I've found that AI opinions can be classified into three different groups:
https://swecyb.com/@troed/116725135832925252
Erik Jonker (npub1jnv…zpfx) Some Guy (npub1duw…xq6f) Max Leibman (npub1388…sxrr)
I left Facebook partly because I really didn't want to spend my time on pointless arguing with people that Really Should Know Better(tm).
The last week or so I've found myself now doing the same thing here on Mastodon, and it bugs me. One of the irritating things my autism decided to bless me with is an aversion to seeing blatant falsehoods being spread.
The topic de jour is of course LLMs. I've found the following to be a good approximation of the state of affairs here:
1) The people who most loudly scream about LLMs being useless are:
- not developers
- don't know anything about how LLMs work
2) The people who ridicule LLMs might:
- be developers
- have decided to not use them and don't really know what they currently can and cannot do
3) The people who post in support of using LLMs are:
- developers
- decided at some point to see what the hubbub was about and found them actually useful
The problem is that people in groups 1 and 2 believe they sit on Truth and that everyone in group 3 is either a paid shill, delusional or incompetent.
(as you can see I'm only talking about LLMs from the aspect of software developer/reverse engineering etc. That because I don't have the competence to have an opinion on them in other areas and so I refrain from even trying)
I could of course stop commenting when I see errors our just outdated information being posted. What I don't like with that approach is that those in group 2 who I genuinely believe would benefit from additional insight keep believing all the crap produced by group 1.
See, group 1 are a bit like anti-vaxxers. They've "done their research" and no amount of facts will ever sway them. They've gone into full "in-group vs out-group" mode and they'll resort to actual threats towards anyone who they perceive to challenge _their whole identity_ as an anti-AI fanatic.
sigh