The Midterms and the Price Already Paid
This post is part of my Medium blog.
On Tuesday the country voted. By Wednesday morning it was clear what happened: Democrats won control of both the House and the Senate. Flipping one chamber in a midterm is normal. Flipping both in the same cycle has happened twice in the last fifty years. Whatever else you think about the result, one thing is settled. The legislative branch changed hands.
I write fiction about the near future, so I pay attention to this differently than most people. The Condition Set is set close enough to now that elections land inside the forecast window. And one of the running threads in these books is that Canada gets ahead of the United States on AI. Not banning it, not pretending it will go away. Regulating it like the serious industrial activity it is, quietly and a little ahead of us, because that's how Canada tends to handle things. I wrote that as texture. Every month it looks less like texture.
So the question people have been asking me this week: does Tuesday change the books? Do I need to revise?
Here's the honest answer. Maybe, but not yet.
The change that matters is this. There is now a Congress that could actually take up AI regulation. Before Tuesday that was theoretical. Now there are committees with jurisdiction, chairmen with gavels, and a two-year clock. Do I have more confidence that Congress takes some rational steps toward regulating AI? Somewhat, yes. I think regulation is necessary. I also think the pressure that produces good regulation rarely comes from Congress on its own. It comes from something breaking loudly enough that doing nothing costs more than doing something.
And some of what would need regulating has already been decided. That's the part I keep coming back to.
Look at the debt. The AI buildout, the data centers, the power plants, the chips, all of it is being financed on a scale that assumes the good outcome happens. This week Oracle announced a twenty billion dollar debt and stock sale and the stock fell nine percent anyway. AMD dropped seven and a half. Marvell fell thirteen after an all-stock deal. The Nasdaq had a four percent day on Friday. And Jim Cramer spent his evenings on air explaining what a real market bottom looks like, waiting for what he calls a crescendo, the punishing session where sellers finally give up. When the loudest bull on television is narrating a capitulation pattern, the story has left the financial press and entered the regular conversation. The market will probably correct for this. Whether it's a sudden break or a slow grind, I don't know. Nobody does. But the borrowing happened first and the questions are arriving second.
That's the thing about some decisions. They get made in the middle of a boom and by the time anyone votes on the consequences, the consequences are already load-bearing. You can change the legislative branch on a Tuesday. You cannot vote your way out of a balance sheet in one election cycle. Some of the price for this buildout has already been set, and it will be paid on schedules that don't care who holds the gavel.
I want to be clear about what I'm arguing for, because it isn't much and it isn't dramatic. I'm not advocating hugely restrictive regulation. I don't want a permit office standing between the country and the most useful technology invented in my lifetime. What I want is what we failed to do with social media. These systems are going to have a dramatic impact on human beings, on work, on what kids grow up believing is normal. With social media we chose to find out what happens first and regulate later, if ever. It cost us roughly fifteen years and I don't think we've finished paying that bill either. AI gives us a shorter window and a bigger blast radius. Something more effective than hope is the minimum.
Am I revising the novels? Not yet. The world I wrote still holds. Canada still reads as the sensible one. The debt still reads as the bill coming due. But here's the commitment I'll make, mostly to myself: as the near future I predicted keeps changing, and it will, I'll update the books. Near-future fiction is a prediction with a publication date, and predictions should get revised when the world answers.
The voters changed one branch of government on Tuesday. The debt is still sitting there. A lot of it was signed before anyone voted.