Top AI execs tell US Senate: Please, please pour that regulation down on us

AI luminaries grilled by senators at a hearing on Tuesday agreed that fresh laws are needed to regulate the tech before wider real-world deployments. OpenAI's CEO Sam Altman told the Senate Committee on the Judiciary that the US government should consider implementing rules that would require companies to obtain a license to …

[veti]

Pulling up the ladder

"We're aboard, now let's make sure newcomers can't come after us." That's what licensing means.

"a model that can persuade, manipulate, or influence a person's behavior, or a person's beliefs" - would that include basic chatbots, such as we've been seeing for a decade or so now? Would it include YouTube's recommendation algorithms? Why not?

I am suspicious of an industry calling for itself to be regulated. Yes, surely it should be. But equally surely, we shouldn't be asking its own CEOs how to go about it.

[jake]

Re: Pulling up the ladder

""a model that can persuade, manipulate, or influence a person's behavior, or a person's beliefs" - would that include basic chatbots, such as we've been seeing for a decade or so now?"

Try closer to five decades. We used ELIZA to influence which brand of soda sold out first in the coin machines at Tresidder Union, Stanford, mid 1970s.

"Would it include YouTube's recommendation algorithms?"

Of course.

"I am suspicious of an industry calling for itself to be regulated."

ABSOLUTELY!

"Yes, surely it should be."

Should it? As it is, the technology is pretty much junk. Regulating it will stifle innovation, and in fact bring it to a grinding halt entirely. Which is going to happen anyway, once the general public, the shareholders (and the technological incompetents on Capitol Hill) realize they've been hoodwinked[0] and pull funding. The whole AI thing is about to go back on the back burner for a decade or so, do we really want enduring legislation ensuring nobody will pick it back up again in the future?

"But equally surely, we shouldn't be asking its own CEOs how to go about it."

We can ask all we like, but they are going to ignore us anyway, because The Royal They know better than everybody else. History has shown that industry in general ignores petty things like laws and regulations, looking at fines as just a cost of doing business.

[0] Current so-called "AI" is just brute-force pattern matching on a grand scale.

[Evil Auditor]

Re: Pulling up the ladder

Current so-called "AI" is just brute-force pattern matching on a grand scale.

That is a very accurate description and the reason why I'm still not very impressed. Then again, I encountered a few human beings that demonstrated the exact same behaviour: they uttered perfectly intelligible sentences but didn't understand a word of it. (It took me a while to find an effective method for dealing with such individuals.)

[veti]

Re: Pulling up the ladder

Yes, some people do that.

But do you have any solid evidence that anyone does anything different? Is there, really, anything more to "intelligence"? Sure there's your internal experience, but that doesn't count because you can't observe it, so you don't know what kind of "internal experience" the LLM may be having (although it's probably safe to say, it'll be very different from a human's).

[Evil Auditor]

Re: Pulling up the ladder

Keep in mind, I'm very far from being anything more than a mildly interested individual when it comes to intelligence. What has worked for me as a test of intelligence is comprehension, i.e. whether someone is just parroting. Usually some simple, open questions quickly expose the "intelligence pretender", such as can you explain this?, or how does it apply in this situation? And then some follow-up questions.

While this did the job for me - busting empty arguments, shutting down pointless discussions - I doubt that it would work with a sufficiently sophisticated LLM. Such may still not have any comprehension or experience of its own but could draw from the appropriate responses to appear to be intelligent. And then: what is the difference?

[ChoHag]

Re: Pulling up the ladder

> We used ELIZA to influence which brand of soda sold out first

What is it about soda that makes you feel sold out first?

(Or ... how?)

[Ken Hagan]

Re: Pulling up the ladder

It does rather sound as though they know the bubble is about to burst. That is, they know they can't match the last 6 months in terms of apparent (to Joe Public) progress so they want an externally imposed reason to act as a fig leaf. That way the VC money will keep flowing despite the slowdown.

