This is a situation where AI needs to be correct nearly 100 percent of the time. Errors mean wasted time from the employees correcting orders and wasted food from incorrect orders. An employee that makes hourly mistakes is a problem. No AI proponents are saying that Gen AI is going to be 100 percent correct because the goal is Gen AI is to provide a probable answer.
I’m not sure if you have read this, but here is an example of a more simple interface where Claude was asked to run a vending machine. At times it acted out an existential crisis or even attempted to call the FBI.
This is a situation where AI needs to be correct nearly 100 percent of the time. Errors mean wasted time from the employees correcting orders and wasted food from incorrect orders. An employee that makes hourly mistakes is a problem. No AI proponents are saying that Gen AI is going to be 100 percent correct because the goal is Gen AI is to provide a probable answer.
I’m not sure if you have read this, but here is an example of a more simple interface where Claude was asked to run a vending machine. At times it acted out an existential crisis or even attempted to call the FBI.
https://www.anthropic.com/research/project-vend-1