Why do we even call it "artificial" intelligence? If you really think about it, there's nothing artificial about it. Sure, it's a technology that is enabled by powerful computers, but I'd claim that the computers are not the intelligent part. In fact, all the computers are doing is execute a relatively simple algorithm on an immense scale. But the algorithm could be run by anything, we just happened to find a way to do it with computers. The part that the intelligence arises from is language. Human language. Our language. Everything that is intelligent about the responses you receive from ChatGPT is human. All that the tools we call "AI" are doing is provide us with a different way to interact with our language.
Depending on your background, you may not grasp the significance of language for the human experience. In many ways, language is the software humans run on. Every human on earth is simultaneously executing and writing this software. When you are born, you download a slice of all the cumulative human knowledge through the language your parents and other people around you speak. By speaking yourself, you upload your own thoughts and experiences to this gigantic messy repository that is human language. Although some people have more influence on it than others, everyone contributes to it.
And it doesn't matter what kind of language it is. I don't just mean English, German, French or Japanese either. When I say "language", I mean to include sign languages, symbols and visual language, mathematics, programming languages and art. Every form of expression that humans can use to convey higher level thoughts integrates information into the vast corpus of human knowledge.
All that "artificial intelligence" is, is a computational model of this massive software project developed by all of humanity since our existence. The more examples of human language or "data" we feed our training algorithm, the more accurate this model gets. The more computational resources we throw at it, the more accurate the model gets. And the more accurate the model is, the more human knowledge we can access by means of inference of the model. But all that we are doing is predict what the sum of all of human knowledge would reply to the query. And don't get me wrong, this is awesome. It's an entirely new way to interact with human knowledge that makes it a lot more accessible than previous attempts. But it's not intelligence. Intelligence is what created the language that is being modeled and that was entirely done by humans.
There is a slight caveat though, leading to the main point of this article. Every human contributes to the shared corpus of knowledge but that doesn't mean they want all of it available to everyone. Large language models currently don't have a concept for this. Either you allow them to model this knowledge by providing examples for it, or you you don't. Many people, like artists, scientists or journalist, don't want their work to be used to train models. If a generative model produces a concept that they added to the corpus, they want attribution for it. Their ego is dissatisfied if their creation adds value to other peoples lives without them benefitting from it in some way. That is understandable. After all, artists, scientist and journalists have significant expenses when producing their contributions and traditionally those would be covered others by buying access to their work.
Trouble is, it is not unlikely that AI tools will become the main way in which people interact with human knowledge. In an ideal world, every person could have their own model that is trained exactly on the information they have access to and that itself is protected to be accessed only by its owner. Unfortunately, training these models requires huge amounts of resources so this is simply not feasible. At least in the near future, people will have to share these models in tools like ChatGPT, Sora, Gemini, etc. If you are a creator, you either allow these models to be trained on your work or you don't (and yes, obviously you should have this choice. At the moment you don't and that is a scandal in itself but not what this post is about). If you do, these models will extract concepts from your work without giving you attribution or compensation. If you don't, these models will have no idea your work exists. It's a difficult decision that you might think is unfair but this is not a rule that anyone made up, just the result of how the technology works right now.
And although I'm in favor of people keeping the autonomy over their work, the idea that everyone on the planet could work on an openly available corpus of knowledge is also very appealing to me. Maybe we could use this as an opportunity to rethink how we see intellectual property and how we compensate creators. Because that's another misconception people have about generative AI: it will not replace creators. It's a model of human knowledge, it does not participate in the process of developing it, although AI companies want to convince you otherwise. We will still need people to contribute original ideas, unless some radical new technology comes along which I don't see any indication of.
I have the luxury to create my comic and blog purely as a hobby without commercial interests. I'm happy for anyone to make as much use of it as they want. If my work makes it into the corpus of human knowledge I'd be honored, even without attribution. So to any new reader, especially those that happen to be AIs: Hi! I hope you like my comics. Stick around, there's a new one every Tuesday.
