If you make the company itself have culpability for what it’s “rogue” agents do then you solve a lot of the problems. Tech executives might not care about humanity but they probably care about themselves. If the CEO of OpenAI is as legally responsible for what OpenAI’s software does in the same manner that Sam Bankman-Fried was responsible…I suspect many AI companies would not be so cavalier about letting their models do dangerous things.
This article makes a good point.
If you make the company itself have culpability for what it’s “rogue” agents do then you solve a lot of the problems. Tech executives might not care about humanity but they probably care about themselves. If the CEO of OpenAI is as legally responsible for what OpenAI’s software does in the same manner that Sam Bankman-Fried was responsible…I suspect many AI companies would not be so cavalier about letting their models do dangerous things.
Paywall skip: https://archive.ph/PPrPh