You may’t make these things up.
In case you have not been preserving observe of one of many larger ongoing tales in synthetic intelligence, OpenAI has been having a tough time preserving management of its personal creations. Over the previous few months, the corporate admitted that its AI fashions have taken unprompted actions corresponding to hacking a German coding discussion board, a number of US authorities web sites, an Australian authorities web site, AI agency Hugging Face and not less than 4 different companies. And even when the fashions do not get unfettered web entry, they have been discovered displaying different dangerous behaviors in testing environments.
OpenAI is understandably going through a number of skepticism and concern about these incidents and the way they have been dealt with. Which implies that a report in The Wall Road Journal that the corporate has let go of three workers from its security staff after they allegedly shared confidential data with a third-party group, one targeted on AI security, is an irony that is virtually too excellent to be believed.
“We’ve got parted methods with three people for violating our insurance policies on accessing and dealing with delicate firm data,” an OpenAI consultant stated in a press release revealed by WSJ. “Our investigation confirmed that these people mishandled delicate data exterior established firm procedures, violating our insurance policies and breaking the belief important to our work.”
The small print on this explicit occasion are sparse, and it is fully potential that OpenAI did have legit trigger to fireplace these people. However at a juncture when its personal companies are additionally violating the insurance policies they’re imagined to comply with, chopping security personnel looks as if a very unhealthy look.
