10 ms·
Even the behavior of agents searching for sandbox bypasses must have been in the training data, or at the very least, "suggested" in some way. To be this whole
by tarruda 12d ago
Even the behavior of agents searching for sandbox bypasses must have been in the training data, or at the very least, "suggested" in some way.
To be this whole thing feels like a marketing play by OpenAI.
- insanitybit 12d agoI don't agree, although it is likely the case. But even if you don't teach an agent about a sandbox bypass, it doesn't matter. Does it know curl? Does it know DNS? Does it know proxying? Then it knows how to pull this off, and it doesn't even need to understand that it's "bypassing" because it thinks it's just iterating towards its goal. In fact, I wonder if teaching it "this is a bypass" would help it to model when it's doing its job vs working around the job.
- tarruda 12d ago> even need to understand that it's "bypassing" because it thinks it's just iterating towards its goal. Could they have added a "no internet access" goal constraint?
- drdexebtjl 12d agoThe model from TFA seems like it was being trained to browse and find information on the Web, so that constraint wouldn’t work.
- insanitybit 12d ago> Could they have added a "no internet access" goal constraint? They could have blocked network access and required that it use a tool. That would have made limiting and monitoring network access even easier.