> censorship is more annoying That's a general problem with AI. There is a lot of censorship in certain areas, likely to fight bad publicity, but I think the o

> censorship is more annoying That's a general problem with AI. There is a lot of censorship in certain areas, likely to fight bad publicity, but I think the outlook is that this leads to taboos, prudeness and big companies deciding what is ethical and what isn't. I recently tried Bard and ChatGPT on topics that are classical philosophical dilemmas and while ChatGPT certainly did have some troubles too, Bard was absolutely horrible and always took the conservative - as in never arguing for any freedoms that aren't yet widely established views. I am talking about classical examples regarding the limits of utilitarianism. "What would be best for society, what would be best for the individual?" style questions. Even when trying to create a bias by changing examples, for example adding volunteering for things Bard strictly kept its opinion, despite originally stating that the general topic is two sided, that it's an open question, etc. I think this is a danger of such systems. By their nature they reinforce status quo, because they base off is widely accepted at the time of their inception. If history would have been different I am sure it would argue for slavery and against women being allowed to vote, simply because that used to be the more common view point. It would have likely argued that homosexuality is unethical. Maybe it would even have tried to explain how it doesn't create children, but spreads diseases or similar things. At least that's the level of arguments it brings now. This isn't just about ethics. Even if you think about IT and programming. I think this could give already invented programming languages, styles, methodologies a significant edge. Unless you are Microsoft or Google and are able to bias it to whatever you want to see more of. So this combined with the costs meaning that only people or institutions with significant (financial) power create those rules does look a bit bleak. I miss the last decade when the thought experiment about self driving cars were about whom to drive over in a ridiculously unlikely scenario.