Are you saying that large-model capabilities would make a substantial difference in a military conflict within the next five years? Because we aren’t seeing any signs of that in, say, the Ukraine war.
We do see the signs and reports, you just have to look. LLMs are being adopted to warfare, with drones or otherwise, there is progress there, but it's not currently at the level of "substantial difference". And 5 years is huge time from progress perspective in this domain - just try to compare LLMs of today with LLMs of 2020.
small scale drones are in use in that conflict. On device AI would be a game-changer no?
It’s not impossible, but also highly nontrivial. Apart from the actual AI implementation, power supply might be a challenge. And there is a multitude of anti-drone technology being continuously developed. Already today, an autonomous drone would have to deal with RF jamming and GPS jamming, which means it’s easily defeated unless it has the ability to navigate purely visually. Drones also tend to be limited to good weather conditions and daytime.
In terms of countermeasures, what's the difference between having a human drone pilot and having an AI (computer vision plus control) do it over cloud? I know I'm moving the goalposts away from edge compute, but if we are discussing the relevance of GPU compute for warfare it seems relevant.
Assuming human-level AI capabilities, not much of a difference, obviously. But I also don’t think that human operators are a bottleneck currently. Cost, failure rate, and technical limitations of drones is. If you are alluding to superhuman AI capabilities, that’s highly speculative as well with regard to what is needed for drone piloting, and also unclear how large the benefits of that would be in terms of actual operational success rate.