Solaris One: Artificial Intelligence from 1960

This series of blog posts began as a set of observations about literary research on the novel A Confederacy of Dunces (Confederacy), by John Kennedy Toole, but I have extended it to include other topics. This blog entry starts a new topic.

My current research interest is the novel Solaris by Stanislaw Lem. I will not yet post my overall thesis for the project, but I am learning things along the way which are tangential to my thesis. These I will share.

In addition to his fiction, Lem published some non-fiction. A major work from the early 1960s, which was revised several times, is Summa Technologiae. I have felt the obligation to read this text to gain insights into Lem's conceptual framework when he wrote Solaris. I have now read (part of) the recent translation by Zylinska which was published by the University of Minnesota in 2013.

Summa Technologiae had almost nothing that furthered my own thesis, but I did notice a number of interesting details. I plan to have several blog entries about points I have discovered in this book.

Today's topic: Lem's understanding of computer science, or as it was called at that time, cybernetics. Lem, writing in the 1960s, seems to have an overly optimistic view of the ability of computer systems to display intelligence. That is, he is too optimistic about what we now call artificial intelligence. He doesn't understand that a major function of computers in the 1970s through today will be to keep track of records in database systems. Computer systems have not needed to be intelligent to accomplish such a role: information management does not require sophisticated computations, just efficient data retrieval.

In the late 1950s and early 1960s, many computer theorists speculated that artificial intelligence would not be so computationally difficult. There is the famous story (possibly an urban legend) that the CIA at that time funded a Russian language translation system. They tested the system by inputting an English phrase for it to translate into Russian. Then they took the output and had it translate that back into English. The proverb, "The spirit is willing, but the flesh is weak," was translated into "The vodka is good, but the meat is bad." (Cue the rim shot.)

The recent advances in AI have only come about after great strides in machine learning and neural network designs. That, and an enormous increase in computing power. Reading what Lem wrote in 1960, I think, well, he was sixty years too early. Only products such as IBM's Watson, Google's AlphaGo, and OpenAI's GPT 4.0 have shown that computers are achieving something that might be called real intelligence.

An interesting sidetrack: the split between language smarts and mathematical smarts. The SAT test is famously divided between a verbal test and a mathematical test. In early 2024, the journal Nature reported on an AI system that is able to solve math competition problems better than many humans. Especially noteworthy is the fact that it is not a single system; it is two systems glued together. The first system is a language parser, which translates the verbal description of the problem into mathematical formulations. The second system then solves the math problem. Even among AI systems, the SAT test is split between verbal and mathematical skills.