Post
BTW most effort is going into agents, as that is where the profits will be (tools are much more generic, and not enough money there) the problems may get solved. But if we see nothing better by 2027 I am right.
I want agents. Particularly I want an agent that can drive my car. Mēh!
(1/3)
@worik
> Uses of AI as tool are looking good.
> Uses of AI as an agent, not so much
That's more or less where I'm coming down on it. @pluralistic has a few good pieces exploring the distinction between "centaurs" and "reverse-centaurs", including;
https://pluralistic.net/2025/09/11/vulgar-thatcherism/#there-is-an-alternative
... and ...
https://pluralistic.net/2022/04/17/revenge-of-the-chickenized-reverse-centaurs/
(2/3)
OCR is "AI". As is machine translation of natural language. Both things I endorse as useful, and which can, in theory, be developed as fully libre tech. After all GNU FM incorporated an "AI" recommendation algorithm over a decade ago. No proprietary software or data required.
(3/3)
What I object to in the current "AI" hype bubble is not the concept of automation, or even "smart" automation, but the MOLE Training long con and everything that comes with it;
https://disintermedia.substack.com/p/invasion-of-the-mole-trainers
* corporate control of the tech
* proprietary nature of existing MOLEs
* the DataFarming its all depends on
* the corruption of OSI
* defunding of community-driven software that isn't "AI"
* the extractive nature of the hyoerscale datacentres it runs on;
https://techwontsave.us/episode/241_data_vampires_going_hyperscale_episode_1
etc
A space for Bonfire maintainers and contributors to communicate
