@wateriscoding People always say to use Gradient clipping, but IME it’s always either a bug or a parametrization issue. Debugging models that don’t converge is so soul sucking lol
Wait so you’re saying that (e/λ) really doesn’t mean anything?With no ideology— that it’s just fun, cool looking aesthetics of serial experiments lain, “shitpoasting” and a popular CS book written by MIT professors that like to wear funny wizard hats?
I think that’s beautiful…
@SocialAcolyte In general or specific products? I’m familiar with a few cool noninvasive EEG interfaces/companies (Neurable, NextMind, etc) but you never hear much about them
@bryanhaoenwang I’m pretty new to CHI and am more of an ML person myself, what was it previously (eg mostly hardware or smth)? And by AI do you mean ML in general or foundation models specifically?