I had taken along a (quite early) copy of the GNU Emacs manual. The FSF was selling them, but I'd put this one together myself. Running TeX on the texinfo source, converting the output for the Imagen printer, and then taking it to Kinko's to be spiral bound, including my imitation of the yellow cover that the FSF version had.
I asked Minsky for his autograph. He looked at what I presented, understood what it was, an autographed inside the front cover, "Marvin Minsky, friend of Stallman".
In April of 2011 in an airport in Honolulu, I presented that same manual for an autograph to Richard Stallman. He looked my manual over for a long time. IIRC, it documents Emacs version 16 or 17. Then he signed it, below Minsky's autograph,
"Richard M. Stallman - Friend of Minsky"
While he was in Hawaii, I also paid to fly him to the "Big Island" (Hawaii) to speak with people (mostly astronomers and their children, though some people drove up from Hilo.)
Since I was paying for his ticket, I knew he would be at the airport.
I lived in Hawaii at the time.
Back in Las Vegas (where I lived at the time), the book, and emacs, were new to me. I took it to read then.
"What are you doing?", asked Minsky.
"I am training a randomly wired neural net to play Tic-tac-toe", Sussman replied.
"Why is the net wired randomly?", asked Minsky.
"I do not want it to have any preconceptions of how to play", Sussman said.
Minsky then shut his eyes.
"Why do you close your eyes?" Sussman asked his teacher.
"So that the room will be empty."
At that moment, Sussman was enlightened.
- From his talk On Game Software Development, in 2001 I think (from Technetcast.com)
Just to be clear: I'm not questioning the story, just the details in the recollection :)
I'm sure Hillis knows more physics than I do (though I knew the equations of motion pretty well at one time), but he could easily have just mis-spoke. I didn't pursue this line of thought, but considered it might have been something to do with him deriving an expression for the vertical position in a gravitational field, perhaps in terms of horizontal motion or something.
Yes, but they will be our children.
--Marvin Minsky http://web.media.mit.edu/~minsky/papers/sciam.inherit.html
"Because of the random wiring, it had a sort of fail-safe characteristic. If one of the neurons wasn’t working, it wouldn’t make much of a difference—and, with nearly three hundred tubes and the thousands of connections we had soldered, there would usually be something wrong somewhere. In those days, even a radio set with twenty tubes tended to fail a lot. I don’t think we ever debugged our machine completely, but that didn’t matter. By having this crazy random design, it was almost sure to work, no matter how you built it."
"So Sussman began working on a program. Not long after, this odd-looking bald guy came over. Sussman figured the guy was going to boot him out, but instead the man sat down, asking, “Hey, what are you doing?” Sussman talked over his program with the man, Marvin Minsky. At one point in the discussion, Sussman told Minsky that he was using a certain randomizing technique in his program because he didn’t want the machine to have any preconceived notions. Minsky said, “Well, it has them, it’s just that you don’t know what they are.” It was the most profound thing Gerry Sussman had ever heard. And Minsky continued, telling him that the world is built a certain way, and the most important thing we can do with the world is avoid randomness, and figure out ways by which things can be planned. Wisdom like this has its effect on seventeen-year-old freshmen, and from then on Sussman was hooked.]"
Do TensorFlow/CNN builders use random initial configurations, or custom designed stuctures?
It's related to the No Free Lunch Theorems. It basically says that if an algorithm performs well on a certain class of learning, searching or optimization problems, then it necessarily pays for that with degraded performance on the set of all remaining problems.
In other words, you always need bias to learn meaningfully. More you have (the right kind of) bias, faster you can learn the subject in hand and slower in all other kinds. In neural networks the bias is not just the weights. There is bias in the selection of random distribution of the network weights (uniform, Gaussian etc.) There is bias in the network topology. There is bias in the learning algorithm, activation function, etc.
Convolutional neural networks are good example. They have very strong bias baked into them and it works really well.
For some tasks, you may wish to initialize using a network that was already trained on a different dataset, if you have reason to believe the new training task is similar to the previous task.
But the weights are often initialized to be really close to zero.
Closing his eyes did not make the room empty. It made him not know which things were where.
Randomizing the neural network did not remove all the preconceptions from the network. It made him not know what the network's preconceptions were.
