The computing world mourned the loss of a legend last week when AI visionary and inventor Marvin Minsky passed away. Many technical communities have paid their respects over the last week, from AI researchers and roboticists to hackers and supporters of open source methodology.
Minsky was highly educated, like many in the field, holding a BA in mathematics from Harvard and earning his PhD from Princeton only four years later in 1954 with his thesis Theory of Neural-Analog Reinforcement Systems and Its Application to the Brain Model Problem. In 1958 he joined the MIT faculty, which he was a part of for his entire life, until his death this year. In 1959 he founded the highly influential MIT Computer Science and Artificial Intelligence Lab with computer scientist John McCarthy, becoming indirectly (as well as directly) responsible for much of the early innovation in the field of computer science.
In Pursuit Of Artificial Intelligence
It was while he was still a student, in 1951, that Minsky started to put his brilliant and creative mind to work, building the world's first randomly wired neural network machine. It was called SNARC, or Stochastic Neural Analog Reinforcement Calculator, and has been pointed to as the first ever artificial self-learning machine. The idea of computers and machines functioning in the same manner as their biological counterparts was a concept that Minsky returned to multiple times over his life, driving much of his research into artificial intelligence -- he was even known for referring to humans as 'meat machines'.
Minsky helped Seymour Papert build the first Logo "turtle" robot -- a robot built low to the ground with a roughly hemispheric shell, hence its namesake. This kind of robot is still a common project for robotics students and enthusiasts, and has even led to the development of technology like iRobot's Roomba. Much of Minsky's work on artificial intelligence was theoretical, however, publishing multiple books including the controversial Perceptrons, another collaboration with Seymour Papert. Its pessimistic predictions have even been credited with contributing to the so-called AI winter, when funding was cut from studies of neural networks.
In the 70s, Minsky developed a theory called the Society of Mind, publishing a book on the subject that boded a little better for the field of artificial intelligence than his earlier work had. The Society Of The Mind was written for a general audience, unlike many of his earlier academic writings, and covered a theory of AI whereby intelligence could be the product of an interaction between multiple non-intelligent parts. The main example of his theory was Minsky's experiments on using a robotic arm, a video camera and a computer to build with children's blocks.
Rather fittingly, Minsky was contracted as an adviser on one of the most well-known fictional depictions of artificial intelligence ever: Kubrick's 2001: A Space Odyssey. He was even mentioned in Arthur C. Clarke's novel of the same story:
Probably no one would ever know this; it did not matter. In the 1980s, Minsky and Good had shown how neural networks could be generated automatically—self replicated—in accordance with any arbitrary learning program. Artificial brains could be grown by a process strikingly analogous to the development of a human brain. In any given case, the precise details would never be known, and even if they were, they would be millions of times too complex for human understanding.
Letting Hackers Be Hackers
Unbeknownst to many, the hacker community also owes itself a debt to Marvin Minsky. Steven Levy, while working on a book about hackers, discovered that all of hacker culture originally started at MIT -- moving from the Tech Model Railroad Club to Minsky's Artificial Intelligence Lab in the 1960s. Although Minsky was never a hacker himself, he encouraged those who were exploring the uses and limitations of computers at the time, becoming as Levy describes it, a 'ringleader' for that community.
He put up with a multitude of pranks from this newly created group of hackers -- even behaviour that occasionally bordered on criminal. Of course hacking isn't entirely built on illegality and pranks (even if that may be a large part of it) and under the general environment of freedom that hackers experienced in Minsky's lab, these hackers began to create the essential principles of open source software, according to a farewell to Minsky on opensource.com. Says Rob Tiller:
Along with his impressive intellectual achievements, he left a legacy of the seeds of open source. He inspired his students to experiment with early computers, and they did so with an enthusiasm that bordered on mania. He encouraged development of games and play with what were then fantastically expensive machines.
More than anything, Minsky seemed to be a great mind who wanted to foster ideas from other great minds, rather than focusing always on his own concepts. This can be seen through his years spent at MIT, mentoring generation after generation of computer scientists. His protégés are as numerous as they are talented, including such minds as Danny Hillis, who founded supercomputer manufacturing company Thinking Machines Corporation.
The Most Useless Machine
The best known of Minsky's work may be the more absurd things that came out of his mind. For one, he came up with the concept of the 'useless machine' -- that popular internet sensation, a machine with the sole purpose of turning itself off again. While the original machine was constructed by Claude Shannon, it was Minsky who had the original idea that led to its conception.
While this concept has led to a number of different iterations after finding fame in the depths of silly YouTube videos, the original, unadulterated useless machine is still a classic. Minsky originally called it his 'ultimate machine', but obviously this name did not catch on like the more honest description of its capabilities did.
While not as popularly known, Minsky was also keenly interested in the idea of alien intelligence, and proposed a theory that aliens would have to function in similar ways to humans, and therefore communication with alien intelligence would be entirely possible.
When first we meet those aliens in outer space, will we and they be able to converse? I'll try to show that, yes, we will–provided they are motivated to cooperate–because we'll both think similar ways. My arguments for this are very weak but let's pretend, for brevity, that things are clearer than they are. I'll propose two reasons why aliens will think like us, in spite of different origins. All problem-solvers, intelligent or not, are subject to the same ultimate constraints–limitations on space, time, and materials. In order for animals to evolve powerful ways to deal with such constraints, they must have ways to represent the situations they face, and they must have processes for manipulating those representations.
These two reasons were 'economics' -- the idea that intelligence would develop symbol-systems for representing common objects, causes and goals -- and 'sparseness' -- the reasonable assumption that every evolving intelligence will eventually stumble upon the same concepts, such as arithmetic, economics and causal reasoning.
Minsky also developed the first head-mounted graphical display in 1963 -- the original predecessor of devices such as the Oculus Rift, Google Glass and Gear VR, along with a multitude of other headsets that many big companies are currently developing or improving.
The Father Of Artificial Intelligence
Like many of the great minds profiled in this series, Minsky was a recipient of the Turing Award, the highest honour in the field of computer science. The folks over at Wired even partnered with an automated news-writing bot from Automated Insights to get a robot to write his obituary last week. While it's not quite as lyrical as some of the human-written farewells that have surfaced, it only seems fitting that one of his figurative 'children' has written this simple send-off.
These Are Your Numbers is a new Lifehacker series where we profile great minds that have made significant contributions to robotics and computing.