A lot of us here hate AI because of how it was built: training data gathered without the creators’ explicit, opt-in consent, data centers that negatively affect communities’ access to clean water and energy, a technology design that is inherently prone to hallucinations, etc. At least, those are the main reasons why I hate it.
I think I might actually want to support an AI project if I thought it was being done right. Maybe we could get more people away from exploitative models if there was a non-exploitative alternative.
So what would it take to build AI ethically, in your opinion? And do you know of anyone trying to build AI without these issues?
There’s https://apertvs.ai/ probably as close as it gets. Government funded, made by universities. Afaik its datacenter is powered by hydro. But it is an academic project, so still uses common crawl and other publicly available datasets, which is considered ok practice in academia but still means consent is opt out, if something is publicly available. And of course no one uses this, because no marketing and it’s not as ‘good’ as models trained on stolen data.
Plus you could still argue that the energy and tax payer money could be better spent elsewhere.
That looks far better than the mainstream AI tools, but I don’t think respecting opt-outs is quite enough. It would be so much better if it were built from solely opt-in training data. As far as I can tell, it’s not attempting to tackle the hallucinations or environmental impact issues. Still, it would be a major change for the better if ChatGPT users switched to something like that.
yes, agreed. And while right now its environmental impact is low due to renewable energy use, I have no doubt that if this caught on and had to serve more users, the environmental part would get just as bad as other LLMs.
Universities are the tip of the baby killing military industrial complex. you people are all so silly. There is no such thing as ethical AI.
We live in capitalism, nothing doesn’t contribute to the orphan crushing machine, but we can still try.