Some weeks ago I heard an ad on the radio where the host was lamenting how some of the questions that toddlers ask are hard to answer — particularly those dreaded whys. Why does this do that, why is the sky blue, why do cars move, so on and so forth. The solution put forth by this ad was a toy powered by a large language model that would talk with your kid and answer those questions for you. Don’t worry about knowing the answer to the question: the machine will answer your kid for you.
There’s something so very haunting about the idea to me. We know that these large language models that fuel generative artificial intelligence are prone to wild inaccuracies, we know that the machine’s guardrails can be circumvented, we know these things are riddled with concerns. And this is the tech being advertised for toddlers? We want them to play with a toy that might tell them how to make napalm?
But myriad safety issues aside, why is this something we would even want a machine to do? Curiosity is a wonderful thing and part of the joy of getting an answer is the process of learning something. That practice of cracking open Wikipedia is learning how to be curious as much as it is learning the answers. To say nothing of it being a form of socialization. Finding an answer together is fun, even as adults — how calloused do you have to be to outsource quality time with your child to a glorified autofill algorithm?
That’s all these AI chatbots are: algorithmic text. There’s nothing behind the words it generates, nothing beyond the illusion of meaning. It takes advantage of our pareidolia as we seek meaning. It is the Heider-Simmel experiment in action. But it is a tale told by an idiot, full of sound and fury, signifying nothing. We, as a species, are predisposed to find meaning in madness — isn’t that what language and the written word are? So when we see the nonsense spouted an LLM we reflexively try to understand it, look for reasoning, assume that there is something more to it. But there is none. There is no mechanical turk, there is no ghost in the machine; it is a phantom monkey at a typewriter getting things right by probability. There is nothing behind the words but the meaning we ascribe to it. But why waste time trying to find meaning where there is none?
I’ve said before how I write this blog because I enjoy the process. Similarly, I respond to texts and talk to friends because I, well, like them. I don’t want to spend my time trying to parse generative text when I’d rather parse a friend’s drunken text messages. I don’t want to ask ChatGPT what play I should see in New York when I could ask a friend and strike up a conversation. Life is busy and chances to rest and connect are so fraught in modern adulthood, I want to spend my precious life finding those small joys that make life worth living, not striving against some crude imitation of life. I want to take the time to walk my parents through some tech, answer a kid’s questions, and give friends book recommendations. Life is short; connection is a gift. I don’t want to delegate that some unthinking machine.
Essays, Not Rants! 