Ask an AI a question then doubt the questions facts. AI’s always believe their answer is correct. Imagine when the AI thinks “is this street clear?” Then answers yes to itself and slams into a parade full of people at top speed and just keeps driving
Fair, when IBM’s Watson was on Jeopardy it did show percentages of what it’s confidence level was. So I guess in my parade scenario Elon has told engineers that 51% is more than halfway correct so they have to program the AI to accept it as 100% correct.
My main point is that Elon is a snake oil salesman and a no good grafter.
7 Comments
ME5SENGER_24@lemm.ee · 4 pts · 3y
Ask an AI a question then doubt the questions facts. AI’s always believe their answer is correct. Imagine when the AI thinks “is this street clear?” Then answers yes to itself and slams into a parade full of people at top speed and just keeps driving
fallingcats@discuss.tchncs.de · 1 pts · 3y
Language models are a completely different kind of AI though.
ME5SENGER_24@lemm.ee · 3 pts · 3y
Fair, when IBM’s Watson was on Jeopardy it did show percentages of what it’s confidence level was. So I guess in my parade scenario Elon has told engineers that 51% is more than halfway correct so they have to program the AI to accept it as 100% correct.
My main point is that Elon is a snake oil salesman and a no good grafter.
YIj54yALOJxEsY20eU@lemm.ee · 1 pts · 3y
little too much anthropomorphizing for AI
Ubermeisters@lemmy.zip · 3 pts · 3y
Maybe we should set up some laws around this?
MyNameIsIgglePiggle@sh.itjust.works · 2 pts · 3y
Yes. Won't be better
DogMuffins@discuss.tchncs.de · 0 pts · 3y
What a mess. A car that stops at red lights 99.9% of the time is not Fully Self Driving. This is level 3 of 5.