Post

I think a lot of people in tech are probably feeling something close to despair about what to do about LLMs, since it’s clear that those with power are not going to relent with their advocacy, and they have enough uses that they’ve gained critical mass, yet the harms & risks are so wildly egregious…

I think the answer has to be a multiplicity of responses. There need to be *many* alternatives, created by many entities. The answer to McDonald’s being unhealthy isn’t “make another fast food joint” — it’s getting rid of food deserts and improving nutrition overall. Give power back to the people…

21 reposts223 likes

I've used the McDonald's analogy a lot with smaller software creators or service providers who get bummed out seeing exploitative garbage attract a lot of funding and attention. McDonald's exists. Don't worry about them. Figure out what your role is, what you offer, etc.

I think people fixate on the flashy startups with billion dollar valuations and miss out on all the successful small to medium businesses that provide a solid living to a handful of people. Success doesn't have to be what the media tells you it is. You get to define success.

If people want some of the things AI can do, have those technologies be provided by tools & platforms created by cooperatives, by unions, by universities and governments, by municipalities and by individual creators, or as open source owned by nobody. Destroy the economic value of proprietary tools.

“Open source” doesn’t mean “owned by no one” … but I think you’re right that opening up the model’s training data and weights is a good 1st step. I’m not an expert on libre software licenses as they would apply to an LLM, but having a competitive, public domain AI is in the public interest for sure

There is a long history of communities effectively competing with closed platforms — and certainly a worker’s coop could make a tool that is better for the environment, better about intellectual property, not prone to hallucinations, and that has safety guard rails so it won’t incite self-harm.

Coming back to this thread... After deeper research I'm finding lots of people who've built LLM products who are excited about the prospect of individual software developers spending $200 a month on side projects. Sounds to me like some people are dreaming of a world where wealth builds code.

And it could even be regular-ass companies! An insurance company could make an LLM that has a lot of use for a lot of folks and isn’t evil! Or a supermarket. Like… it’s not exclusive to giant evil tech tycoons!

Love this. I just started thinking about what it'd take to provide an LLM to at least my neighborhood community. Do folks in this thread know good resources to start learning in this direction?

There has been a forty year effort to make this sort of thing happen in software that has been basically ignored by the organized political left for aesthetic / vibe reasons (this is the reason I despise the political left). I'm not holding my breath I expect vibes will once again carry the day.

If you despise people for not understanding stuff then you kind of have to be a psychopath. For example, you don’t understand human nature.

Watching the last decade of shit unfold onto the world as opposition is completely and, to my eyes, wilfully unprepared for it has been a souring experience. I’ve been working these trenches a long time. The left don’t want to do what it takes to win the 21st century, and that may well kill us all.

Hard agree here because I’m a Canadian watching the US careen off into extreme RW chaos and accept so many awful things, while every individual American I know is kind and decent. Most worldwide don’t have education to understand tech like this, but we expect our Governments to take precautions.

oh that was a poor choice of analogy, I see, sorry I wasn't meaning to focus on the energy cost of ai, just that a worker owned company isn't a total solution because the workers that now own the company would like to make money

strongly agree with this but it is hard to imagine a world where coops/etc have both the resources and the inclination to lawlessness to recreate the most powerful models, leaving users to choose between kneecapping themselves or participating in the problem

not using this as thought termination, more restating the power imbalance problem; current models are so powerful precisely because they are made by people who give no fucks, competing with today's SOTA with any level of care will be quite hard and may take years, while SOTA continues to advance