First stop on the road to regulating AI? Finding humans to do the job.

Matt Levin Mar 25, 2024
Heard on:
HTML EMBED:
COPY
Because generative AI is so young, there isn’t a huge talent pool for regulatory offices to draw from. And big tech companies can pay way more than governments. Clement Mahoudeau/AFP via Getty Images

First stop on the road to regulating AI? Finding humans to do the job.

Matt Levin Mar 25, 2024
Heard on:
Because generative AI is so young, there isn’t a huge talent pool for regulatory offices to draw from. And big tech companies can pay way more than governments. Clement Mahoudeau/AFP via Getty Images
HTML EMBED:
COPY

Earlier this month, the European Union passed the world’s first major legislation to regulate artificial intelligence, the AI Act.

The law bans employers from using AI to read worker emotions, requires some AI-generated images and audio to be labeled as computer generated, and compels high-risk AI systems to be safety tested.

Passing those rules was hard. But arguably harder for the EU, the United States or any government trying to regulate AI? Finding the humans who will enforce those rules and prevent all those nightmare AI scenarios we keep hearing about from actually happening.

The newly created European AI Office recently hosted a virtual after-work event for tech talent, trying to fill dozens of technical positions in the next few months.

The computer scientists, software and data engineers, and hardware experts it’s targeting are the folks who will help test whether the most cutting-edge AI systems in the world are biased or can offer the recipe for a biological weapon.

The webinar suffered some technical difficulties, freezing twice. And sure, glitches happen to everybody, although maybe less so for the OpenAI hiring presentation.

Gerard de Graaf, senior EU envoy for digital to the U.S., said we shouldn’t read too much into the those glitches.

“I like to think it was because maybe the session was oversubscribed, or blame it on the video conferencing facility,” de Graaf said.

He said the perception that governments aren’t great at tech isn’t really a hindrance, at least for Europe.

The big obstacle in luring the young Stanford and Berkeley grads he’s trying to recruit is the obvious one.

“The money. If you’re offered, like, maybe a couple of hundred thousand dollars from an AI company here in Silicon Valley, that’s hard for us to compete on,” said de Graaf.

It’ll vary by position and experience, but de Graaf said the beginning salary for a junior level EU technologist with a master’s degree would be between $65,000 and $80,000.

The Googles and Metas of the world can offer a lot more, which is why de Graaf appeals more to heartstrings than wallets.

“If you want to have an impact on society, if you want to be able to tell your parents and in the future your kids, ‘Look, I was there, I made sure that we put AI to good use,’ then to work in the EU AI safety office is a fantastic opportunity,” de Graaf said.

Because generative AI is so young, there really isn’t a huge talent pool regulatory offices can draw from yet.

“I think it’ll be a few years till you have a surplus of labor here,” said Jack Clark, co-founder of the AI research company Anthropic. Clark is also a member of the federal government’s National AI Advisory Committee.

Clark said the good news is that for plenty of AI safety jobs, you don’t necessarily need a Ph.D. in computer science or math. It’s less hacking and more the ability to trick machines into doing things they’re not supposed to do — like making that biological weapon.

“If I’m testing out an AI system, I basically come up with prompts or tests that are mostly written into plain text,” Clark said. “You don’t need very, very rare elite-tier qualifications.”

If you look at the EU or U.S. federal government’s job postings around AI, you’ll see lots of openings for technical talent. people good with computers

Considering how many parts of life AI will and already does touch, the public sector should also be looking for more social scientists, like economists and anthropologists, said Brandie Nonnecke at the University of California, Berkeley, Goldman School of Public Policy.

“We cannot regulate or oversee this technology just from the technical side,” Nonnecke said. Good regulators shouldn’t just be experts on AI. They also need to be experts on humans. 

There’s a lot happening in the world.  Through it all, Marketplace is here for you. 

You rely on Marketplace to break down the world’s events and tell you how it affects you in a fact-based, approachable way. We rely on your financial support to keep making that possible. 

Your donation today powers the independent journalism that you rely on. For just $5/month, you can help sustain Marketplace so we can keep reporting on the things that matter to you.