I contributed some technical opinions to an article titled “Researchers Demonstrate an AI Language Generalization Breakthrough” on the Pure AI web site. See https://pureai.com/articles/2023/11/06/systematic-generalization.aspx.

The Pure AI article summarizes a research article titled “Human-Like Systematic Generalization Through a Meta-Learning Neural Network,” by Brendan Lake and Marco Baroni.

The topic is a bit tricky to explain. “Systematic generalization” in humans is the ability to more or less automatically use newly acquired words in new scenarios. For example, humans who learn the meaning of a new (to them) word such as “gaslighting” (psychological manipulation that instills self-doubt in a person), can use the new word in many contexts. AI researchers have debated for decades whether neural network systems can model human cognition if the systems cannot demonstrate systematic generalization.

The Pure AI article explains an interesting experiment in the source research paper. Take a look at this image:

In the image, there are four example primitives: tav = blue, yig = olive, muk = red and poz = green. There are 10 example functions. For example, “muk frabet yig” = red olive red. Based just on this information, a human should be able to determine the output for “tav frabet poz.”

If frabet is interpreted as a function, and muk and yig are interpreted as arguments, the example function can be thought of as frabet(muk, yig) = muk-color yig-color muk-color = red olive red. Put another way, the frabet function accepts two arguments and emits three colors, where the first argument color is in position 1 and 3, and the second argument color is in position 2. Therefore, “tav frabet poz” = tav poz tav = blue green blue.

Typically, humans are quite good at problems like this but AI systems cannot generalize as well as humans. But the research paper descibes how an AI system trained specifically for this type of problem can learn systematic generalization.

I am quoted in the Pure AI article. “The question of systematic compositional generalization has been a key point of debate for decades. The Lake and Baroni paper, while not absolutely definitive, strongly suggests that the current generation of large language model systems can be significantly improved in a relatively straightforward way.”

I always admired author Michael Crichton (1942-2008). He had the ability to apply systematic generalization to a wide range of plot ideas. He’s best known for the techno-thriller novel “Jurassic Park” (1990) but he wrote many different types of stories. Here are three of my favorite movie adaptations of his books.

Right: “Rising Sun” (1993) is a murder mystery but also explores the vast cultural differences between the U.S. and Japan. A key element of the movie was the idea that video and photographic evidence can be manipulated — a very novel idea at the time. I give this movie a B grade.