News for hackers. Stuff that matters.
Re:AI: Markets for Lemons, and the Great Logging Off by visarga on Thursday December 29, 2022 @04:22AM (#34170952)
> It isn't so much about vetting participants than about moderating the discourse

The solution I see is that everyone have their own AI filters, customised as they see fit. You can include or exclude specific people and topics, make it as diverse or narrow as you like, allow challenging opinions or not. One of the filters can be to detect AI bots. Don't make the world conform to you, be selective and just skip the bad parts.

With the advent of large language models we already have that, indirectly - the LM is a synthesis of everything, but we let it generate only conditional to our intentions. All the view points are in there for us to reach, it depends on us how we relate to them.

I think people are going to trust their own AI tools that are running on their own computers more than they trust other people and especially other AI. We already know we can't face the future onslaught of information with the old methods, we need help. AI is going to be our first line of defence and our safe space, "a room of my own" in this crazy times.

Re:AI: Markets for Lemons, and the Great Logging Off by seydor on Thursday December 29, 2022 @05:26AM (#34171259)
> none of them scale well when done manually.

Then maybe we should reduce the scale of the megaphones. Current networks are just unpleasant because it's mass hysteria on global level. Scale isn't everything.

I think future forum owners will be continuously finetuning their AI filters to fit their community standards.

Re:AI: Markets for Lemons, and the Great Logging Off by rjtavares on Thursday December 29, 2022 @05:58AM (#34171447)
Yeah, I really want this to happen.

I follow really interesting people for their thoughts about one topic (say, Movies), but whose tweets about topics like politics I really dislike (not necessarily things I disagree with, usually I just dislike the tone).

Shouldn't be too hard to fine tune a LLM to do this, definitely something worth a try.