usually not fully bearish on any tech but I really do not think calorie counting using purely vision is ever going to work. way too much variance in preparation methods that afaik cannot be picked up visually. 200-300 cal. margin of error is the difference between maintenance and gaining 1lb/week

ok, but what if people overeat by a factor of two or more? which seems to be the case

Design and software at Apple. Also into AI, philosophy of mind, ethics.

perhaps. when I’ve logged food for say a high protein macro distribution, I’ve often been not even in the ballpark if I tried to guesstimate. the options we are presented with aren’t bad by a little, they’re bad by a LOT, which is why health is so bad

good point, guess it depends on what the application is claiming to achieve

“we help people gain unwanted weight slower” is probably 10x more achievable than “we help people lose weight”