If I could personally have my way, I think it would be to re-envision the rules and regs. It seems like we threw away the rules and regs when large models were being trained in the first place.
A lot of these large models were essentially predicated on a lot of illegal activity to start with. Whether it be training on child sexual abuse material, non-consensual intimate images or simply copyright material, it proliferates in all of these large models. It was illegal to start. Now we're in this bad position. I would argue that we need to take a step back and say, “The rules we had in place weren't adhered to. Are we now just accepting that this is the new reality?” I take issue with that.
I'll go back to what I said earlier relating to accountability. You have to be accountable. The way these AI models work, it's not like you can seek to have everything fit under section 230 in the U.S. In terms of the outputs of these, we accept that no one's responsible for them, because it's not an individual. It needs to be attributed to the companies using them. There was an Air Canada example where one of their agents was discussing a flight with a customer: Here's the price. Air Canada went back on it. Well, no, that's not acceptable. If you're going to use AI in this way, you need to be responsible for the actions that come out of them.
I think both are actually important. One is probably a little more likely than the other in terms of the genie being out of the bottle, but there you have it.
