2 ms·
I'd love to see AI driven robot discovering optimal swordfighting "AI zero" style.
by scotty79 5y ago
I'd love to see AI driven robot discovering optimal swordfighting "AI zero" style.
- stefs 5y agowell ... it would be interesting, but i fear that an AI robot vs. human swordfighting style would probably be really boring (given that it'd be possible to make a robot with human attributes, i.e. strength, sensory input, etc.). my guess is a very small set of moves would emerge that are, for practical purposes, un-counterable, depending on the proportions of the particular human opponent. that said, how to train the AI? robot to human test battles are very limited in training set size, so it would have to be AI robot vs. AI robot. that would probably yield a strategy very close to the style of the AlphaGo/zero/*: very conservative, maximizing the space of winning scenarios, winning each one with as little of an advantage as necessary. edit: i've got not the slightest clue about sword fighting (except that it doesn't look like in the movies), so i could be completely off the mark here.
- andrewflnr 5y agoGiven the convergent evolution of fighting styles around the world, I think the result would be pretty familiar. By far the tricky part would be evaluating a "style": in real life, footwork and using biomechanics to eke out a tiny edge in speed play a role. And of course you have to start over moving to a different type of sword, or a different opposing weapon...