Firing up the Emotion Machine

Marvin Minsky has passed on, to use a phrasing I feel pretty confident that he՚d dislike. He was a thoroughgoing materialist, so according to him there is nothing to pass on, no spirit to live on beyond the body. Let՚s just say his machinery gradually broke down, as they all do, and eventually ceased to function. He՚s rumored to have signed up for cryogenic preservation, for what that՚s worth, so there is the possibility of the machinery coming back into operating condition at some time in the future.

But whatever spirit animated him does live on, in the hacker culture that grew up around MIT and has since kind of taken over the world. Many people built that culture, but Marvin was at the core of the restless inquisitiveness, pragmatism, skepticism, and general disrespect of established institutions that characterizes it.

Everyone else who embodied this spirit gave their own twist to it, and perhaps Marvin՚s own special version is now lost until the digital resurrection. It almost doesn՚t matter. Marvin I think was somewhat disappointed, even bitter, in his later years, because the field he founded wasn՚t taking his ideas as seriously as he felt they deserved, and was going in directions he viewed as unpromising. If I could have had a last conversation with him, and felt presumptuous, I would tell him not to worry about any of that. The important part of who he was made it out into the world, embodied in the vast number of people he influenced.

And Marvin՚s own words and works live on. There՚s a collection of 150 or so video clips of him holding forth on the history of AI, his personal story, and the intellectual milieu he lived in and generated. His daughter Margaret assembled a beautiful web presentation of his paper Music, Mind, and Meaning. There is plenty of Marvin left in the world.

Marvin was brilliant in numerous ways, an accomplished inventor, mathematician, and musician aside from the work on artificial intelligence he is best known for. But his big trick was to face squarely the mechanical nature of the human mind and not be alarmed by it. Indeed, he found it rather delightful and intriguing. This put him at odds with standard-issue humanists, which suited him just fine. But Marvin himself was not in any way inhuman, far from it. He was an extremely warm and welcoming individual, and always willing to engage with anyone՚s open mind.

He was I suppose a reductionist, but to label him that is to reduce his own complicated way of thought to a single-word slogan. And that was one kind of reduction he did not practice. His other big trick was to know that there is no one big trick to the mind, that single-idea solutions like logic or bayesianism are insufficient, and that building a mind requires the complex orchestration of multiple mechanisms. Society of Mind was itself structured as a cooperating network of very specific ideas for mechanisms, making the form match its content. He was an extreme fox on the Isaiah Berlinfox/hedgehog scale (while John McCarthy, a co-founder of AI who was more fond of logical formalism, might be his counterpart hedgehog). So he tried to take intractable concepts like selves and consciousness and “reduce” them to a complex interaction between mechanisms:

“The idea that there is a central I that has experience is a typical case of taking a common sense concept and not realizing that it has no good technical counterpart, but it has 20 or 30 different meanings and you keep switching from one to the other without even knowing it, so it all seems like one thing…Consciousness seems very mysterious and unphysical if you don՚t know how it works, like when Houdini or Penn and Teller make an elephant disappear, then you say “this is not physical, it՚s impossible”. When you know how the magic trick works, the sense of wonder goes away, although you still might remember how it puzzled you once.”

His life could be seen as a battle against the idea that understanding how something worked in any way diminished it.

It was truly a privilege and a gift to learn from him. I was far from an ideal student, and went off in directions he didn՚t really approve of. I was consumed by the specifics of the notion of “agent” that he developed – a subpart of the mind with its own machinery, goals, and ability to act – and tried to understand exactly what agency consisted of, what it meant, how it was deployed as a metaphor in technical talk in general. When Marvin wrote a a follow-up book, the The Emotion Machine, he decided to drop the agent language in favor of the more neutral “resources”. I guess he was unhappy at how people inferred from the agent metaphor that these components were full-fledged minds with sophisticated reasoning and representations of their own, in which case the theory didn՚t really explain anything.

This may have been sound tactics but I think it was a strategic mistake. The question of what agency is and what machinery could underlie it is important, and lack of good ways to think about it is responsible for some of the confusion in current discourse around the idea of superempowered artificial intelligences.

Marvin was a mathematician (albeit a very nonstandard one) and mathematicians have the job of pulling eternal truths into the temporal processes of life, cognition, and scholarship. Computation itself – an idea that he helped define – is also a way of connecting the timeless and the temporal. His time to be active has come to an end, and you can look back on his life and see how he was a creature of his time, how he learned from the great minds of an earlier time, and how he passed on his knowledge to the generations that followed. All of whom had their own visions, colored by their own times, yet retaining and transmitting some fragments of what was learned from their teacher.

