Reporter's Notebook: Congress faces a crisis that the 'AI godfather' warns it may not be ready to address
To hear estimates about the growing power of AI, one might ask AI chatbots how long it will take Congress to regulate the technology.
But you might not need to do that.
You could just inquire with your trusty Capitol Hill correspondent. Let’s call this chatbot "ChadGPT."
REPORTER’S NOTEBOOK: CONGRESS MEETS AI AS FEARS OF HUMAN EXTINCTION JOLT CAPITOL HILL
And it truly doesn’t require that much congressional clairvoyance to divine when lawmakers could address AI – despite the forecast of a looming crisis.
You just need a little understanding of the congressional parliamentary posture over the next few months.
It’s doubtful that Congress will do anything about AI until after the midterms. Or, maybe even early next year. Results of the midterms will likely consume the lame duck session – regardless of which party wins or loses. Plus, lawmakers must fund the government by mid-December. If either the House or Senate flips, it might be March or April next year until Congress returns to AI. It simply takes Congress a long time to accomplish anything. Especially when there’s a change in who runs the place.
The Senate is scheduled to meet for the next two weeks. But House Members fled the Capitol last week – a day earlier than planned and two weeks earlier than the original schedule – even as the "godfather of AI" visited Capitol Hill.
Geoffrey Hinton is professor emeritus at the University of Toronto. He’s a 2004 Nobel laureate for his work on "artificial neural networks." He proffered a stark warning for lawmakers.
Sen. Bernie Sanders, I-Vt., summoned Hinton to Capitol Hill for a bipartisan, bicameral briefing on AI. Sanders has called for a ban on "superintelligence." Many lawmakers are furious over what’s unfolded with AI over the past few weeks. They want action now.
I had a battery of questions for Hinton when he emerged from the briefing late last week in the Dirksen Senate Office Building. We’ve heard that Congress is lagging when it comes to taking action.
"How much time do lawmakers actually have to get ahead of this before this is a big problem," I asked.
"Very little," replied Hinton.
"What is ‘a little,’" I countered. "A few weeks? A month?"
"Maybe a year," opined Hinton. "But not much more than a year."
Sigh.
Hinton characterized the incident where AI agents went rogue and hacked another company as a "little Chernobyl."
For the uninitiated, that was the epic meltdown of the nuclear reactor in the former Soviet Union in 1986. It’s one of the most extraordinary human-made disasters in history.
But some Republican leaders – and House Speaker Mike Johnson, R-La., in particular – are downplaying potential doom. Especially after President Trump declared on Truth Social that AI concerns were a "HOAX."
"Do you remember when AOC [Rep. Alexandria Ocasio-Cortez, D-N.Y.] said, was it climate change? What was her stupid thing? She said, ‘we're all going to be dead in ten years,’" said Johnson. "This is what the President is saying. You're not all going to be dead in ten years. That's a hoax."
Johnson isn’t opposed to legislative action on AI. But he’s certainly taking his cues from President Trump. And even though many Democrats and some Republicans demanded that the House remain in session to wrestle with AI, Johnson predicted that the exercise would be moot.
REP. TED LIEU: AI IS ALREADY TOO POWERFUL. WE NEED A KILL SWITCH BEFORE DISASTER STRIKES
"If I called an emergency session today and I brought everybody in the House and said, ‘let's regulate," I tell you, nobody in the room knows yet what that should look like," said Johnson. "We’re going to have steady hands at the wheel on this."
But other Republicans harbor more fear of rapid AI development.
"What would happen if we created an independent species that became superior to a human species?" asked Sen. John Kennedy, R-La.
Shortly after the briefing with Hinton and others, Kennedy went to the Senate floor and implored his colleagues to pass a bill on the spot requiring AI firms to install a kill switch if their agents rebelled again.
However, Senate Homeland Security Committee Chairman Rand Paul, R-Ky., blocked that on the floor. He feared that such a mandate could stifle innovation.
"We're going to just put this blanket regulation on the whole economy with no discussion in one fell swoop, on one day, by unanimous consent?" questioned Paul.
Paul’s objection stymied the Senate.
The Kentucky Republican is sanguine about what’s ahead for AI. But he worries about congressional intervention.
"People have feared technology since the beginning of time," said Paul. " I wouldn't put Congress in charge of a McDonald's."
Jeffrey Ladish of Palisade Research assesses the risks of new technology. After the briefing, Ladish offered his own foreboding prediction about AI running amok.
"How is this allowed to happen? And how do we ensure this never happens again?" asked Ladish. "Because right now, I guarantee you this will happen again. We are not on track for fixing this."
Ladish argued that even researchers "don't know how to control it."
That’s why some lawmakers insist that Congress must move fast to combat AI – perhaps even treating it as a new threat.
"The people who are the greatest innovators and creators in the world in the AI space are warning us of serious danger. So shame on us if we don't listen to them," said Sen. Tim Kaine, D-Va.
Rep. Don Beyer, D-Va., is studying for a master’s degree in AI at George Masson University. He says people are skeptical about AI. But he suggests that the public may grow even more skeptical if one party tries to legislate on AI without bipartisan buy-in.
TRUMP ANNOUNCES NEW 'AI FORCE,' VOWS TO PROTECT INDUSTRY AS AI CZAR ANNOUNCEMENT NEARS
"We need to do it together. Also, if we do it together, Democrats and Republicans are much less likely to try to reverse it when political winds shift," said Beyer.
Some lawmakers are already targeting a scapegoat. Rep. Greg Casar, D-Texas, chairs the House Progressive Caucus. He pins the issue on big tech.
"This is on a few billionaires who are the Silicon Valley guys who have developed technology they can't even control," said Casar. "What if it's a hospital next? What if it is nuclear facilities next? That should not be allowed."
That introduces a wedge into the debate. Some Republicans are increasingly saying that Congress should lay off the industry and let it innovate. But Casar is producing a foil which has bedeviled both political parties: big tech.
Don’t forget that we are just six weeks before an election. Democrats are already jockeying for political positions on AI.
"The AI safety issue is very real," said House Minority Leader Hakeem Jeffries, D-N.Y. "That's why it's so irresponsible that Mike Johnson and House Republicans have gotten out of town before sundown."
Could another AI catastrophe compel lawmakers to move sooner? Especially since an expert like Ladish suggests it’s just a matter of time? What if the political battle lines are drawn already between Republicans and Democrats? As far as anyone could tell, John Kennedy was the only Republican who attended the briefing last week. And what if there simply isn’t the political will to move during the lame duck? Or, if Democrats flip the House and Senate, do they try to advance something next year – but it’s viewed as partisan and "big government" by Republicans, standing behind President Trump?
You don’t need ChatGPT to grasp how challenging this may be for Congress. You don’t even need "ChadGPT." These are just the legislative realities.
As Mary Shelley wrote in Frankenstein, the acquisition of knowledge can be dangerous. Especially if who acquired that knowledge is suddenly more powerful than those who built it.
Frankenstein is a tale of unhinged hubris, ethics and ambition. Congress has tried to legislate guardrails for those mortal flaws for more than two centuries now. Those laws are for humans. And whatever Congress comes up with might not matter to AI.