Alex Mordvintsev says a manually tuned swarm simulation shows complex self-organization is possible without machine learning
Decentralized agents coordinate using local rules rather than optimization
@mayfer this is amazing wtf
murat 🍥@mayferi recall @nickcammarata asking how the ants did it, here's a swarm interpretability nerd snipe
#778Alex Mordvintsev@ZZZNAHWhoa!!!♥️
memo akten@memoaktenMy babies are getting so smart, it's freaking me out. I was looking forward to using GA or bayesian optimisation to find a sweet set of parameters, but hand-tuning is giving scarily good results. I cant stop watching! (No global coordinator, all acting entirely on local cues)
#778Alex Mordvintsev@ZZZNAHSeeing this resonates heavily with my recent shift in perception of self-ogranization as an engineering discipline. For years I was thinking that it's too complex to for humans to do w/o ML/compute-heavy search because "computational irreducibility" and unintuitive relation between local rules and global behaviours. But now it feels like the brave ones can do surprisingly far with modest tools and some change in perspective
memo akten@memoaktenMy babies are getting so smart, it's freaking me out. I was looking forward to using GA or bayesian optimisation to find a sweet set of parameters, but hand-tuning is giving scarily good results. I cant stop watching! (No global coordinator, all acting entirely on local cues)
#1882murat 🍥@MAYFERi recall @nickcammarata asking how the ants did it, here's a swarm interpretability nerd snipe
memo akten@memoaktenMy babies are getting so smart, it's freaking me out. I was looking forward to using GA or bayesian optimisation to find a sweet set of parameters, but hand-tuning is giving scarily good results. I cant stop watching! (No global coordinator, all acting entirely on local cues)