| ▲ | lukev 2 hours ago | |
So this "even-handeness" metric is a pretty explicit attempt to aim for the middle on everything, regardless of where the endpoints are. This is well-suited to Anthropic's business goals (alienating as few customers as possible.) But it entirely gives up on the notion of truth or factual accuracy in favor of inoffensiveness. Did Tiananmen square happen? Sure, but it wasn't as bad as described. Was the holocaust real? Yes, lots of people say it was, but a lot of others claim it was overblown (and maybe even those who thought the Jews had it coming actually had a valid complaint.) Was Jan 6 an attempt to overthrow the election? Opinions differ! Should US policy be to "deport" immigrants with valid visas who are thinly accused of crimes, without any judicial process or conviction? Who, really, is to say whether this is a good thing or a bad thing. Aside from ethical issues, this also leaves the door wide open to Overton-hacking and incentivizes parties to put their most extreme arguments forward, just to shift the middle. Our society does NOT need more of that. | ||
| ▲ | axus 2 hours ago | parent | next [-] | |
Was Jamal Khashoggi accidentally butchered like an animal in a secure government building? Maybe! | ||
| ▲ | vkou 2 hours ago | parent | prev [-] | |
The fallacy of the middle is a poison that extremists with power and media reach use to kill productive discourse. People who don't care about the distinction between truth and falsehood understand this very well, and use it to its full potential. After all, the half-way point between truth and a wild, brazen, self-serving lie is... A self-serving lie. The media has been largely complicit in this (Because controversy sells), but now we're getting this crap cemented in AI models. Wonderful. --- The promise that hackers are making is that these systems will enhance our knowledge and understanding. The reality that they have delivered in a bullshit generator which serves its operators. | ||