There's an urban legend that I once got into a fistfight with Marvin Minsky, which does about as well as anything to illustrate the crazy, crazy things that people have been known to believe about me.
We have temporarily misplaced a great mind. See you later, Professor Minsky.
This kind of statement gives me great hope, and in particular represents the kind of fundamental mindset change that helps counter many of the painful aphorisms commonly pulled out when someone dies. I find it deeply unfortunate how rarely it applies, but as mentioned elsewhere in the thread, it applies here. Thank you.
But, as someone who plans on being cryo-preserved eventually, I'd say that whatever chance there is of reviving whatever remains of this individual, it should be taken. I'd want to live in the future, even if that meant not having my full cognitive abilities. Maybe not as very cognitively-impaired individual, but I guess I'll put that type of stipulation in the contract if I was worried about it.
Everything eventually ends, I don't see the appeal in pushing that only to suffer.
It is possible to lose an entire brain hemisphere and retain complete functionality
There is a huge amount of redundancy in the brain, and most brain matter is only concerned with I/O, signal processing, and life support. It's one of the reasons I have a lot of hope that cryonics is feasible -- massive loss of brain tissue need not mean irreversible loss of an individual.
Brain injury kept Roy Walford from being cryopreserved, though there it was clearly an extension of his own thoughts on the matter: http://www.cryonet.org/cgi-bin/dsp.cgi?msg=24045 I see that as a terrible shame; it is guessing in advance as to the limits of what can be restored.
I mean apart from the technical expertise on freezing etc, the rest is mostly wishful thinking about how such a facility will survive long enough in tact (and the US wont itself go the way of the Assyrian, the Persian, the Roman, the British and other empires, in 1 or 2 centuries time), technical/medical resurrection will be possible, and future people will care to resurrect those in there.
Even if a great current mind was preserved there, if people of the future are, say 2x brighter than us (not to mention having access to advanced AI) it would make little sense want to resurrect them for that alone. And as for having access to 20th-21st century info, with our trillions of bytes of video, images, texts and sound recorded every day, they'll likely want LESS, not more information about our times.
EDIT: I find it ironic - as he was my first exposure to the science of mind and thought etc... I find it ironic that his own death was due to a failure of the brain/mind in some way given how much he has contributed to the idea of thought and mind in his career.
This man and I started talking about intelligence, ML vs. symbolic, and more... he truly knew many intricacies of AI! Eventually, for some amazing reason, out of nowhere he asked me if I wanted to come to Marvin's house that evening! Of course I said yes! At the time, the only paper I had on me was ironically Patrick Winston's thesis printed out in my backpack, so this man wrote the name "Henry Lieberman" (a colleague of Minsky's) on the cover and gave me Minsky's address!
I went to Marvin's house that evening, and it was simply wonderful! We talked about SoM, and I was included in these discussions and was treated like a colleague. Marvin answered all my initial questions, but only created more within me! He engaged me! I really felt included. It was one of luckiest days in my life.
I'm sharing because I'm reading other stories about people's encounters with Marvin, and while I was reading them I didn't feel as sad. Perhaps mine might do the same for someone somewhere.
I asked him if it was true, and he said that it wasn't true that he thought the problem was easy, but it is true that he had a first year undergrad student that he decided to put in charge of his grad students working on the problem. The first year student was Gerald Sussman.
Of interest: https://en.wikipedia.org/wiki/Neats_vs._scruffies
I find it interesting because Minsky did a lot of the foundational work in Neural Network research yet he philosophically identified as the opposite on the Neat/Scruffy spectrum of most NN researchers today. Much like Bayes, I think there is some immense wisdom from his research that will not even be acknowledged as wisdom for decades.
EDIT: You can get a little intro to his thoughts on the matter starting about 27:16 in this video  (linked at the time marker). If you watch for about 10 minutes, he demonstrates some of the difficulties of using single abstractions for something as complex as human intelligence.
On top of his work, Minsky taught me that you should meet your idols. If they're worth it you walk away enriched and invigorated.
Not to mention that he was a brilliant mind, and his loss is a loss for humanity at large.
Anyway, RIP Mr. Minsky.
Why Programming is a Good Medium for Expressing Poorly Understood and Sloppily-Formulated Ideas
P.S. Web of Stories has an extensive, autobiography style interview with Marvin Minsky .
