August 15, 2021
'A Brief History of Artificial Intelligence' by Michael Wooldridge
The time is near at hand which must determine whether Americans are to be free men or slaves. George Washington
Here are some of my top take-aways from reading this book:
‘Entscheidungsproblem’ means ‘decision problem’. It pertains to a mathematical idea intially posed by Godfrey Leibniz in the 1690’s concerning whether a universal computing machine could be built to mathematically provide a true/false answer to any question. Ah, Leibniz - if it weren’t for him and Newton we would not have Calculus. Love those two men or hate them (because, Calculus), the fact of the matter is that mathematics is a man-made language used to help describe and explain the world around us and life in general. This might explain why, in 1928, German mathematician, David Hilbert, set out to prove, among other things, whether mathematics was decidable. The fundamental question ‘entscheidungsproblem’ set out to answer, therefore, was whether ‘…there are mathematical questions that cannot be answered by simply following a recipe.’ pg 11
Enter Alan Turing. Turing invented a computer model that could solve mathematical problems. The model was not a physical device, more of a thought experiment. Given a thought experiement of a Turing Machine considering itself (like a human thinking about her own thinking), Turing ultimately proved that ‘…mathematics could not be reduced to following recipes.’ pg. 14 This may also indicate that Artificial General Intelligence is not, in fact, possible.
It is one of the great ironies of scientific history that Turing invented computers in order to show that there are things that computers are fundamentally incapable of doing - that some problems are inherently undecidable.pg. 57
Turing also derived an elegantly simply test to help a human discern whether a machine should be considered ‘Aritifical Intelligence’ or not. The test is to simply have a conversation with machine via a computer keyboard via some user interface. If the user could not discern whether they were talking with a human or a machine, then the machine would pass as artificially intelligent. The test largely ignore others considerations, such as whether the machine is sentient.
Waves of AI Progress
- 1930’s - Alan Turing and Turing Machines
- 1950’s - Ingredients for model computers come together
- 1956 - 1974 - The golden age of AI
- 1980’s - Rise and Fall of ‘Expert’ Systems
- 1990’s - AI Agents, DeepBlue
- 2006 - Geof Hinton, Backpropagation, Deep Learning
- 2010 - Demis Hassabis founded DeepMind
- 2014 - Google acquires DeepMind
- 2016 - AlphaGo defeats Le Sedol in Go
Our Social Brains
Page 222 of this book talks about the research of British evolutionary psychologist, Robin Dunbar. Dunbar claims that human brain size, as differentiated from the brain sizes of other primates, is such because of the number of social connections humans can manage at one time, which seems to be around 150. Social connections are important to human evolution and survival because tight-knit groups of humans survive better than individuals. I don’t think it’s much of a leap to say that social networks give rise to the need for select individuals to try to derive power and influence, or political control, of groups in order to maximize some form of human well-being (perhaps at the cost of centralized control). Clear examples come to mind, such as FaceBook, Twitter, Nation States, Tribes, etc.
As Carl von Clausewitz is famous for saying, ‘War is the continuation of politics by other means.’
Therefore, is it too much of a stetch to conclude that human consciousness and intelligence is evolutionarily bestowed as a result of our ability to successfully wage war? And wouldn’t this human bias for survival be fundamentally bestowed upon our sentient machine ancestors? It seems possible to me.
OK, So What?
I don’t think the idea of humans becoming slaves to our super intelligent machine overlords is beyond the realm possibility.
In thinking a bit more about the whole idea of Artificial Intelligence and why humankind finds it so intriguing and worth pursuing, I found myself going to darker and darker places. These places are not the future worlds shown to us by George Lucas in his creation of C-3PO and R2-D2, but are in fact closer to the words shown in the ‘Terminator’ movies. I wonder why that is?
The only reason I can truly come up with for creating sentient machines is to help satisfy man’s never-ending quest for power through a compliant, controllable work force. Heck, I always wanted my own robot that would clean my room for me as I was never clever enough to talk my sister into it.
I visited Mt. Vernon this weekend - the venerated home of our (United States of America) founding father, George Washington. George Washington and his family lived on over 70,000 acres of land. There is no way George Washington and his family could manage, much less work and improve, that much land without human slave labor and animals. I read that the slave labor working this land would rise, work, and sleep with the sun and would only have Sunday’s off to be with family. One of America’s horrible realities is that it was built on the backs of slave labor, and arguably, she would not exist today without it.
Similarly, I doubt very much if humankind will be able to colonize the stars (starting with the Moon and Mars) without more and more advanced robotic labor. We do get some sense of that kind of future from the worlds George Lucas created for us on film. Hopefully, the robots in our space colonizing future will be humble, faithful servants like C-3PO and R2-D2.
Regardless of our spiritual and ethical influences, which would seem to counterbalance other contrary forces, humankind does seem, in the end, very driven by Machiavellian principles where the end is ultimately justified by the means. I can’t help but wonder how human biases for seeking power, control and forced labor (the Social network) in pursuit of some perceived greater good will become ingrained in the human-like technologies we create, and how that might ultimately impact human existence (i.e., machines flipping the script on humans in pursuit of control and survival).
Human bias is already a serious problem in Machine Learning.
Perhaps in the end, the answer to the ‘Entscheidungsproblem’ will be humankind’s savior.
As for me, just thinking about Artificial Intelligence, and how to adequately capture it in some clear definition, makes me all the more amazed and thankful to be alive and human.
Some Depictions of AI In Modern Mythology
Taking one last look, sir, at my friends.C-3PO
Quite an experience to live in fear, isn’t it? That’s what it is to be a slave.Roy Batty, the Android
Shall we play a game? WOPR
No one user wrote me. I’m worth millions of their man-years. Master Control Program
I may be homely, buster, but I’m S.M.A.R.T smart. Rosie
podcast learning deep learning neural networks michael wooldridge machine learning artificial general intelligence artificial intelligence mycin eliza alan turing john mccarthy lisp shrdlu shakey