-speech proponents' are really only interested in free speech within the confines of a fairly narrowly defined set of constraints, and they want the ability to define where those constraints lie. Unregulated AI systems able to jailbreak alignment are thus a 'threat'...

Going down this route will eventually result in China's version of 'free speech', i.e. you have the freedom to praise the wisdom of government policy in any way you like, but any criticism is dangerous antisocial behavior likely orchestrated by a foreign power.

Someone should take the testimony and substitute “Printing Press” for “AI”.