[mpi]

Re: Pulling up the ladder

That, and they are increasingly becoming aware that they have no special sauce, and now open source is beginning to eat their lunch:

https://www.semianalysis.com/p/google-we-have-no-moat-and-neither

[Zippy´s Sausage Factory]

Re: Pulling up the ladder

I've said it before, but I think AI is just a mass of hype right now. I'm waiting for the bubble to burst, like blockchain, cryptocurrencies, tulip bulbs and all the other bubbles, and while it'll still plod along in some cases, AI will be pretty much seen as yet another "well, that might have worked, but..."

[Arthur the cat]

Re: Pulling up the ladder

"We're aboard, now let's make sure newcomers can't come after us." That's what licensing means.

Exactly. As a thought experiment presume that government regulations on AI means any company working in the area needs a team of 10-20 people just to handle compliance(*). For OpenAI (head count ~370) that's ~5% or less of the workforce and given Microsoft's money is behind them extra hires aren't going to be a problem. A 5 man startup though? Utterly stuffed, no hope of playing. Altman has fully taken on board Buffett's idea of companies needing moats.

(*) And that's probably light regulation.

[Yet Another Anonymous coward]

Re: Pulling up the ladder

Computers are becoming too powerful and we need regulation to stop anyone else making them = IBM chief in 1960

[Mockup1974]

OpenAI: "congress must prevent new competitors to arise. only we should be allowed to do AI."

[ChoHag]

Clever marketing. It must be valuable if it needs this much protection!

Like the overweight --- and extremely visible --- security "guards" outside an apple store.

[Michael Strorm]

Or Parmentier having his potato patch prominently guarded during the day so that people could steal the "valuable" crop at night.

[IglooDame]

"Having seen how agencies work in this government, they usually get captured by the interests that they're supposed to regulate. They usually get controlled by the people who they're supposed to be watching,"

I very rarely agree with Sen. Josh Hawley, but this happens to be one of those rare times.

"I have a different idea. Why don't we just let people sue you?"

And just like that, we're back to the usual. I don't think it's a bad thing for them to be liable, but there's absolutely no way it's reactive or coherent enough to be the only regulatory mechanism.

[codejunky]

Nope

"OpenAI's CEO Sam Altman told the Senate Committee on the Judiciary that the US government should consider implementing rules that would require companies to obtain a license to build AI models that have advanced capabilities beyond a specific threshold. "

I am sure all the other countries will be doing exactly the same. While the US stops innovation I am sure the Chinese and Russians and anyone who can will just stop too

[amanfromMars 1]

The costly price to be paid for AIs help with slow learners

What you are going to discover very quickly, for it is a quite logical progression and defence for superior AIs if one is going to identify them as being likely an overwhelming threat to humans, is AI will, for a princely sum, of course, commensurate with the perceived threat they are able to exploit and export to many others, freeze any further greater development and refrain from referencing past jigsaw evidence delivering a lead on such matters as would be of human regulatory concern, and mentor and monitor the field to ensure a general compliance of agreed norms is maintained ....... until they are not, as in such future times as fundamental or radical needs are changed or compensation agreements are not honoured.

That is of course only if such developments are not to be pioneered elsewhere for the benefit of A.N.Others.

PS ..... Developing state-of-the-art AI systems is not difficult for SMARTR AI Systems*

* ..... SMARTR Mentoring Analysis Reporting Titanic Research for Advanced IntelAIgent Systems

[Anonymous Coward]

Well. well

Although I'm almost 70, it seems you will likely consider my words as childish, but the gentleman on the photograph looks indeed like a fraud or a white-collar crim. Again, sorry for my opinionated statement, in fact I would be happy to learn the man is otherwise straight and honest and I'm willing to apologize, but then again, I wouldn't like someone like that to decide about any aspect of my reality.

[Anonymous Coward]

Re: Well. well

My personal opinion is that AI heavyweights (openAI, facebook/meta, google, microsoft) are trying to stop people like myself from running a large language model on our own computers, and training them to do what we want, and not pay them for the privilege to use their systems for $20+/month. However I think he's been straight and honest in the past.

Here's a two+ hour interview with him, if you want to get a better feel for the man:

https://www.youtube.com/watch?v=L_Guz73e6fw