[shagie]

I believe the key word in that passage is "decisions".

You can't use ChatGPT (or other OpenAI offerings) to grade essays, decide who is the least risky tenant, assign risk for insurance, filter resumes, approve a loan, determine sentencing...

Those are things that require human agency for "a person decided this" rather than "we fed it into the program and took the answer."

[elcritch]

Which of course means humans will use it for those decisions until there’s some lawsuits and maybe some laws.

[snoman]

Humans have been using bad data to make decisions for a while now.