A little while back we reviewed the reading algorithms doom loop. If you haven’t seen it take a look here or some context. I wanted to add to it based on recent articles which are coming out about the elephant in every room. Artificial intelligence. AI.
A new kind of doom loop is forming, and nobody will have any say in the end result of it but you. And me. And your friends. All of us. But it is starting to look unlikely that the decision will be made by government agencies or the law in the near future. The doom loop I am referencing is the doom loop for AI.
There have been many, MANY articles in the news recently about ChatGPT, and other large language models and their use of data to construct those models without reasonable payment to the source material. Here is just a small sampling of relevant articles:
- In India
It is starting to seem like nobody wants large language models in use. Similar lists of lawsuits can be found by artists and the loss of revenue for the alleged theft of art work for items like Midgard, Dalli, and other digital art generators. They have even won competitions in some cases, driving the ability of artists to make a living down hill.
I have read articles which say between 45 % and 67 % of people think that large language model companies should have to pay for the data on which they are trained. Europe even released reports on the possible detriments of LLMs. If so many of us don’t seem to want this, why is it happening anyway and what can we do about it?
First for transparency: I have dabbled with both art generators and writing generators. None of the pieces on this website have ever been AI generated, though part of the problem is you can’t prove that. Contrary to the hope that AI can spot AI, for now, it can’t. I have, on the whole, been grossly disappointed with the output of both the art and the chat conversational skills. I have caught it in lies, and I have caught writing just poorly regurgitated information. It is a skimming of surface information without any depth, because for now, it doesn’t understand anything. Not really. It isn’t general AI, and it isn’t sentient or even close to taking over the world, so worries about Skynet coming can relax. And. Well. It’s boring. It’s not me making those statements even if it could make good content.
What I am going to say for the concern is something entirely different. AI in this format is theft. If tomorrow I would write a story taking place in Westeros I would need to give credit to the creator, and give royalties if I make any money on the use of that intellectual property. LLMs and the people who use them do not. There is a fine line between reading something and taking steps forward with the information provided, synthesizing the data and coming up with your own understanding, but that is not what LLMs do. They regurgitate wholesale items which were scraped up to train on. The NY Times claimed they were able to find the content buried wholesale in a LLM of their websites content on the other side of paywall content that was not paid for.
We have a choice as a group.
It will be a tangled mess for years in courts to decide what happens to AI and its right to use your material. But we can decide today what we want to do about it. Just as what you read determines what happens in the world, what software you use does to. In defense of writers, artists and creatives I will not knowingly use AI generated art, I will not use AI generated text, and I will no longer “play,” with these software online, because to do so is to patronize something I do not agree with.
You have a choice too. AI has its place in many other contexts, and AI is not coming to conquer our world, but it is starting to conquer our artists. If you wrote a book tomorrow, would you want someone else to write the second novel in a day with an LLM? If you made a work of art and Midgard stole it, then made ten versions for ten customers, and you never see a dime, wouldn’t you want credit and cash? If you entered a contest with all your artistic skill, and a hack entered the contest with some basic AI prompts, and got an AI to create a masterpiece, would you be willing to lose the competition to the AI? Would you want your writing entry skipped because the magazine needed to make room for ten regurgitated AI written pieces?
We as a group get to decide what to do about it. I’ve made up my mind. How about you?
Comments