No computer has ever been designed that is ever aware of what it's doing; but most of the time, we aren't either.
In general we are least aware of what our minds do best.
And, come to think of it, a variation was in the novel Rainbows End by Vernor Vinge.
From SICP preface, which is the quote that actually matters, this is inspired by Minsky's quote. I hate that the quote that went public were all the other less inspired quotes from SICP.
Jokes and their Relation to the Cognitive Unconscious
Marvin Minsky, MIT
Abstract: Freud's theory of jokes explains how they overcome the
mental "censors" that make it hard for us to think "forbidden"
thoughts. But his theory did not work so well for humorous nonsense
as for other comical subjects. In this essay I argue that the
different forms of humor can be seen as much more similar, once we
recognize the importance of knowledge about knowledge and,
particularly, aspects of thinking concerned with recognizing and
suppressing bugs -- ineffective or destructive thought processes.
When seen in this light, much humor that at first seems pointless, or
mysterious, becomes more understandable.
1) We don't always know who is one of "the greats" of today's generation, until much later.
2) Today, content generation is so democratized (eg wikipedia, soundcloud) that there are less individual superstars like Pushkin, etc.
3) Even intellectuals who were once highly regarded would today have tons of comments on their blog nitpicking and debating every detail of what they said, with many of the comments being low quality. See eg Sam Harris vs Noam Chomsky debates.
It seems like a world in which it's harder to be great like Alan Turing or Claude Shannon or Richard Feynman or Marvin Minsky. At the same time, maybe there are more great people than ever, and we just don't always know who they are until much later.
"Communication with Alien Intelligence" is another favorite of mine. The idea of enumerating all possible Turing Machines and looking for ones that do something meaningful is brilliant.
Chapter 3 talks about McCulloch-Pitts.
I remember being impressed with chapter 14 ("very simple bases for computability") as a kid. Finding UTMs with minimal number of states etc. are great riddles. I also fondly remember the discussion of the halting problem and related problems ("does program P output X") in chapter 8. This was my first introduction to this procedure and Minsky made the idea of reducing one problem to another totally straightforward. Many years later I realized that not a few CS students find these ideas confusing.
I went to Patrick Winston, and asked him if it was worth going to Marvin's lectures given that I keep falling asleep. He said - of course, we all know you are overworked, but marvin may say something that will change your life.
Not that I knew anything about him... it was just an impression.
Rest in peace, Mr. Minsky.
Though not as strong and fast-talking as he once was, Marvin's humor and wisdom shine through.
Here's a link to the video:
Danny begins his introduction at about 41:29.
> True story: in my final essay of my AI module at uni, I disagreed with everything Minsky believed, so my professor, a fan, flunked me.
Machine Learning is applied science. Artificial Intelligence is science.
It'll be fun.
Have you chance to elaborate why this guy deserved this? Did he build first neural network, lisp machine, ALICE chatbot, break image net. I always considered him as some kind of celebrity from science, while other guys, which names nobody remembers, actually pushed AI movement by doing real things, while working on Google Brain, Watson, cyc, trying to catch spam, terrorists and fraudsters.
You won't see the top menu links.
Seems like the admins have pretty much done away with the black bar thing.
When I see a black bar on Hacker News, it serves as an alert to me to scan the news more carefully. Although I knew who McCarthy was, Jobs, etc, I didn't know this person. The black bar alerted me that someone in this community that is considered quite an influence died. Now I'm reading all about the person in various articles. The black bar helped to alert me to that.
In one of the lectures, Minsky told a story about how when he was at Princeton, Oppenheimer invited him to lunch. When Oppenheimer brought Minsky to the lunch, there were two other people there - Gödel and Einstein. Talk about brainpower, that must have been an interesting table conversation.
It would be easily dismissed if said by someone much less accomplished. In this case, I just take it as a frank statement of belief.
Regardless, I think it's safe to say Minsky was a great intellect, of a caliber not often seen.
I would have preferred Asimov to say something like "Sagan and minsky are the brightest minds I've encountered in my life." And if you think Asimov is super smart, then you'll respect Sagan and Minsky that much more. And you'll respect Asimov too for not being so egotistical. It doesn't even matter if it's actually true or not.
Minsky, in my experience, demonstrated humility when talking with or listening to him. For that I gain admiration beyond the talent.
