Hacker News Reader

Comment by guywithahat | Hacker News Reader

I mean this seems objectively good. All of the early issues with OpenAI were the AI refusing to respond to a prompt for political reasons. Generic models should be as politically neutral as possible.

View on HN

Is that the most likely outcome, do you think?

What does "politically neutral" mean? I see a roughly equal amount of pro-vaccine and anti-vaccine content on my Facebook lately, is that what you mean?

It is truly sad that a stance on vaccines is somehow a political position. Fast-forward to 2030s where flat vs sphere is determined by an election.

It generally means unopinionated. You should be able to access all the research on vaccines that exists and draw your own opinions.

At some point, there is a gray area where opinions and facts are hard to differentiate, but we haven’t reached it yet. We’re still at a stage where ChatGPT won’t answer questions if they’re critical of unions (or at least that was the case, I haven’t checked recently).

Oddly enough if you access & summarise all the research on vaccines you're going to appear pro-vaccine.

Right, that's the goal. If you're right, our collection of knowledge will defend you. There's no need to fill it with fake info or hide legitimate research.

I agree you should be able to research these topics. But if I ask chatGPT if the polio vaccine works, how do you think it should respond?

What if I ask it if the world is a globe (also something I see debated hotly on my facebook feed for some reason)?

Reply on news.ycombinator.com