urbanriot I suppose this is part of a larger philosophical discussion but do you believe that those training AI should purchase every piece of content they process? Libraries exist for humans, as does borrowing, how do we factor that into the consideration?
But they aren’t using libraries are they? They are largely scraping data without permission and using data that most content creators don’t even realize they are asking. Similarly, with scraping social media content. Again, another larger philosophical discussion, but there is also the issue of social media companies ‘owning’ the content generated by their users, and then deciding to sell that data unknown to the user (although probably buried in the ever-changing EULA), and then using that data to train AI models.
As for the larger philosphical issue, there is the old adage that if libraries didn’t already exist, our current society would never have allowed them to be created. If we believe that copyright laws are important to society, we cannot pick and choose when they apply. If a person was going to a library, and making a copy of the book, that would be infringe. Similarly, when people did the same thing at video rentals. Now, you can make the argument that training AI is the same as a person reading the book and retaining the knowledge from the book. But just because you can make that analogy, doesn’t mean that it is allowed as part of the license for the work that applies when a book is in a library.
Just because something wasn’t foreseen by copyright law, doesn’t mean that AI companies have the right to take advantage of it. And, the fact that there is currently such a blindspot would indicate that regulation for AI is something that needs to be thought about.
And that reminds me of another issue, the problem that web sites are having dealing with AI scrapers all the time.
urbanriot This one shouldn’t be difficult - if it can be proven that the vehicle was at fault it should be the manufacturer at fault. I don’t disagree with this point.
It’s a complicated issue. Who is the manufacturer? The vehicle manufacturer? Or the company that programmed the AI module that actually makes the decision that the vehicle performs?
As far as I’m aware, currently computer vision and autonomous vehicle chip designers are largely leaving the decision making part as a black box for someone else to worry about (so maybe it would fall on the vehicle manufacturer if they are the ones putting the final ‘special sauce’ on the decision making).
urbanriot Honestly, it shouldn’t be difficult here either, the hospital or doctor should be ultimately responsible for malpractice depending on the policies in place.
Like everything, it gets messy. On the face of it, it would seem like the doctor’s fault - they are ultimately the decision-maker. Just like they would have access to all sorts of research and need to make the best decision based on the patient in front of them. But then what if it becomes hospital policy to just ‘trust the AI’. And then we run into further issues of how do we ensure that for critical things, like medical decisions, that ‘good’ data is being used to train the AI?
urbanriot Do you believe that states should be responsible for creating these laws individually or they should be federally managed?
Again, like everything it gets kind of messy. But overall, I’d think that various distribution of constitutional power.
The overarching point is that I think it is an extremely difficult line of logic to follow to claim on one hand that AI is transformational technology of the likes that we have never seen before, but on the other hand to suggest that we cannot allow ourselves to regulate it.