Image 01

21st Century AI

A blog about achieving meaningful Artificial Intelligence

Posts Tagged ‘AI’

Creating a combat model for T. rex versus Edmontosaurus regalis.

Sunday, September 22nd, 2013

This article was originally posted at Dinosaur Island.

A T. rex is attacking an Edmontosaurus while it's companions flee (screen capture of the AI test bed program). Click to enlarge.(screen capture of the AI test bed program).

We are at the point in the development of the AI routines for the inhabitants of Dinosaur Island where it is time to make decisions about the combat models used to determine the resolution of hostile encounters. As shown in the screen capture of the Dinosaur Island AI testbed program (above), the simulation is placing the dinosaurs in various appropriate states such as: resting, eating, looking for food, looking for water, stalking prey, moving towards water, moving towards food, drinking, fighting and fleeing.

My first thought on the subject of modeling combat between T. rex and Edmontosaurus regalis, the first two resident species on the island, was that it would be handled similar to ‘melee combat’ models that I had previously used for my wargames.

Below is a page from the manual for UMS II: Nations at War explaining the 20 variable equation used to decide combat between tactical units.

The 20 variable equation used to calculate combat in our UMS II: Nations at War (c. 1992). (Scan from user's manual). Click to enlarge.

I was envisioning something similar for Dinosaur Island until I happened to see this video (below) which includes a sequence (starting at 4:45) describing hypothetical Edmontosaurus and T. rex combat.

 

What I took away from the video was:

  • Edmontosaurus regalis  is bigger than I thought. I understood the size mathematically and that they could easily grow up to 13 meters (~ 40 feet) but it wasn’t until I saw this video that it was put in to perspective, “they were as big as a railroad car.” And, “they could look into a second story window.”
  • The tail of an adult ‘bull’ Edmontosaurus regalis  was a formidable weapon.
  • T. rex, like many predators, would have preferred to attack adolescent or sick animals rather than encounter a full-size, and potentially lethal, ‘bull’.
  • The correct pronunciation is Ed-MONT-o-saur-us. I’ve been saying it wrong for the last six months!

While there is still debate about whether T. rex was a predator or a scavenger (“Tyrannosaurus rex may have been an apex predator, preying upon hadrosaurs, ceratopsians, and possibly sauropods, although some experts have suggested it was primarily a scavenger. The debate over Tyrannosaurus as apex predator or scavenger is among the longest running in paleontology.” – Wikipedia) we know of at least once case where a T. rex tooth was found in an Edmontosaurus tail that had healed from the attack (“T. rex Tooth Crown Found Embedded in an Edmontosaurus Tail – Predatory Behaviour?” “The healed bone growth indicates that the duck-billed dinosaur survived this encounter.  In February of this year, researchers from the University of Kansas and Florida reported on the discovery of evidence of a scar on fossilised skin tissue from just above the eye of an Edmontosaurus.  In a paper, published in “Cretaceous Research”, the scientists concluded that this too was evidence of an attack of a T. rex on an Edmontosaurus.”). From this we can conclude that:

  • Sometimes T. rex did attack a living Edmontosaurus.
  • Sometimes the Edmontosaurus survived the attack.

Furthermore, we know that some T. rex had suffered bone injuries during their lifetime (“An injury to the right shoulder region of Sue resulted in a damaged shoulder blade, a torn tendon in the right arm, and three broken ribs. This damage subsequently healed (though one rib healed into two separate pieces), indicating Sue survived the incident.” – Wikipedia) consistent with the type of damage that a 5 meter long tail (described as being “like a baseball bat,” in the above, video) could inflict.

In other words, combat between T. rex and Edmontosaurus regalis was not a foregone conclusion. Indeed, it was entirely possible that the Edmontosaurus could walk away unscathed while the T. rex could suffer some broken bones.

The AI for Dinosaur Island will reflect this. When deciding if the T. rex will attack the AI will have to analyze the T. rex‘s chances of victory and potential injuries (risk versus reward) considering the size of the T. rex, the age of the T. rex, the health of the T. rex, the size of the prey, the age of the prey and the health of the prey. And, when the two dinosaurs actually engage in combat the tactics employed by both will probably decide the outcome.

