Marvin Minsky Reflects on a Life in AI [video] | Hacker News Reader

Marvin Minsky Reflects on a Life in AI [video]

technologyreview.com

56 pointsadenadel16 comments

Back in the 80s he wrote a book called "The Society of Mind" which I think is an amazing book. The core theory, which really resonated with me and stuck with me, is that we do not consist of a single mind. Our consciousness is a set of smaller subsegments (he calls them agents) that work (generally) together.

If you think about the fact that looking at our brain you can see the evolutionary differences compared with other organism,s this starts to make a lot of sense. We have these different systems that have varying ages, with natural selection putting more pressure to develop a new system or enhance an old one, and the brain winding up almost like a tree trunk with different rings. It makes sense these smaller chunks do their own thing then "talk" to other parts. we even have language referencing it, "I'm of two minds about XYZ", and diseases like schizophrenia and hallucinations we trace to various parts of the brain not recognizing each other or not working properly together. We have studies that seem to indicate some decisions being made before we consciously think about them.

About a decade ago I really started to feel like I've begun to recognize different "aspects" and understand why I feel the way I do by understanding this theory. An example is, if I have a big decision, and I find that I'm undecided and really unhappy while trying to come to a conclusion, it's usually because I've actually already decided, and I'm unhappy because I'm arguing with myself, I'm not accepting that decision. Rarely have I been unhappy with the decision I'd already made, too.

He's a brilliant guy with some amazing work going back decades.

I'm not familiar with Minsky's book, but the core idea seems to have a lot in common with the modular theory of mind [1], which was explored in a MOOC that I found fascinating, and which is now available in a self-paced form [2].

[1] - https://en.wikipedia.org/wiki/Modularity_of_mind

[2] - https://www.coursera.org/learn/science-of-meditation

Haven't read The Society of Mind, but I enjoyed Incognito, which explores similar ideas: http://www.amazon.com/Incognito-Secret-Lives-David-Eagleman/...

And from a therapeutic perspective, I've gotten a lot out of Voice Dialogue - http://www.voicedialogueinternational.com/

Their book is geared toward practitioners but still quite good. I don't think you can do Voice Dialogue on your own. It helps to have a trained therapist.

dunno know if it's necessary to go all the way back to the 60's to see interesting approaches to cognitive science.. hofstadter's 1995 book is a good read

https://en.wikipedia.org/wiki/Fluid_Concepts_and_Creative_An...

I'm in the middle of reading that book right now. So far, I'd say that I highly recommend it. I'm actually planning to do some work trying to implement a program for the Le Compte Est Bon game.

I've also been reading Surfaces and Essences, also by Hofstadter, and recommend it as well.

"Let's get another group of beginners in to see what they can do... and fire the experts."

What a thing to say.

When trying to bring about a big change, his point seems to be that sometimes new simple thinking is more important than old complicated thinking. That's very much in line with the MIT Media Lab and their way of looking at the world.

It's the right thing to say though, especially for technological progress. Old ideas and old people last to long in this business, they need to be overthrown more quickly. Punctuated equilibrium is directly due to old hierarchy handling the helms for too long. Out with the old (including myself) and in with the new.

> Old ideas and old people last to long in this business, they need to be overthrown more quickly.

I'm not sure how literally you mean this, but have you considered that old people are still humans, and after a certain point it's a lot more likely that they have a family to support, need more medical care, plus a bunch of other things? You can't just throw people out because they are no longer as fresh and efficient as they were in the beginning. Being old comes with it's own advantages. There's a place for everyone in the business.

I am one of the most humane persons you will probably never meet. I am in total support of taking care of older people, I just think they hang around too long and impede technical progress. I would posit there is a strong correlation between youth of faculty and innovation of a given university in any research area.

There is value in the old and they young, and ideally you would have both on a team or faculty. You need youth for innovation and boldness, and you need age for experience and wisdom.

He did say this was a gloomy suggestion..

Minsky certainly knows what the experts can do: he and Seymour Papert more than decimated research in neural networks by writing "Perceptrons". https://en.wikipedia.org/wiki/Perceptrons_(book)

He remains critical of neuroscience:

http://luttrellica.blogspot.com/2007/01/marvin-minsky-bashes...

I've sometimes thought that AI would be solved only once Minsky passed from the scene.

To paraphrase MM: Let's fire all the experts in NN and make room for someone naive enough to reinvent GOFAI.

Great to hear I'm not the only one. I can't make up my mind regarding the sign (plus or minus) of his contribution to the field.

I have heard him speak a few times, very engaging.

I have one minor disagreement with something he said about in the past they would make some big improvement every few days, now it is every few years.

With all due respect, I think that he has it backwards. I have to some degree been working in AI and machine learning since the 1980s. I am blown away at how fast new good results are achieved.

Reply on news.ycombinator.com