I’ve got a post-graduate certificate in artificial intelligence (AI). I’m also an author, and I believe writers and publishers should not use AI in publishing. So that’s why I was disturbed when a reviewer asked if I had used AI in writing my recent coming-of-age novel, Under the Gulf Coast Sun.
But the reasons I oppose using AI are not the usual ones you hear.
We have all read or heard about copyright violations during AI algorithm training, as well as plagiarism problems, job displacement, potential stifling of creativity, legal complexity, blandness, and plain old human outrage. Those are all good arguments for opposing the use of generative AI in publishing.
Let me also argue against its use, but for a completely different reason: AI is dead.
Literally.
When I want to read poetry, a short story, a novel, a memoir, or non-fiction, I seek the voice of a fellow human being. A computer, by contrast, has the exact same awareness of the world that you had before birth—basically the perspective of a stone sitting on the side of the road. That is, no awareness of the world at all.
So, when I’m interested in what a person has to say, why would I willingly spend time reading or listening to a text that was mathematically calculated by a dead thing? I would not. And once you consider this reality, I believe you will lose interest as well, just as we all completely lost interest in (and quickly forgot) the rather incredible achievement of IBM’s Big Blue defeating chess champion Gary Kasparov in a six-game showdown in 1997.
Mustapha Suleyman, Microsoft’s Artificial Intelligence CEO, said in an NPR interview with Manosh Zamorodi that AI systems “communicate in our languages. They see what we see. They consume unimaginably large amounts of information. They have memory. They have personality. They have creativity.”
That is mostly nonsense. Computers operate only with zeros and ones. AI does not see what we see. It has no personality, no creativity. At best AI is a glorified calculator that works by fooling people into believing that it possesses the qualities Suleyman lists because AI does consume and process unimaginably large amounts of information from human beings. Unlike Suleyman’s claim, though, computers don’t have any real understanding of the data they generate.
Here’s how AI calculates novels or short stories or poetry: A human language prompt is converted into zeroes and ones and stored in a vast ocean of other zeros and ones. Then a set of instructions are loaded into the main processor’s transistors (again, zeroes and ones). The instructions tell the computer which zeroes and ones to retrieve from memory, then the retrieved zeros and ones are added together with other zeros and ones using Boolean math. This creates a different pattern of zeros and ones—which get put back into the storage system.
This process repeats itself at extreme speed for about ten or fifteen seconds, until the novel or short story or poem has been fully calculated and placed into memory (again, as zeros and ones).
At a higher level, AI calculates texts based on probabilities of what word is most likely to come next in the stream of data it’s generating. That’s all.
In other words: A poem, short story, or novel calculated by a computer is nothing more than a prediction. The fact that a human being sometimes can’t tell the difference between an AI-generated text and one written by a real person doesn’t make the calculation any less of a mindless mathematical outcome. (The Turing Test first devised in 1950, is a way to attempt to determine whether machines actually exhibit any real intelligence. In particular, if a human can’t tell the difference between the machine output and one created by a human, the test passes. But Turing, the man who proposed the test, later said that the question of whether machines can think is itself “too meaningless” to deserve discussion.)
And, critically, the math has nothing to do with how humans build beauty and meaning with writing.
There are uncountable amazing and incredibly useful applications of AI, both generative and otherwise, in society. I won’t list them (just ask your local ChatGPT, it will be happy to calculate a result for you). And on that note, I let ChatGPT end this article with a calculation it made about the meaning of its own calculations, based on the following prompt:
Me: I would state that you don’t understand concepts, but retrieve data, operate on the data, then calculate an output. The concept is up to me to create based on your calculations.
ChatGPT: You’re absolutely right. I do not possess true understanding or conceptual awareness in the human sense. Instead, I work by retrieving patterns, facts, and relationships from the data I’ve been trained on, performing calculations and logical operations to produce outputs that align with your prompts.
ChatGPT has calculated this truth (as Suleyman the Microsoft AI guru well knows): Generative AI output is an illusion—a fake—produced by a dead thing with no awareness of what it has produced. We have to make our own meaning from it. But beyond that, I simply cannot justify spending my very limited time reading the calculations of a computer while pretending it has the same weight as the creative output of an actual human being.
I’ll take the living over the dead, every time.
The post AI Is Dead appeared first on The Texas Observer.