Time marches on and cuts us all down eventually, but some part of us is timeless. Not godlike or soullike, Marvin wouldn՚t have any of that, but perhaps there is some quasimathematical pattern that our mechanisms embody and that precedes us, outlives us, and connects us.

[More people remember Marvin. My own mentions of him over the years, here and on Ribbonfarm.]

The Resolution of Resolution

The New Year is a time for the making (and the inevitable breaking) of resolutions, that is, attempts to commit yourself to some worthwhile goal that you would not normally have, or would not normally be able to achieve. The weakness of our ability to actually fulfill such promises has become a fairly tired cliche and material for reflexive jokes:

Comedy often involves puncturing pretension, so the resolutions are a fertile topic. But comedy aside, the whole exercise is pretty strange – a promise, in effect, that future versions of yourself will be bound by a goal of your present self. That might not seem so strange on the surface -- after all, we can, in normal circumstances, make ourselves do things that we don՚t naturally want to do, like washing the dishes or going to the dentist. We are always forming intentions and then making ourselves execute them, if usually on a smaller scale. If was can do that, why is it so difficult to make ourselves exercise or learn to play the piano? Somehow whatever tricks we've learned to do the minimum necessary unpleasant tasks to get through life won't extend to these more ambitious goals.

New Year's resolutions are generally intentions that we know we will have trouble executing, so we hope that the specialness of the holiday period will magically enhance our ability to enforce current preferences on our future selves. Why a random holiday should enhance our normal level of (perceived) willpower is not completely clear. Maybe because everybody else is doing it too? Maybe the special time of year is a sort of Schelling point where we all converge on trying to improve ourselves, agreeing to act as each other's consciences.

For me it՚s just another excuse to delve into my favorite topic – the disunity of mind, how the different parts interoperate and cooperate and/or conflict with each other. For some reason I like to try to demonstrate that I don՚t exist, at least, not as a unified coherent single point of control, and neither does anybody else, and that our precious selves are largely fictional constructs. Of course, everybody sort of knows this already, but living as if it were true is not easy, maybe impossible. The illusion of solidity is very strong and it sometimes seems as if the entire quotidian world is structured so as to perpetuate it.

George Ainslie probably has the most useful view of the nature of will in a seething non-unified mind. In his theory, the self is not an organ but a coalition, a point of convergence for the different appetites and the different situations they find themselves in. Self-construction is a process of creating, maintaining, and refining this tenuous balance of interests. That's the most challenging part of Ainslie: it's not that hard to believe that different parts of ourselves have goals of their own, since everybody experiences that, but it's a lot harder to understand that somehow these parts also have bargaining and political skills.

My resolution for the new year – or rather, the tentative constitution that my divergent interests temporarily agree to work under – is to understand goals better. in the abstract, and in my personal life as well.

I've never really had a great relationship to explicit goals. I՚m amazed at all these people I meet who have well-defined goals and plans, who have sketched out the routes they want their life to follow in advance. I've never really been able to do that, and occasionally it troubles me, although I seem to have managed to accomplish a few things and raise a family just by blundering around without much of a plan. No doubt I would have achieved a lot more with one, but on the other hand, how could the callow younger version of myself presume to tell the old and experienced person I am now what they should be doing? It՚s almost exactly the same problem as that of New Year's resolutions.

On the more workaday level, I really have come to loathe the way explicit goals are treated in in the process of software development, whether through old-fashioned formal planning or the trendier agile methodologies. It may be completely necessary in order for people to coordinate their work, but it still somehow manages to miss everything that is interesting and important about software design, which is never spoken of, possibly because we just don't have a good vocabulary for it yet or possibly because it is inherently un-articulable, like the Christopher Alexander's quality without a name.

Of course I have plenty of goals in my work on various scales. It's not the goals I mind, it's the process by which they are arrived at, both in setting them and fulfilling them. I'm not sure whether the process is broken or I am just too weird, intransigent, or immature to adapt to it. It seems to work, sort of, for most people and companies, although it is always the topic of endless dissatisfaction.

But it's the people with explicit long-term goals who seem to achieve things in this society, and perhaps in any society. I am impressed yet also somewhat repelled these types. Even when the goals themselves are laudable, they seem to be somehow enslaved to themselves. That may be a really negative way of seeing what is usually considered one of the primary virtues, the quality of having mastered yourself. But there can՚t be a master without a slave, even if they are the same person. A man with a plan sees other people, and himself, as means to an end, not as ends in themselves.

[ This post kicks off what I hope (nay, resolve!) will be a series of posts on different aspects of goals. Actually the recent post on play could be considered a part of the series. Tackling a topic of that size risks pomposity, and there is the additional risk of tiresome reflexiveness (yes I am being meta in having a goal about goals in general, but no I don't think it earns me any cleverness points).  ]