Symbol Grounding and Proportional Analogy

If symbols must be grounded in perception, how does this grounding happen? How do we learn to create mappings between language and perception? For example, how does the word “rabbit” get tied to the perception (visual, tactile, whatever) of a rabbit? AI algorithms for assigning textual labels to photographs are not yet able to approach human performance on this task. The problem is somewhat similar to statistical machine translation, which exploits parallel corpora to learn mappings between two different languages, although the difference between text and photographs is more extreme than the difference between any two written languages. Perhaps ideas from statistical machine translation are applicable to symbol grounding. The translation algorithm of Lepage and Denoual, based on proportional analogy, seems particularly appropriate, since it makes minimal assumptions about the structures of the languages.

Meditation, Language, and Evolution

There are many kinds of meditation, but a common theme in meditation is control of attention or awareness. In particular, several meditation exercises involve focusing attention on the immediate present, which seems to involve stopping or altering the internal monologue that usually fills our consciousness. It seems to me that this internal monologue, this constant flow of language, is the main thing that distinguishes us from our nearest living relatives, the chimps. Some types of meditation, in stopping the internal monologue, may be altering our consciousness in a way that brings us closer to the consciousness of chimps. (This hypothesis is not intended to denigrate meditation.) I’d like to explore this idea and see where it leads.

Logical Atoms

In predicate logic, the concept red ball is represented as a combination of the concepts of red and ball. We can define the predicate RedBall(x) as (Red(x) & Ball(x)). Logical atomism views the world in terms of compound predicates, such as RedBall(x), that are built up from atomic predicates, such as Red(x) and Ball(x). Good old-fashioned AI (GOFAI) research almost always assumes a kind of logical atomism. Cyc, for example, represents knowledge using a form of logical atomism. Even those researchers who reject GOFAI still tend to assume logical atomism. Statistical and connectionist models of concepts typically view red ball as a combination of red and ball. I believe that we should turn this view on its head. That is, red ball comes first (is more basic, more primitive); red and ball come later (are more complex, more refined).

Math and Art: Differences and Similarities

Mariana Soffer has made a list of some differences between math and art. In a contrarian mood, I will go through the points in this list and discuss the similarities between math and art.

Ada Lovelace Day

Alfred Ayache brought to my attention that today is Ada Lovelace Day, and that “Bloggers are asked to post about women they know and admire in technology.” I’ll list a few and you can add a few in the comments.

• Marti Hearst
• Jan Wiebe
• Claire Cardie
• Ellen Riloff
• Karen Spärck Jones
• Kathy McKeown
• …

Beyond Proportional Analogy

For some time now, I’ve been experimenting with algorithms for solving proportional analogies. A proportional analogy has the form A:B::C:D, meaning “A is to B as C is to D“. For example, quart:volume::mile:distance means “quart is to volume as mile is to distance”. Multiple-choice proportional analogy problems were part of the SAT college entrance test until 2005, and they are still part of the GRE and MAT tests. It seems that proportional analogies capture an important aspect of cognition, but it also seems that they are a bit simplistic, when compared to the kinds of analogies that we use in the real world. For example, consider the analogy between the solar system and the Rutherford-Bohr model of the atom. We could express the analogy as sun:planet::nucleus:electron, but this proportional analogy is overly simplified. Recently I was able to extend my previous work on proportional analogies to include more complex analogies.

Edgy Stuff

I found an old list I had made in 2000, my top ten favourites in fiction, nonfiction, music, cinema, and art. I was thinking of posting it on my blog, but I decided I wasn’t really happy with the list. I suppose that’s a good thing, since it means I’m not static, I’ve changed since 2000. The problem with the list is not that I no longer like the things in it, but rather that they seem a bit stale. They’re not edgy to me anymore. So here’s a new list, one item per category, of post-2000 edgy favourites.

Structural Realism

Günther Greindl pointed out that my blog post on Attributes and Relations is similar to a philosophical position known as Structural Realism. I read the Stanford Encyclopedia of Philosophy entry on Structural Realism and, indeed, I would have to say that I am a structural realist. Thinking about structural realism, I saw that there is a common theme in my views on science, analogy-making, functionalism, qualia, semantics, and ethics: it’s all about relations, not attributes. Attributes supervene on relations.

SVD, Variance, and Sparsity

There is a steady trickle of visitors to my post on Why Does SVD Improve Similarity Measurement?, so I gave this question a bit more thought. In that post, I offered three hypotheses about why SVD helps — high-order co-occurrence, latent meaning, and noise reduction — and I said that I didn’t know which hypothesis was correct; perhaps they are all really saying the same thing. On further reflection, it seems to me that all three of these hypotheses imply that, in the limit, as the corpus size approaches infinity, SVD would not be necessary. That is, SVD is a way of coping with small corpora. Of course, a corpus is like a hard drive: no matter how big it is, you will eventually discover that it’s too small. But I think there are (at least) two aspects to smallness, and we may be able to separate them, to see which is most important.

Language, Cognition, and Evolution: Modularity versus Unity

When I first started reading about evolutionary psychology, I was excited by the insights into the human mind. The Adapted Mind gave me a new way of thinking about the mind. These insights were reinforced by my research on the Baldwin effect. As an AI researcher, I was eager to apply these new (to me) ideas to my own research. Arguably, the most relevant lesson of evolutionary psychology for the AI researcher is the modularity of mind. I began thinking of the mind in terms of modules (vertical, domain-specific modules, as opposed to horizontal, general-purpose modules), and I thought about how I could implement some of these modules in software. However, after a few years of trying to push this idea forward, with little success, I began to doubt the modularity of the mind. I now believe that the mind has much more unity than most evolutionary psychologists suppose.