One could rightfully ask how seriously you can take a senate hearing when the person supposed to testify is allowed to dazzle its members with what Emily Bender calls a Magic Show first.
CNBC:
Sam Altman Wows Lawmakers at Closed AI Dinner: “Fantastic…forthcoming”
OpenAI CEO Sam Altman spoke to an engaged crowd of around 60 lawmakers at a dinner Monday about the advanced artificial technology his company produces and the challenges of regulating it.
The dinner discussion comes at a peak moment for AI, which has thoroughly captured Congress’ fascination. […] “He gave fascinating demonstrations in real time,” Johnson said. “I think it amazed a lot of members. And it was a standing-room-only crowd in there.”
And of course it gets worse from there.
During the hearing of the Senate Judiciary Subcommittee on Privacy and Technology itself, Sam Altman ostensibly calls for regulation, licensing, and oversight. But what he’s really asking for is a monopoly for OpenAI.
- Form a new government agency charged with licensing large AI models, and empower it to revoke that license for companies whose models don’t comply with government standards.
- Create a set of safety standards for AI models, including evaluations of their dangerous capabilities. For instance, models would have to pass certain tests for safety, such as whether they could “self-replicate” and “exfiltrate into the wild”—that is, to go rogue and start acting on their own.
- Require independent audits, by independent experts, of the models’ performance on various metrics.
Translation:
Please create a new agency to oversee and license large AI models that isn’t the FTC, because the FTC has already made it repeatedly clear that they will clamp down on actual risks and what makes LLM technologies actually dangerous, concerning consumer safety, unsubstantiated claims, fraud, discriminatory results, training data transparency, privacy concerns, using artists’ work without permission (aka theft), and so on and so forth.
So no, not the FTC! Not the agency that’s been called “Washington’s most powerful technology cop!”
Instead, please create a new, better agency that will concern itself with imaginary threats like “self-replicating rogue AI acting on its own.” An agency that audits “performance on various metrics” and sets standards for AI in the same way governments “regulate nuclear weapons.” An agency, moreover, that gives out government licenses and makes it prohibitively expensive for potential competitors to enter the market. And, for good measure, please stack this new agency with our own experts, if that isn’t too much to ask.
It’s true, there’s the general problem of recruiting experienced professionals for oversight agencies that are not deeply in cahoots with the industry they’re supposed to oversee—a problem that, for example, has most visibly plagued the Federal Communications Commission since forever. And it’s not as if Altman were in any way subtle about it:
Still, many of the senators seem eager to trust Altman with self-regulation. Sen. John Neely Kennedy (R-La.) asked if Altman himself might be qualified to oversee a federal regulatory body overseeing AI. Altman demurred, saying he loves his current job, but offered to have his company send a list of suitable candidates.
You can’t make this up.
For some senators, at least, having a giant sign nailed on your forehead that says “Hello I’m a Fraudster and Con Man!” reliably counts as a stellar letter of recommendation.