AI That Shifts the Blame

Kyle Kingsbury believes that going forward, some people will be hired essentially as the ones responsible for ML systems, even if it's not explicit. He describes these people as "meat shields" who take the blame when the system goes wrong.

That responsibility can take various forms. It could be an internal responsibility, like at Meta, where a person re-checks the results of automated content review. Or it could be an external responsibility, like a lawyer who gets sanctioned after submitting an LLM's lies to a court.

In other cases, it becomes fixed as an institutionalized role, like a Data Protection Officer. He also points out that companies can use outside contractors like Buscaglia as a way to offload blame for the entire system's malfunctions onto someone else.