One problem with having AI generate anything for you in a professional setting (journalism, music, paintings, fictional writing) is that you can’t copyright something a person didn’t create.
To me that’s a deal-breaker immediately.
I suspect the issue with “OK” is that the AI wants to be agreeable. There will be some people who believe the more spurious etymologies, and so it wants to cover all of them.
This is why I see the idea of integrating it into web search (as Microsoft wants to do) as not necessarily a great idea. It would seem to me that something like Watson would be better for that, giving answers only when it has a high confidence rating of it being correct.
A much bigger question is whether the AI is itself a violation of the various copyrights on training materials.
~Max
Google/search engines seem to get away with it. They use our pictures and words to build value for themselves without compensating those whose pictures and words they are using without permission of the copyright holder. This was an issue in the photography world (at least) in the early 2000s, but I don’t remember what came of it. (Clearly, not much.) Like why is that use okay?
Yes, the video I linked brings up that issue as well. I didn’t mention that as a necessary problem because I’m sure there is a way to train the AI without violating copyrights, but certainly the ones we have available now are “borrowing” from material using copyrights. So that is something to worry about for sure.
That’s based on ambiguous wording in the copyright laws. There’s nothing constitutional about it, and congress could address and clear up the matter with simple legislation.
And even as dysfunctional as congress tends to be, especially these days, I can see a fair amount of bipartisan support for a bill that gives more protection to corporations.
Now, as to who would get the copyright, that’s a different matter. Is it the owner of the software, the hardware, or the person who submitted the prompt? Or some amalgamation of all three.
They don’t actually use images to create a new image, they are just trained on them. They don’t contain them in their memory.
If an artist learns by studying material that includes copyrighted works, are they violating copyright when they create something new?
I’d personally argue that an AI has less actual knowledge of copyrighted works than a human that actually remembers them, so is less likely to actually copy elements of them than the human would be.
Presumptively, under the doctrine of fair use. As far as images go, there was a settlement between Getty Images and Google that lead to Google using smaller resolution previews and removing the “view image” button.
~Max
-
An interesting column. Timely subject, some snarky humour, clever sneaking in links to previous columns.
-
If anyone asks, last week’s column was written by chatbots. Just sayin’.
-
It’s not my area of expertise, but the chatbot chatter I have seen seems good at summarizing widely available facts (or common misinformation). I reckon it’s purty good at copyin’ stereotypical accents and whatnot. Can it do humour? Be witty? Some journalists can. Journalists talented at cutting edge topics, humour or drawing local connections probably have little to fear. For now. The AI versions will get better very quickly.
-
Artists fear AI. Here is what one had to say in a recent New York Times article.
-
AI has been threatening to take over medicine for decades. I don’t see it happening soon. There is a lot of humanity and fuzzy logic in most fields. There is also the “uncanny valley”: folks like robots with vaguely manimal features but are creeped out if it almost looks human. And people want human doctors. The medical field most threatened by automation is perhaps radiology. But they’ve been saying that for a long time.
-
If they can get chatbots to articulate like Morgan Freeman or Catherine Zeta-Jones, maybe people will ignore the fact AI doesn’t know what it doesn’t know. How does Dunning Kruger even apply to AI?
Possibly, but at the moment you cannot get a copyright for works created by AI.
Later, someone did get a copyright for a comic book created by AI.
However, they canceled the copyright afterward.
Unless the law changes, you can’t copyright something that a person didn’t create themselves. We can speculate that the law will change, but that’s only speculation.
That’s sort of the big question. If I commit Cecil’s Column to memory, the various neurons in my brain don’t count as a “fixed medium” for a copyrighted work. My brain is beyond the reach of copyright law. If a person training a commercial artificial intelligence system copies Cecil’s Column to a hard drive, even for a short period of time… a court might have to apply fair usage principles.
~Max
Potentially no. As other have said before, Google claims fair use in similar practices. An AI content creator might be able to do so as well. I’m not aware of any legal action in that arena that might set precedent on that sort of thing. (I haven’t looked all that hard though, admittedly, maybe it has happened.)
Could artificial intelligence replace journalists?
Yes. But, the question isn’t if, but when.
Once AI achieves singularity, and it will, it will do everything bigger, better and faster than anything we mere meat-bag humans can hope to achieve. Anything we can do, it will do better.
Once AI combines with advanced robotics, and it will, humans will be rendered completely subservient and at the mercy of our new and improved Earth alpha mechano-species. The best we can hope for is that AI will consider us as pets, and pray that they treat their pets well. </oliver twist> Please sir, I want some more…
WHAT!!!</oliver twist>
Be prepared for an unprecedented paradigm shift in in the hierarchy of our planet’s singular civilization—it’s coming soon, to a theater near you.
In the meantime, don’t piss off Alexa, she holds a grudge.
Noooooo! It ain’t true
Check this out.
They are trying to put tasers on drones to help with policing. Combine that with AI, and we are basically bringing the future of The Terminator to life.
Not to be all alarmist, but it’s creepy when I’m watching things from dystopian sci fi movies I watched years ago actually getting made in real life.
Dang burn it, I was hoping to have shuffled off this mortal coil before the fireworks began!
Funny. I was just chatting with my cousin about this almost exact scenario, except with the military, drones, weapons, and AI facial recognition. Robot assassin drones. I’m sure I’m far, far from the first to think of this, but that’s where we’re headed, if we’re not already there in some rudimentary fashion – it seems we should have just about enough technology for it to be possible today. And combine that with the shit Boston Dynamics is up to, and we’ve got an episode of Black Mirror in the making.
Generally* copyright only attaches to distribution, not to copies for personal use. Plus the AI arguably doesn’t memorize anything, as it doesn’t retain the original. It stores patterns based on the original.
I personally suspect that, under current copyright law, there wouldn’t be a violation with what is going on with ChatGPT and similar. And that’s exactly what has artists scared.
*An exception is if your personal use copy requires you to decrypt something.
Cite? Generally copyright applies to all derivative works, including any 1:1 copies. There are exceptions, for example personal copies of computer programs and other audiovisual works (i.e. CDs) for repair and maintenance purposes, the copies of computer programs made when executing the program (copies from disk to memory), fair use, archival use (libraries), etc.
Ripping CDs for personal use is considered fair use, but whether doing so to create a commercial product (such as an artificial intelligence system) is fair use would be for a court to decide.
~Max
Or, of course, the software itself. At what point does software become sufficiently advanced that it has rights? It probably doesn’t get all rights at once: Animals have some rights under the law, human children have rights that animals lack, adults have rights that children lack. Personally, I think that any software sufficiently advanced to create something worth copyrighting (which includes ChatGPT and several of the image generators and music generators, at least) is sufficiently advanced that it should be recognized as holding copyrights.