Someone used Midjourney to AI-generate images of politicians cheating on their spouses — though claims that it was well-intentioned.
There are lots of open source projects involving AI that you can run on your personal computer. I think the community-driven projects are heading in the right direction, but it’s completely opposite for the ones owned by corporations as they’re only driven by profit margins, not people.
The problem with “open source” in the context of AI is that the source code is a much smaller factor than the training dataset. AI companies running around and scraping everybody’s data as if they own anything they see is a real problem raising massive ethical and legal concerns.
That’s great (genuinely), unfortunately having to work outside of the mainstream brings its own hurdles -this isn’t on the same level but consider twitter vs mastodon or reddit vs lemmy: the corporate solution is shiny and easy and requires very little to no effort from the end user to use, while the other requires a little more understanding and effort and comfort with technology, and might not appeal, or even be known, to many. Sure, people can look it up and learn it, but that looking and learning are hurdles, and when it comes to accessibility devices, those hurdles tend to be more significantly in the way.
To be clear, I am not trying to shit on the open source stuff, I do genuinely think it’s great, but like so many of the solutions we currently have to work with, it’s a band-aid on a cancer. We need to remove the cancer.