2026-04-28 21:37 UTC
Replies (16)
-
@PrimozR@social.vivaldi.net 2026-04-28 22:53
@anildash Here a handful of people made something useful, something that actually solves a problem. Meanwhile, bajillion-dollar valuated companies create "agents" that ignore instructions and delete production databases. Quite clear who's in it for just the money and who cares about the product.
-
@mossman@social.vivaldi.net 2026-04-28 23:36
@anildash I saw their video about this a few weeks ago and it does truly look like a useful tool for video production which they want to allow people to use as freely as possible. Seemed quite impressive and the collaborative attitude was refreshing as well.
-
@petersibley@mastodon.social 2026-04-28 23:46
@anildash nice writeup !
-
@hankg@friendica.myportal.social 2026-04-28 23:48
@anildash This is so inspiring and your bullet points are so nailing it. Thank you!
-
@neutrinodisaster@social.vivaldi.net 2026-04-28 23:53
@anildash I feel like this is useful as it provides the scaffolding for how we can clearly articulate the problem with current "AI" as contrasted with more helpful uses of machine learning. The "AI" we dislike is an inhuman *other* from us, with closed data, a black box algorithm, and a goal of replacing our human joys of creating, research, and communicating with others -- it is an imposter pretending to be human, but really just an idol of capital. This, meanwhile, is a clever use of math and data to fill a specific hole *in* those pursuits, really a "tool" (even though that's overused by proAI") used for creative purposes. That's what separates the machine learning for photo sorting, subject selecting in photo editing, and all those useful features, from the ghastly, over-pushed, soulless machine learning that pushes out slop and dehumanizes our very creations, pursuits, and lives. I think if we realize those ethical grounds and differences, we can probably convince so much more people to really care about why we should stop using AI -- because there will always be those who say that hallucinations will be lessened, or we can de-corporatize it, and they may be right -- but what won't naturally change are the ethical, unhuman problems in the process and function of "AI" that we all know and understand.
-
@cutesobri@cathode.church 2026-04-29 00:59
@anildash fuck off ai!
-
@nyhan@fediscience.org 2026-04-29 01:43
@anildash I also like that it's designed to solve a particular problem. That makes it easier to evaluate its performance. And I have the impression from the "Power hungry processing" paper that purpose-built models used less energy for the same task than general purpose models.
-
@OliviaVespera@spacey.space 2026-04-29 01:53
@anildash exactly the kind of thing we build. neural nets, have so much potential before even getting to large language models.
-
@kitkat_blue@mastodon.social 2026-04-29 02:27
@anildash "He was able to build the tools that carried out this training by asking one of the current popular commercial AI tools to help." If you gen-ai corp datacenters to do anything within a project, it taints the entire project with gen-ai and all the bad that includes. Focusing on "localizing" gen-ai involved tools is just the gen-ai version of greenwashing. 'localwashing'? 'openwashing'? whatever. "useful" isn't an excuse. "Good" ai is ai that isn't tainted with gen-ai.
-
@negative12dollarbill@techhub.social 2026-04-29 04:07
@anildash Just so you know, I can't read this post on my phone. This is what it looks like, and I can't close the citation box by clicking the x top right.
-
@n_dimension@infosec.exchange 2026-04-29 06:12
@anildash "He was able to build the tools that carried out this training by asking one of the current popular commercial AI tools to help." Good indeed.
-
@Lyle@cville.online 2026-04-29 11:42
@anildash Thank you for sharing this. I have been struggling to think through a helpful direction for AI thinking especially about Illich’s work on “convivial” tools. I agree this is the closest I’ve seen. I also liked that piece at the Met a while back https://www.moma.org/calendar/exhibitions/5535
-
@Papa_Joe@www.minds.com 2026-04-29 12:04
A great look at big AI and what a few people with their head and heart in the right place can accomplish as open source, single purpose, ethically trained AI.
-
@franciswashere@mastodon.social 2026-04-29 19:56
#Reinventing_Green_Screen
-
@mpanhans@social.librem.one 2026-04-30 00:54
@anildash This resonates so strongly with me. I find with so many new technologies (AI, blockchains, prediction markets), people on one side declare them as evil bullshit, and people on the other side over-hype them to try to quickly make loads of money off of people. And I always find myself in a middle ground that they are cool technologies, how can we deploy them in ways that empower people, rather than extract and harm to enrich a few.
-
@shimon@kolektiva.social 2026-06-09 04:29
@anildash@me.dm I eagerly clicked on the link, expecting to find an open-source LLM to run on my laptop so that I could run my agent-based models locally, but instead it was a dedicated video-processing app... nice of them to make it open-source, but it's not quite an AI (even by the LLMs' limited standard).