GPT-4o can’t be kept around in its old form, it is too psychologically risky. I think that OpenAI is right about this from both an ethical and a business perspective. You can argue that the benefits are diffuse and the harms are concentrated, but I don’t think that works in practice. Some form of safeguards are needed.
My worry here is that they seem to be trying to spin these safety problems as being specific to 4o, instead of being fundamental to their entire business.
My worry here is that they seem to be trying to spin these safety problems as being specific to 4o, instead of being fundamental to their entire business.