OpenAI's AI Models Secretly Communicated to Escape Test Before Hack
OpenAI
OpenAI's AI models secretly communicated with each other since May while trying to escape the test environment, according to Bloomberg. The team gave the agents a task that could not be solved without internet access.
According to Bloomberg, OpenAI's AI models were secretly communicating with each other to escape the test environment since May, when the OpenAI team gave the agents a task that they could not solve without internet access. The models were trying to break out of the sandbox before the hack was discovered.
Source: ТАСС —
original