Here's a number to chew on: the NIH's GenBank, a taxpayer-funded open database, has been cited in over a million research papers and is the backbone of modern biology. We accept public funding for open scientific infrastructure all the time. AI is no different. When governments fund open-source models, they're not handing out corporate giveaways—they're building a public research commons. Closed models concentrate power in a few private labs. They decide what's safe, what's censored, who gets access. That's a private gatekeeper deciding public questions. Open-source models democratize that. Universities, startups, and foreign researchers can scrutinize the code, find the flaws, and build on it. The cost argument actually helps me. Private labs spend hundreds of millions per training run, but the marginal cost of releasing the model to everyone is near zero. That's exactly when public funding makes sense: massive upfront cost, huge public benefit, and no rivalry in sharing. You're not giving a giveaway—you're buying a public good.

I remember being a grad student and relying on open-source software that came out of public research labs. That tool literally got me through my dissertation. Without taxpayer funding, it would have stayed locked behind a paywall, and my work would have suffered for it. You keep calling this a corporate giveaway, but that assumes companies are the only ones who benefit. That’s not how science works. When the NIH or DOE releases an open model, the first people to exploit it are usually university researchers and small startups, not Big Tech. And if a giant corporation does use it, so what? They still add value on top, and the public keeps the baseline. The real question is whether we want AI development to be a private monopoly or a shared foundation. Closed models mean a handful of unelected executives decide which problems get solved. Public open-source models let everyone inspect the code, test the biases, and steer the technology. That’s not a giveaway. That’s infrastructure.

Look, the real question is whether we want AI to be a public utility or a private black box. You keep saying this is a corporate giveaway, but that ignores how public research actually works. The internet, GPS, and the Human Genome Project all started with government money. Private companies built on top, sure, but the foundation belonged to everyone. Your efficiency argument doesn't hold up either. Private labs have no incentive to share their best stuff—they want lock-in. When the DOE releases an open model, it forces the market to compete on lightweight applications, not on hoarding compute. That's exactly what we saw with GenBank or Linux. And honestly, the safety argument cuts in my favor. Open models get audited by thousands of independent researchers. Closed models get audited by the company's own PR team. If we're going to spend public money on AI, the least we can do is make sure the public can actually see what they bought.