With the current ruling in the Authors vs Anthropic case, I thought a few words of commentary might be in order.
First, a summary:
The court found that Anthropic’s use of existing works for training purposes was NOT infringement but met the definition of a “transformative” use of those works (if this bothers you, harken back to the Hugo Awards committee’s decision that the AO3 website that publishes fan fiction engages in “transformative” works and was therefore eligible to win a Hugo Award), in essence echoing the same conclusions we came to here at Amazing HQ – that using existing works to train AI LLMs is essentially the same as an author studying someone else’s work to aid their own, or an artist studying a painting to work out technique – although admittedly done on a much more wide-ranging and detailed basis (not to mention on a far more granular scale). Basically, AI training does what people do, but on a much more efficient level.
The courts will be hard pressed to rule otherwise (especially in the face of the well-established principal of the first sale doctrine – the purchase of a book grants the purchaser the right to do with that book whatever they like – read it, lend it to a friend, use it for kindling, share it with an AI, and, apparently, write my own version based on it – but not the right to copy and distribute and sell) as absent some new laws that separate normal and customary human activities from those same activities with an AI in the mix would be necessary – and that would probably lead to a slippery slope of trying to determine how much of a work was technologically aided and how much wasn’t. “Oh, sorry, you used a word processor to write your novel, you can’t copyright it now….”
One thing I find very interesting (and telling) is that in every PR on the subject that I’ve read from sources representing “creatives”, this aspect of the court’s ruling is greatly down played. It is, in fact, the major part of the ruling. Anthropic’s use of a “pirate” library as the source of the works was infringement, but infringement of a completely garden variety of illegality. No new law was created or expanded with this part of the decision.
In fact, if I got together with a goodly number of my collector friends and each of us scanned and uploaded copies of the publications in our collections to a “library” that was then used to train an LLM, we could probably come close to the number of works used, and that training would be completely legal (First Sale) – as none of us are distributing and selling anything. We’re just showing an LLM some of our favorite works….
Which brings me to another aspect of this. Yes, I know, the primary focus for most creatives is the lost revenue/not being paid a fee for use and rights being infringed – but if that is the case, those agreeing to the current settlement are selling themselves short, which brings in to question actual motives.
Why do I say that?
Because – The owner of a registered copyright can choose to recover statutory damages instead of actual damages and profits. For each work infringed, statutory damages range from a minimum of $750 to a maximum of $30,000, and – IF the court finds the infringement was willful, the maximum statutory damage award can increase to $150,000 per work.
PER WORK, while what is currently being offered in the proposed settlement is a mere $3,000. A bit more than the minimum of $750, but a LOT less than the maximum, especially considering that I think the case for “Willful Infringement” is a pretty solid one.
(Note that this only applies to works that have REGISTERED copyrights. Apparently, as we have learned, there are a fair number of infringed works out there that the publisher failed to meet their contractual obligations and did not register the copyright – there’s a whole slew of potential lawsuits coming down the pike in those circumstances, lawsuits not based on copyright infringement, but on breach of contract.)
If the current settlement is enacted and upheld, it would be good evidence for establishing that “willful infringement”, in which case it would seem kind of foolish for anyone to join that class action suit, rather than hiring their own attorney and going after the big bucks. (One third of 150k is still $100,000.)
But that’s a long aside, which should be accompanied by a strong suggestion that you consult with your own attorney on these matters. On the basis of the preceding, what I really want to address is the following:
Why are you all so upset that from now until the end of time, a technology that is going to permeate and dominate our world, was influenced by YOU?
Most of you are not earning a living from your writing, even with multiple novels currently on the market. Many of you are earning a nice “supplemental” income, but still need to keep your day jobs, and, unfortunately, that’s unlikely to change for the foreseeable future.
By the same token, there are very few among you who would admit to writing only for the income. The vast majority of you write because you need to, and you’d write and seek publication whether you got paid or not (just as I am doing here). The money you get paid is a nice affirmation of your skills, but if it were suddenly mandated that no one can ever receive any form of compensation for doing creative, artistically related work, you would all still do it.
I find myself actually quite flattered when a question asked of an AI comes back with thoughts and words resembling my own. (When it comes to Amazing Stories and the history of the SF genre, I see a lot of that; I even had one offer up my own words as a quote to support its response.)
Owing to the nature of AI, that “data” is going to be perpetuated throughout time. It’s part of the training…there are nodes in the neural network related to MY stuff. AI will be using my words and thoughts as it responds to queries from people generations from now (which, admittedly, some of you may find horrifying, but this is apparently what happens with AIs when you put a lot of your stuff out there for anyone to read).
Maybe, at some point in the not too distant future, I’ll get some egoboo from being able to tell an AI “Hey, you’re quoting ME!” and getting back some canned positive responses about how insightful and creative I am.
It won’t pay the bills (on the other hand, neither will $3k from Anthropic – that’s barely enough for grocery shopping these days), but it will likely mean that some kid, decades from now, will win an online argument using my words.
Similarly, if your works have been used to train AI “how to write good”, generations of future authors will be your legacy. YOU will be among the “giants” upon whose shoulders they stand. Just like Melville, or Steinbeck, Shakespeare or even Homer. They may not know you by name (and, lets face it, how many kids these days know the names I just mentioned?), but your influence, your creative DNA, will carry on, long after the last copy of your last book is printed and sold.
A. Bertram Chandler (one of my faves, who is rapidly fading from the scene) once wrote about a “Hall of Fame”, a pocket universe in which all fictional characters still lived – so long as someone, somewhere, was still reading them. AI now offers us all a place in that Hall of Fame, permanently (or at least so long as AI is with us, and, I suspect, probably long after we are gone).
You, or a part of you that I strongly suspect you hold near and dear, has been granted a form of electronic immortality.
Steve Davidson is the publisher of Amazing Stories.
Steve has been a passionate fan of science fiction since the mid-60s, before he even knew what it was called.
