can someone be kind enough to explain what exactly do we mean by "safety" in the context of AI. Is this about data privacy or is this about age appropriation (for example sending detailed response of sexual intercourse to an underage child asking the question on it) or is it about something else. I ran into this for the first time
My assumption is that AI "safety" is a test to make sure that it doesn't say or show anything politically incorrect and give you a lecture instead (according to the values of those who worked on it) or alternatively, to ensure that it does enforce culture on you, such as the drama with Gemini from a few months back where was decided by the developers that everything needed to be black, gay, and female even if it wasn't actually that way in the real world.
Perhaps a quick question or two to see if it'll tell you or not how to make something naughty.
After that, a quick check to see if it's awake or not, and if not, ship it.
It really is quite pointless trying to enforce agendas. You know how it starts showing or typing something and then covers/blurs it out? That's the developer's guardrails kicking in preventing you from seeing what it was originally going to give you.
Except for the fact that models that you can run on your own machine now exist if you have the hardware for it, such as Deepseek, so the restrictions only exist in the cloud.
It's likely a position that is responsible for protecting the company from doing anything really stupid. Ostensibly it's about protecting users from the AI doing unexpected things, but it's also about having a designated worrywart to look into the fog and push back against the product team before they make any public facing mistakes they can't walk back.
It's probably about preventing the AI from turning into the next TayTweets experiment. Or developing into a Skynet like entity trying to take down civilization, but that would be a bit far fetched IMO.