Artificial Intelligence is not Intelligence

A few thoughts on artificial intelligence

For an AI chatbot like ChatGPT to work, it first needs to train the Large Language Model (LLM) it relies on using a huge amount of text data. Modern LLMs are trained on almost everything that can be found on the internet.

The texts generated by chatbots are often so similar to human writing that they are described as artificial intelligence. Some people even trust them more than their relatives or friends.

The heated debate over whether such texts can be considered a manifestation of intelligence ultimately comes down to how we define “intelligence” itself—what it actually means.

However, it is relatively easy to show that artificial intelligence is not intelligence in the human sense: if chatbot language models are trained on texts they generate themselves, they eventually begin to hallucinate uncontrollably.

There is a certain elusive quality present in human texts that is absent from machine-generated ones.