Disciples, Apostles, and Saints!
There is a word that people have been using in tech news around large language models (LLM) like Claude and ChatGPT. It is a word to describe these supposedly AI systems when they make up an answer. They call it “hallucinating.” I’m never going to use that word.
Here is the thing about all of the AI hype and what the Pope has rightly argued in his encyclical on the subject: these are artificial creations that don’t think or have consciousness. In that way, they aren’t real.
I grew up watching Data on Star Trek: The Next Generation. I am the perfect target for sympathy and a nerd for philosophy. So I love asking questions about human consciousness and what it would take for artificial intelligence to be considered real intelligence. Like another sci-fi icon: I want to believe.
This isn’t that, however. Because these are systems, not consciousness. They aren’t human. Which means they aren’t doing human things, like hallucinating. They are, on the other hand, built to produce inaccuracies in the data as if it were a mistake or an expression of creativity.
There is a desire to humanize these machine systems, but it isn’t fair to the LLMs or to humans. And this is where so much of the confusion lies: that it isn’t only about the cheating in school or the ecological impact or the ease of use or the systems that get streamlined. It is the desire to pretend this stuff is what it means to be human, and as a result, degrade what it means to be human. That’s why terms like “hallucinate” or “think” aren’t innocent. They are setting humans up to fail at being human against the greater computational power of LLMs.
We should be careful not to too humanize what isn’t human.
With love,
Drew+
