how does that work if it supposedly has to know the answer to check?

Reply to this note

Please Login to reply.

Discussion

It looks at answers from thousands of people to understand an object

Probably assigns confidence scores and then rotates with least and most confident answers to the same question to try to identify an object

Could be, in my experience it's always a traffic light or a fucking bus, and the images many times are similar, so not sure it's doing anything too advanced

That’s because they are using it to train autonomous vehicles 😆