RIP Marvin Minsky

AI

Jan 2016

Much of engineering and programming is about solving puzzles, learning and making discoveries. That there are many emotions (like joy, fun, awe, pain, surprise) in thinking were one of the lessons prof. Minsky taught me. And, it was one of the ideas of starting my blog “thinking_on thinking”.

The Society of Mind lecture by Prof. Marvin Minsky at MIT on OpenCourseWare

It was very sad news to hear that Marvin Minsky passed away. In honour to prof. Minsky, I want to share some email exchanges we had:

2003

My first email to Marvin when I was studying semiconductors:

Dear Prof. Minsky,

lately I was thinking that computing systems can never be “intelligent” (I haven’t thought about a definition of intelligence yet, but I mean the function of solving problems autonomously). As far as I understand VLSI systems, they are all finite state machines.

Brain are finite-state machines, too. So far as present-day physics can say.

That kept me thinking and I had to ask more questions. I was wondering what the evolutionary reason was, that humans have a certain number of neuron cells.

For example, an Intel Pentium processor has around 400 milion transistors (90nm gate length). Moore’s Law is predicting several billions transistors in some years. At the moment the biggest bottleneck is seen in the power consumption (=heat production). Do you know if there is a similar problem for the human brain?

That is a very interesting question! In fact, many evolutionary biologists think in terms of power consumption, and regard the brain as so power-hungry that this is an important factor in natural selection; they conclude that the reason why so few animals are intelligent, is that it doesn’t have very much cost-benefit!

Here is an amusing fact: a typical human being consumes about 130 watts of power. All this power comes from pumping protons through the membranes of the mitochondria, against a potential of about 300 millivolts. Therefore we are powered by a current of about 500 amperes!

I am still busy in reading some of your work from your homepage… I think there will many question popup. Just one last remark, I have read your article about music and meaning. I found there were extremely interesting thoughts in it.

Thank you. Many composers have told me that, so far as they know, that is the only good article specifically about theories of why people like music. Isn’t that strange, considering that many people spend such a large fraction of their lives (and incomes) in it?

ps: I have recently read the book of Bertrand Russel “Problems of Philosophy”. Do you have suggestions of other books where perception of things/reality is examined ?

Yes, very few philosophers have as much common sense as Russell did. The only ones I respect today are Aaron Sloman, Daniel Dennett, and John McCarthy. ( Sloman and McCarthy have good Web pages.) It is peculiar how few modern thinkers look so deeply into questions about how the mind really works. But I still find important ideas in much older psychology books, especially Sigmund Freud, William James, Francis Galton, Wilhelm Wundt, and some yet more ancient works by Aristotle and Augustine. Needless to say, very few contemporary students or researchers have read any of these!

2004

Then, I started reading more about symbol processing in our nervous system. Not sure if this message came via a newsgroup.

minsky@media.mit.edu schrieb:

The power of consciousness comes not from ceaseless change of state, but from having enough stability to discern significant changes in your surroundings. To “notice” change requires the ability to resist it, in order to sense what persists through time, but one can do this only by being able to examine and compare descriptions from the recent past. We notice change in spite of change, and not because of it.

2006

Then, in 2006 I reflected about moving to Boston to study computers at the Media Lab. Still, moving from Germany to the US would be quite some jump for me at that time. Here is some resasoning of Marvin.

minsky@media.mit.edu schrieb:

Well, it was not necessary to travel, because I was in the right place at the right time. When I was in college in the late 1940s, some of the greatest mathematicians in the world had already come to Harvard, as refugees from Europe and World War II. Then when I went to Princeton in 1950, I had the opportunity to work with von Neumann and other great mathematicians, as well as physicists like Oppenheimer. I did not get interested in logic until later, and never liked it much because I recognized that most reasoning involves analogies, which no logical systems can handle well.

I also asked questions on music, what I was listening at that time:

I like Scarlatti’s ideas enough, but prefer to be guided by Beethoven and Bach, because they show how to think several thoughts at once.

And about traveling in general:

As for visiting countries, I don’t much like cultures, because each of them only shows different mistakes. To me, only the culture of Science shows progress.

2008

The last email conversation:

Marvin Minsky minsky@media.mit.edu 11/26/08 to me On Nov 12, 2008,

at 3:23 PM, Mulder Patrick wrote:

I just discovered your writings at wiki.laptop.org on how education could put incentives on discoveries for children.

Yes. I’m right now trying to finish Memo 5, which is about whether we can explicitly teach children new ways to think, etc. I think I should copy those memos to my web page.

I recently learnt about values within “teams” and how they prevent on developing individual views:

Interesting issue: does teaching “cooperation” conflict with learning to develop one’s own ideas?

For example:

I couldn’t quickly extract a clear idea from these. Does Jonathan Haidt have a good paper on this?

I did not understand too much how to dig for papers in social sciences at that time. But I liked to asked more questions about books in general. Also about writing.

So I asked: It is not possible to regularly order your first book on computers. Are there plans to re-publish the book? Do you have it as PDF or text?

Yes, I’ll try to republish it. I encountered resistance, the last time I tried, because publishers insisted on something new. But I wrote it so that it would never be outdated!

  • I have your book on Perceptrons. Does topology play a role in modern computers? How can it teach us to make better abstractions?

Well, to begin with, just deciding if there is more than one object in a visual scene need a topological process.
Or counting. Or X is inside Y, etc. Topology abstracts relationships that don’t depend on specific shapes.
In any case, I suspect that it would be a better subject for children than arithmetic is. I have an age 10 granddaughter whose hobby is abstract category theory; I think she’s likely to become a powerful mathematician.

What’s interesting about that Perceptrons book is that, so far as I can see, no current neural-network practitioners understand its ideas at all!
So far as I can see, all of its conclusions still apply to multilayer networks.

From these conversations, I learnt how important it can be to ask questions and how to search for new answers. These are the origins of my blog thinkingonthinking.

Leave me feedback

Follow me on Twitter here.

comments powered by Disqus