So, I asked the ChatGPT what is wrong with Donald J. Trumps decisions. AI is scary.

[highplainsdem]

15. This is a pretty clear explanation of how LLMs work, from a university in Australia:

https://teche.mq.edu.au/2023/02/why-does-chatgpt-generate-fake-references/

It is important to outline that ChatGPT is a ‘large language model’ designed to output human-like text based on the context of the user’s prompt. It uses a statistical model to guess, based on probability, the next word, sentence and paragraph to match the context provided by the user. The size of the source data for the language model is such that ‘compression’ was necessary and this resulted in a loss of fidelity in the final statistical model. This means that even if truthful statements were present in the original data, the ‘lossiness’ in the model produces a ‘fuzziness’ that results in the model instead producing the most ‘plausible’ statement. In short, the model has no ability to evaluate if the output it is producing equates to a truthful statement or not.

Anyway, this is why an AI-generated list of books to read this summer included books that don't exist: https://www.democraticunderground.com/10143462646

Another recent thread on how stupid chatbots are: https://www.democraticunderground.com/100220313020