PLease Panic responsibly

Two hundred economists just told us to panic responsibly.

That's the vibe of "We Must Act Now," the statement 200+ researchers and economists (including sixteen Nobel laureates, OpenAI's chief economist, Anthropic's chief economist, Jack Clark, Eric Schmidt, Vinod Khosla) released this week. The pitch: AI could remake the economy faster than the Industrial Revolution did, and we're not ready. Anton Korinek, who organized it, put it well: steam and electricity gave us decades to adjust. AI might give us a few years.

I don't disagree with that. I actually think it's true. And reading it, I kept thinking “wait, is this fucking play about us?”

Because it's not hypothetical for this industry. It's already happening in the parts of fintech that used to be someone's whole job. Underwriting used to mean an analyst reading a file. Now a model scores it in seconds and a human reviews the edge cases, if that. Fraud detection used to lean on pattern recognition built up over years on the job. Now it's largely automated, with people stepping in only when the model flags something ambiguous. Tier-one customer support, the first pass at KYC document review, even a lot of first-draft dispute resolution — all of it is being absorbed by AI right now, not in some five-year projection. What's left for people is the judgment layer: the ambiguous case, the relationship, the decision nobody wants a model making alone yet. That layer is real, but it's also smaller than it used to be, and it's not obvious how much smaller it gets from here. Even Daron Acemoglu and Simon Johnson signed onto this statement — two economists who've spent years being the field's default skeptics on AI hype, the ones you'd expect to push back on a piece like this. Acemoglu told the Times he still doubts AI is moving as fast as Silicon Valley predicts, and in the same breath said recent advances have made him more worried about real job losses. I don't read that as a contradiction. I read it as what honesty actually sounds like when you're this close to the industry — you don't get to pick just one side and defend it.

Then there's aslo the part almost nobody puts in the same breath as job displacement: the climate math. The statement doesn't mention it, and I think that's a real gap. Global data center electricity use was around 415 terawatt hours in 2024 (roughly 1.5% of the world's electricity) and it's on track to more than double by 2030, with AI as the main driver of that growth. Most of that power still comes from gas and coal, not clean energy. So the same systems that might be quickly, exponentially reshaping what my job looks like in five years are also drawing down power at a scale that's starting to show up in local grids and utility bills, not just abstract emissions charts. It's not a side issue. It's the same story, just a different receipt. Those same people that are set to lose their jobs, are living in a more expensive world every day.

Here's where I land though: I don't think any of this is optional anymore. AI is inevitable. Not every prediction about it is true — some are, some aren't, and right now we genuinely don't know which is which. But I and most of my peers already reach for these tools every day at work, and pretending we don't isn't really an option for anyone in my field anymore.

And I want to be honest about something else too: I don't think inevitable has to mean bad. I use AI daily, and most days it makes me better at my job, not smaller. Information access alone is a real unlock: the ability to get a credible first pass on something complex, fast, and then spend my actual expertise checking it and pushing it further, instead of starting from a blank page. Used critically, that's not a shortcut around judgment. It's more room for judgment, because the grunt work took less time. And the applications showing up in fintech specifically are genuinely exciting, not just defensible. AI powered subscription management that catches a price hike or a forgotten renewal before it drains someone's account. Underwriting that can incorporate more signal, faster, for people who'd otherwise get a flat no from a model with less to go on. Fraud detection that adapts in near real time instead of waiting for a quarterly rule update. None of that is hype. It's already shipping, and it's making parts of this industry better for the people who use it.

I guess the real version of my position isn't "AI is scary" or "AI is great." It's both, held at the same time.

So if inevitable is the starting point, the question stops being "should we." It becomes "how, and who's accountable for the how."

Ethical use has to mean something more specific than "be careful." For me that starts with disclosure: not hiding where AI touched a piece of work, whether that's a blog post or a risk model. It means not outsourcing judgment calls that actually require judgment, just because a tool can generate a plausible sounding answer fast. And it means paying attention to whose data trained the thing and who's compensated for it, instead of treating that as someone else's problem further up the supply chain.

Keeping a human workforce isn't going to happen by accident. It's not going to happen because companies feel bad about layoffs, either. It happens if retraining and transition support get funded before the disruption hits, not after — the way the statement itself points out that unemployment insurance wasn't built for this kind of shock. It happens if companies treat augmentation as the actual goal, not a talking point on the way to headcount reduction. Acemoglu's push for AI labs to build tools that extend human labor instead of replacing it isn't just an academic preference, it's the difference between a transition and a cliff.

Not tanking the economy probably means slower is safer than it sounds. Everyone in tech is conditioned to treat speed as the whole point. But an economy where the technology adoption curve outruns every institution meant to absorb the shock (unemployment systems, retraining pipelines, even something as basic as how fast people can find a new job in a new field) is an economy that breaks in ways that are expensive to fix after the fact. Deliberately building in friction, even when you could move faster, might be the most underrated form of risk management available right now.

The statement that all of these leaders and laureates signed unfortunately doesn't tell you what to do. Two hundred people signed it and still landed on "act now" instead of an actual plan. This post doesn't tell you what to do either, and I'd be lying if I pretended otherwise.

What it should do is make you ask, seriously: what does responsible AI use look like in your job, this week, not someday? Not in the abstract, not as a mission statement — in the actual decision in front of you.

I don't have a clean answer to any of that. I don't think anyone does yet, credentials or no credentials. But I'd rather keep asking, out loud, in public, than pretend the question's settled.