But that wouldn't make good banter.
...He also said that he thought McCarthy was an idiot. :)
But Marvin Minsky? My Tuesday wasn't ready for this. He has had such an impact on the field of AI, and even on the social dialogues about it. Not everybody thinks that robots are going to go Skynet on us, and a lot of us that realize that were informed by his work. Whether directly or indirectly, so much of his work has become common knowledge amongst AI enthusiasts and scientists.
I'd be wasting my breath to say that he'll be missed, of course. I wish I could have met him.
Hard to imagine someone more black bar worthy for Hacker News, hope we have one up soon.
Edit: allow me to clarify - just having to choose whether or not someone is "black bar worthy" is distasteful. Ian Murdock didn't get a black bar that I can recall, and I don't remember seeing one since then, even though other people relevant to the community have died. Are we to expect someone to change the color of the bar every time a death in the tech community occurs, and how to we judge relevance in that regard?
No, it's an arbitrary gesture that doesn't really honor anything or anyone. It just gives people something to argue about. Why did someone get a black bar and someone else didn't? Why was this person deserving of it, and not that person? It's best to just remove it altogether.
In which case, it's a personal decision to honor someone they care about by placing a subtle notice on ther site, which happens to be popular. It's no different in spirit to the thousands of tribute blog posts being written as I type this.
I find it extremely distasteful to criticize how someone chooses to honor the dead. As long as they're not doing it by shooting guns into the air or hosting a destructive party next to your house or something like that, what do you care?
It certainly is within the site owners' right to do whatever they like, but for the community it's becoming a spectacle.
Although obviously, as the bar is up right now, my opinion on the matter isn't going to prevail.
Civility is important but it should not be confused for everyone having the same point of view on a topic. And censorship by "civility" causes people to not join discussions. When I come to HN I want a intelligent but lively discussion. However, some issues have reached such a consensus such that even well thought out opposing views put politely get heavily downvoted.
Imagine you were the caretaker of a public establishment like a school or business, and news broke of the passing of an eminent person who was deeply respected and admired by many of the people who frequent that establishment.
So you went and lowered the flag to half-mast, because that is the customary and respectful thing to do. And whilst most people appreciated the gesture and felt comforted by the shared sense of mourning and respect for the deceased person, a small minority erupted into a noisy debate about how appropriate it was to lower the flag, and whether someone else was more worthy of having the flag lowered in their honour, etc.
If you can imagine this scenario in real life, you can understand how dang feels when this kind of argument erupts on a bereavement thread on a site he runs and cares so deeply about cultivating as a pleasant site to visit.
He can't be the one to call people out for being insensitive, but he can at least say "Thank you" to someone who does, and who in doing so, gives him some much-needed reassurance about the level of emotional intelligence around this place.
Discussions about the merits of customs and policies on the site are fair enough, but if we're to be as humane and compassionate online as we would try to be offline, the time and place of the mourning and honouring of a just-deceased person is not the right time and place.
Just reverse that situation and imagine that you are a member of a public establishment and that when certain people you particularly respect pass that public establishment does not follow its usual customs.
That isn't what's happened here.
Seriously, civil behaviour around a bereavement is just not that hard.
Respectfully, I won't be commenting further on this thread.
You don't have to have the same point of view on this topic, just don't take a discussion of a recently dead person being officially mourned by the site owner as an opportunity to criticize that.
If this "censorship" (which is far from it) causes people who are going to argue over mourning to not join the discussion, then mission accomplished.
Meta-discussion of the black bar is intellectually uninteresting at best. Off topic in the middle. And disrespectful of people's grief at worst. The least of these is reason enough to downvote.
If I squint, I guess I can see that use of the word "civilized" as a bit aggressive, but really I think Mike was being hyperbolic for effect in his later description.
(It's probably not comprehensive, but it's clearly composed of computing luminaries and members of the YC community)
Dan Haubert (1984 - 2009)
Robert Morris (1932 - 2011)
John McCarthy (1927 - 2011)
Dennis Ritchie (1941 - 2011)
Steve Jobs (1955 - 2011)
Doug Engelbart (1925 - 2013)
Aaron Swartz (1986 - 2013)
Gene Amdahl (1922 - 2015)
Marvin Minsky (1927 - 2016)