artificial intelligence assistants are currently dumb.
they dont remember persist anything they cached.
The point is, I should be able to tell this thing ... bad, this is what I want.
But currently the only people able to do this are the programmers.
Thats not smart thats generic.
thats not custom thats stupid.
the car autonomously came around the corner and slowly came to a stop in the middle of the road ... lydar 3d data from car elderly woman trying to sweep a duck off the road. car patiently waited then slowly moved off the road. you are not going to make a unit test that says wait for a duck to get out of the road. we have to see things like trees on the ground or tricycles. gotta get out to know how much work you still need to do. come back with a list... were working on this wrong were working on this wrong were working on this wrong. top ten features that are needed for your project. different list... learning list. the things that we could do ordered top to bottom from most learning to least learning. go make me the important list again. Its unrecognizably different from the important list. if you couch failure as learning... then he lost me. if the success or failure doesnt matter that is a bad experiment. if it is an answer you already know that is a bad experiment but a reusable thing. How long would it take to rebuild it now that you know what you know... he says probably ten percent. (yeah, I dont know about that number. This guy says a lot of numbers) (but I guess it depends on who your talking to.) ~ Astro Teller
Hey, we often do the list and learning parts in open source!