my stance on 'AI'
Yesterday I was asked about my stance on AI.
First of all, the question is wrong.
What they wanted to ask me is what I think about Large Language Models (or LLMs, like ChatGPT), “artificial intelligence” is a marketing term and very broad. Anyway.
The answer, as with everything complex, is: I’m not sure.
I don’t believe in copyright as a system for protecting rights.
I believe in a system of collaboration and the universal generation of knowledge.
I believe that we are not as original as individuals, but rather a collective construction.
No matter how much self-help books try to convince us otherwise (and liberal individualism, which is the zeitgeist).
This same belief is what turns my answer into “I’m not sure”.
Because on one hand, LLMs are a first step in building a global cultural tool.
But on the other hand, I am not very happy that two or three companies are governing the outcome of this.
It’s the same old discussion, GNU/Linux vs Windows all over again.
In a couple of points, I believe that:
- It is hypocritical to use copyright as a flag against technological progress, given that we are all a collective construction. No one lists all the books that influenced them, maybe we mention one or two in interviews.
- I am concerned about technological concentration, a few companies dominating the field.
- I look with much love at what is happening with open models (LLaMA 2, BLOOM, etc), although I would like them to be more open: to open not only the weights, but also the datasets, and processes.
- I am concerned that writers with technological aversion and stubbornness may not be able to adapt to the new models and end up without work. Same thing with visual artists and other areas that are affected.
- I am concerned about the second-order consequences of technological progress in general, but this is just anxiety, I can’t do much about it.
- We are going to have a new period of information deterioration (articles generated by LLMs, fake news, a lot of garbage).
And above all (although this is the bottom): - I believe LLMs are the future of interactive fiction.
The use of RAG (Retrieval-Augmented Generation) in combination with models fine-tuned to our writing style will become the norm.
That’s why I think it’s important that, while the world is having the ethical conversation (whether we are part of it or not), we have to learn to use these tools and be part of the exploration.
If you don’t feel comfortable making money out of the results, then don’t charge, but explore anyways, tinkering is what creates gems.