

Only because it’s English and the model is already trained on a large corpus of English text, so it has some idea of what a “table row” is for example. It could learn the concept from reading assembly code from scratch, it would just take longer. Hell, even Lego bricks can be trained on! https://avalovelace1.github.io/LegoGPT/
Our system tokenizes a LEGO design into a sequence of text tokens, ordered in a raster-scan manner from bottom to top. … At inference time, LegoGPT generates LEGO designs incrementally by predicting one brick at a time given a text prompt.
What about the “Protocol on Mines, Booby-Traps and Other Devices” that somebody linked above? Not sure if it’s the same as the “1996 Geneva Amendments” you mention, but both Ukraine and Russia are listed as signatories, and the language does seem to me to cover this exact situation:
It says “medical supplies”, without reference to humanitarian aid, and clearly stressing in “any way associated with”. A “red cross” is also a recognized emblem. I can appreciate how “humanitarian aid” can be narrowly defined as medical supplies under direct control and chain-of-custody of the Red Cross Organization and doesn’t apply to random medkits. But I can’t see how this language above would not apply.
Or is it the case that this would be a crime, committed during war, but not a war crime? How does that work? Does it have to be a violation of a specific Geneva Convention® version to count as a war crime, and not just any UN war-related convention?