depends a bit on the goal of the FOSS as to how important copy-left, licensing, etc. might be but attribution is the crucial underpinning regardless of which licensing you do or don't do. who did the work is vital to know.

I don’t care about “proprietary.” What about intellectual property theft? What about damage to the environment? What about diversion of SO much resources away from things that help people and into things that help investors?

Every time someone uses it, especially as a technology leader who knows better than most, they signal that all those trade-offs are acceptable for the trivial, tepid benefit of writing bland text faster. And the benefit of firing people. And the benefit of wealth aggregation.

In my position, I have the financial means and the knowledge to fight against it. If those of us who CAN choose not to use it, instead choose to support it, how can we expect anyone else to make the pragmatic, better-for-everyone choice?

Nobody gets to hide behind “this is my personal choice and nobody should tell me otherwise.” If a person set their trash on fire in their yard in order to dispose of it, their personal preference doesn’t matter. That’s not ok. Burning fossil fuels to power the most catastrophically inefficient technology ever made is not a personal choice. Climate change affects generations to come, not just people alive today. AI today accelerates climate change. It’s super simple and absolutely directly causal.

As for the hardware:

The mention of doctors reminds me of this piece, which I’ve been thinking about a lot this week: https://www.medpagetoday.com/opinion/second-opinions/122202

My own personal doctor may be an exception. She has sought a unique model that allows her to focus on care. I’m fairly sure she doesn’t like the robots.

I suspect many docs are enthusiastic re: robots because they think it can help free them for care.

Surely, this is one of the primary reasons genAI was released.

Did they *train* the model on that hardware? The model has to come from somewhere, and last I heard, that “somewhere” was billion-dollar data centers.

Also, how the heck did they get that kind of hardware for only €3.5k in the middle of the chipocalypse? I'd probably have to pay that much just to assemble a competent gaming PC.

@bkuhn @copiesofcopies @cwebber @twipped

“You will never be able to afford the hardware to do so” I know people running a model equivalent to top end Claude from ~6 months ago on €3.5k hardware. Not cheap, not awful.

2/3 I would almost certainly start from someone else’s model release

IMO even local models are currently unacceptable when they have been trained on unknown data, by untrustworthy people, using far more resource than any community will ever have access to, with a special privilege to ignore all IP considerations when it comes to accessing and storing their training set and using all web data that no one ever agreed to be used in such a way.

Training anything that way should still be regarded as "the largest ongoing intellectual property and labor theft" and ethically regarded as if in biology someone wanted to clone a human, but if someone creates a local model that fits on world-accessible consumer hardware (not just affordable to first world countries), trained on known data explicitly licensed for reuse (or public domain) maybe it will be time to look at the implications on society (in learning, community building, work for a start...) and collectively and democratically decide if it is worth putting resources in…

@copiesofcopies @cwebber

As for FOSS' success, this wave of LLM infiltration has made it abundantly clear that much of FOSS' success is thanks to corporate sponsorship, in exchange for influence, and now they've decided to exercise that influence by enticing and/or coercing FOSS maintainers into destroying their own projects.

I am no longer confident in FOSS' survival.

And here you are, cheering its demise.

@twipped @cwebber @copiesofcopies

And we shed them as soon as it became practical to do so.

You will never find it practical to shed Claude, because:

1. you will never be able to afford the necessary hardware to do so (Moore's law is dead),

2. you will never obtain an adequate corpus of training data (because you, lacking billions of dollars, are actually subject to copyright), and

3. you will be prohibited by law to do so (because compliance with AI regulations costs billions).

@twipped @cwebber @copiesofcopies

This is the question those LLMs companies have forced us to answer, thereby breaking the alignment that existed so far between industry & “open source” on one hand, and people focusing and user freedom and autonomy.

@copiesofcopies @bkuhn