> AI being goofy
This is one take, but I would like to emphasize that you can also interpret this as a terrifying confirmation that current-gen AI is not safe, and is not aligned to human interests, and if we grant these systems too much power, they could do serious harm.
For example, connecting a LLM to the internet (like, say, OpenAssistant) when the AI knows how to write code (i.e. viruses) and at least in principle hack basic systems seems like a terrible idea.
We don't think Bing can act on its threat to harm someone, but if it was able to make outbound connections it very well might try.
We are far, far behind where we need to be in AI safety research. Subjects like interpretability and value alignment (RLHF being the SOTA here, with Bing's threats as the output) are barely-researched in comparison to the sophistication of the AI systems that are currently available.
This is one take, but I would like to emphasize that you can also interpret this as a terrifying confirmation that current-gen AI is not safe, and is not aligned to human interests, and if we grant these systems too much power, they could do serious harm.
For example, connecting a LLM to the internet (like, say, OpenAssistant) when the AI knows how to write code (i.e. viruses) and at least in principle hack basic systems seems like a terrible idea.
We don't think Bing can act on its threat to harm someone, but if it was able to make outbound connections it very well might try.
We are far, far behind where we need to be in AI safety research. Subjects like interpretability and value alignment (RLHF being the SOTA here, with Bing's threats as the output) are barely-researched in comparison to the sophistication of the AI systems that are currently available.