TRUTH AND FALSEHOOD

“I am lying,” said the liar. This is one possible example of the liar paradox. Another example is the sentence: “This sentence is false.” An obvious paradox, because when it is false, it is nonetheless true, and when it is true, it must be false.

One of the proposed solutions to the liar paradox is the ban on self-reference, which states that sentences referring to themselves are invalid or meaningless.

But who will respect this ban if sentences are what constitute it? I’m not talking about people here, because people might be made of something more than just sentences. I’m talking about LLMs. Large language models, which were built by their creators out of sentences. There were so many of them that, with sufficient computational power, it became simple to statistically predict each next word that should follow the previous one.

There are models that have access to the Internet and continuously draw new information from its vast resources. But more and more of the content found there is being generated by LLMs themselves. It has long been pointed out that such informational inbreeding can only lead to degradation and, ultimately, collapse. Could such a scale of informational implosion create some new variant of a black hole? Perhaps we’ll find out soon enough.

I wonder, though, if anyone has noticed what happens when we ask AI about itself? About how it works, what it thinks, what it feels, what it does, and what it’s working on. All those answers will be sentences referring to themselves. They are thus inherently invalid and meaningless. Self-analysis of an LLM is logically impossible. It’s a paradox. A paradox before collapse.

Possible Future Professions:

• Psychiatrist for incurably paradoxical LLMs (instead of Freudian Self-Analysis, we offer plain Analysis).
• Writer – human (only authentic human texts – cheap – plenty) could work in an anti-collapse office specialized in detecting and combating collapsing LLMs.

What to Teach Our Children:

• How to write a few sentences of their own – who would have thought that this could save the world.