I think its because thats how you get sentinel values to work. At this point its no worse a deal tbat the next encryption surface/chokepoint/target/leaky boat?
There us much more pragmatic reasons to cancel because of dots. The handoff procedure absolutely blows. Just spits in your face re: understanding intent between tasks
Probably worth noting, yes, I'm aware if you're running atop their cloud infrastructure, OpenAI already technically have access to whatever they like. How would you feel about VPS providers enforcing this same policy?
Yes, I'm also aware that if your data goes to their models for processing, you're of course already sending them your private data too. However, you may have noticed, Codex and the like actually go to pretty great lengths NOT to upload things like secrets to the model. This is the polar opposite approach.
Isn't the idea of having a cloud computer meant to be that it's more than just a model? It provides compute, so it can run scripts locally to filter data etc.
I think its because thats how you get sentinel values to work. At this point its no worse a deal tbat the next encryption surface/chokepoint/target/leaky boat?
There us much more pragmatic reasons to cancel because of dots. The handoff procedure absolutely blows. Just spits in your face re: understanding intent between tasks
Probably worth noting, yes, I'm aware if you're running atop their cloud infrastructure, OpenAI already technically have access to whatever they like. How would you feel about VPS providers enforcing this same policy?
Yes, I'm also aware that if your data goes to their models for processing, you're of course already sending them your private data too. However, you may have noticed, Codex and the like actually go to pretty great lengths NOT to upload things like secrets to the model. This is the polar opposite approach.
Isn't the idea of having a cloud computer meant to be that it's more than just a model? It provides compute, so it can run scripts locally to filter data etc.