Discussion
I really do not think the two strong fronts in the AI/LLM debat, which we could call “hard against” vs “strong believers” is the best for our global society.
I’m neither hard against and nor a strong believer. Mostly I’m concerned of all the negative concequences of how the major companies are developing their models, the economic recklessness, the environmental impact and the problematic use of copyrighted material (copy first, take it in court later).
To me this is the most central: Is there a way for the AI/LLM use to develop in an ethical controlled way?
But when the discussion is only “Not at all” or “See this new fancy model which hacked the competitor”, we do not get a balanced value driven conversation. The result is that people keep using the major companies because of their marked share.
@mosgaard open weight models were still made unethically. And legitimizing LLMs, even indirectly, is harmful, given the immense and multifaceted harm they are inflicting directly and indirectly. So I’m an unabashed hardliner against. The fact that they may be useful in some narrow fields is not worth the cost to nature and society, and we were able to manage without them until a few short years ago. Plus being useful is the lowest possible bar. A rock or a piece of used bubblegum can be useful.
@sknob In the case where an open weight model (or a closed sourced model for that sake), is developed ethically, this would be different right?
Can a model be ethically developed (and used for that matter), and can we even believe it can, with the unethical interests controlling the industry? Like this initiative https://apertus-ai.org/
I think people taking control over (otherwise obsolete) hardware or spying televisions is a pretty good cause, working against e-waste and digital surveillance
@mosgaard I think initiatives like that legitimize and whitewashes a technology that should never have left research labs, and that has been weaponized against the world. In my view, there is no such thing as ethical use of generative AI.
@sknob how do you think it has been weaponized against the world?
@mosgaard not the best link, but a good start (I have plenty more) : https://whynoai.info
I’m also not blind for how LLMs can be a strengths for different tasks. And I think these strengths is a valid argument for why we should try to regulate and control the AI industry.
What I can’t see or understand is how people can support the major unethical companies with their clearly uncontrollable “state of the art models”, when you can use less capable - but still capable - models from other companies or local open weight models on your own machine.