4 ms·
I'm sure it was independently evaluated, but I'm sure the folks running the test were not given an on-prem installation of ChatGPT to mess with. It was still do
by upghost 2y ago
I'm sure it was independently evaluated, but I'm sure the folks running the test were not given an on-prem installation of ChatGPT to mess with. It was still done via API calls, presumably through the chat interface UI.
That means the questions went over the fence to OpenAI.
I'm quite certain they are aware of that, and it would be pretty foolish not to take advantage of at least knowing what the questions are.
- jvanderbot 2y agoNow that you put it that way, it is laughably easy.
- ls612 2y agoDepending on the plan the researchers used they may have contractual protections against OpenAI training on their inputs.
- upghost 2y agoSure, but given the resourcing at OpenAI, it would not be hard to clean[1] the inputs. I'm just trying to be realistic here, there are plenty of ways around contractual obligations and a significant incentive to do so. [1]: https://en.wikipedia.org/wiki/Clean-room_design https://en.wikipedia.org/wiki/Clean-room_design
- xigoi 2y agoDo you think ClosedAI cares about that? It’s not like anyone can tell if they violate the contract.