Hallucinations or Lies

My hallucinations by August Natterer

Since ChatGPT appeared in late 2022, I have been warning that the answers provided by Large Language Models (LLM; I refuse to call these tools Artificial Intelligence) are unreliable and should be treated with the utmost caution. Often, these answers seem plausible and are written linguistically correctly, but they are false. These types of answers have been called hallucinations.

This is not surprising. It is a logical consequence of the algorithm used by these programs, which I described in another post in this blog, which I simulated by means of a program with only 18 instructions. The algorithm works by adding words extracted from the most frequent ones that follow the previous words, chosen from billions of files taken from the Internet. It is evident (just think about it) that this algorithm cannot guarantee that the answers these tools provide are true.

What’s the matter with dark matter?

Comparison of observed and expected
rotation curves of galaxy M33
(Wikipedia)

Nine years ago, I published here a post entitled Dark Matter or New Theory, where I pointed out that the motion of stars in galaxies does not match theoretical predictions. The problem can be solved in two ways:

1.      Assuming that galaxies have much more mass than we can see. The mass we cannot see was called dark matter, where the word dark means that we don't know what it is.

2.      Assuming that Newton's laws should be corrected in the outer part of galaxies, where the acceleration of stars is very small. Various corrections to Newton's equations have been proposed that would satisfactorily solve the problem, which are called MOND (MOdified Newtonian Dynamics) theories.

Traitors to the Human Species

The Three Body Problem is a gripping science fiction novel by Liu Cixin, which contains a wealth of information about ancient and modern Chinese history. But I fear it distorts science. And my first golden rule of good science fiction is not to distort science. I think distortions are dangerous because uninformed readers can be led to believe that certain false things are true.

I’m not worried about the assumption that string theory is true. It could be, although it has lost a lot of backing in recent years. But the description in the novel of the three-star system Alpha Centauri has nothing to do with reality, even though that description is crucial to the plot.

Should Barbie chat with children?

ChatGPT and similar tools, called LLMs (Large Language Models), are being used with greater frequency in our daily lives. Google, for instance, has integrated its GEMINI tool with its search engine. Sometimes, when the program behind the search engine deems it appropriate, the question asked is sent to GEMINI, and the response of the LLM appears first, albeit with this warning at the end, in small print:

AI responses may include mistakes.

Of course, they may include mistakes! These responses are not generated by understanding the question, but by using information previously obtained from the Internet, and applying an algorithm based on extracting words that typically appear in that information after the previously generated words. See a post in this blog where I explained that algorithm. Since the information extracted from the Internet can be true or false, and the algorithm can introduce new falsehoods where none existed, the answers obtained may be correct, partially correct, or completely wrong, therefore Google's warning is valid.