GitHub - bytedance/UI-TARS: I think I should go deeper into this, and try to find out if they do anything interesting, or its just a reiteration of what these Large models already do, because GPT 4o, 4.1 are really good at web interaction already.
This is like eye opener on how to model real world digitally for me. Didn’t know something like this is already possible. So, this is what google uses for street view. and it’s arguably realtime too.
olfactory-computer interfaces: imo it’s the next big domain to unlock for entertainment.
Why has it not taken off already? because its the slowest of all 4 sensory system in human body. Vision, Hearing, Touch are all instantaneous, because all three are used for fight-flight response. Olfactory senses are not used for survival. And hence, the hardest to digitise.