I'm a little skeptical of the AI safety field. Nonsense being available in written form is nothing new. If you don't want people to be influenced by propaganda, the only solution is to educate the masses and give them the ability to critique what they're hearing. Telling an AI model "don't give people ideas on how to kill themselves" accomplishes pretty much nothing, when half of 4chan is certain to provide ideas when prompted.
There are so many low-hanging fruit that can be picked to make people safer and more secure; fully-fund public education, eliminate pre-Internet administrative shortcuts (your address gets published on the Internet if you buy property or get a radio license or even register to vote), give people basic income, etc. Putting a filter on a system that we barely understand probably isn't going to have much effect, as this paper finds. We'll be fine without the filter, though.
On top of R&D, it currently takes $$ millions to stockpile/prep data and train a model, so it's only being done by high-profile corporations who can budget for that. Those corporations must operate in hyper-conservative CYA mode because what they can't afford is for their brand to be tainted by bad PR or reactive legislation.
It's now really about the safety of the community at all. It's about the safety of the brands.
My takeaway from the original article was "how dare Meta release model weights when the safety could be removed so easily", which is evidence that Meta doesn't really care. They are being asked to do the impossible, but without any teeth (there are no "AI safety" laws), so they didn't do the impossible. I think that's fine and society is better off for it. People are doing neat things with Llama-2, quickly.
As for safety of the brand, I'm guessing that people don't know that Meta is Facebook and Instagram, so even if an AI safety incident somehow blew up (which my limited imagination cannot even comprehend, perhaps I should ask an AI), people would probably keep using Instagram.
There are so many low-hanging fruit that can be picked to make people safer and more secure; fully-fund public education, eliminate pre-Internet administrative shortcuts (your address gets published on the Internet if you buy property or get a radio license or even register to vote), give people basic income, etc. Putting a filter on a system that we barely understand probably isn't going to have much effect, as this paper finds. We'll be fine without the filter, though.