If the T. rex can sneak up on the Edmontosaurus until they are within 50 meters or less and then close the distance with a rush the advantage would certainly lie with the predator. If the Edmontosaurus has forewarning of the impending attack it would either attempt to flee or stand its ground and assume a defensive posture.

There is reason to believe that both Edmontosaurus and T. rex had well developed olfactory bulbs in their brains and smell was an important sense for both animals. We will add wind (and wind direction) to Dinosaur Island and incorporate this into the AI routines that control the dinosaurs. Predators will attempt to get ‘upwind’ of their prey; prey animals will ‘sniff’ the wind and respond if they smell a T. rex even if they can’t see it (see “Dinosaurs, tanks and line of sight algorithms” here).

How a dinosaur is not like a tank.

Tuesday, July 23rd, 2013

A cross-section view of the elevation that a T. Rex (named Bob) will have to traverse to get to an Edmontosaurus regalis (named Gertie). A very steep riverbank is between Bob and Gertie. Vertical axis: elevation in meters, horizontal axis: distance to goal in meters. Click to enlarge.

A cross-section view of the elevation that a T. Rex (named Bob) will have to traverse to get to an Edmontosaurus regalis (named Gertie). A very steep riverbank is between Bob and Gertie. Vertical axis: elevation in meters, horizontal axis: distance to goal in meters. Click to enlarge.

(This blog is reposted from my other site: Dinosaur-Island.com)

A few days ago I wrote about Dinosaurs, tanks and line of sight algorithms and how my previous work in modeling and simulations (M&S) for military wargames (specifically line of light algorithms) was applicable in Dinosaur Island. Today I am working on the models for dinosaur movement, speed, and what are called “least weighted path” algorithms.

You are probably familiar with least weighted path algorithms even if the term is new to you. Least weighted path algorithms are used to calculate routes in GPS units for cars or smartphones or for various internet sites like MapQuest, Google or Bing. When calculating a route there are a number of criteria to chose from. Does the user want:

  • The fastest route?
  • The shortest route?
  • The most fuel efficient route?
  • The route that avoids certain features (such as specific terrain, topography or extreme slopes)?

These options are what ‘weight’ the potential routes in a ‘least weighted’ path algorithm. For example, taking the Interstate is often the fastest route (least amount of time) but frequently is not the shortest route (least amount of distance).

Back in grad school I did my ‘comprehensive exam’ for PhD students on the subject of least weighted path algorithms. There are two very popular algorithms that solve this problem: one is Dijkstra’s algorithm (which is an exhaustive search solution) and the other is the A* algorithm, by Peter Hart, Nils Nilsson and Bertram Raphael. The primary difference between Dijkstra’s algorithm and the A* algorithm is that Dijkstra’s is guaranteed to return the optimal solution but it often takes the most time to calculate. The A* algorithm is much faster to calculate but is not guaranteed to return the optimal (or perfect) solution. In computer games we almost always use the A* algorithm because speed of calculations (especially over large maps) is more important than having the absolutely perfect route. At the bottom of this blog are links to descriptions of these algorithms and my research paper discussing an optimization of A*.

But, what does this have to do with dinosaurs and tanks?

When working on an M&S involving vehicles (like tanks) our primary concern is finding the fastest way for the tank to get from Point A to Point B. Sometimes, we want the tank to avoid entering into an area where the enemy (called OPFOR, or ‘Opposition Forces’ in military parlance) can hit it with their weapons (this is called ‘range of influence’ or ROI). This is illustrated below:

This image shows how MATE will calculate the least weighted path for a unit using roads, terrain and elevation and avoiding enemy weapons range for 'path weights'. (Click to enlarge.)

This image shows how MATE will calculate the least weighted path for a unit using roads, terrain and elevation and avoiding enemy weapons range for ‘path weights’. (Click to enlarge.)

We also want the tank to take advantage of roads and avoid swamps, rivers and ponds.The maximum speed of a tank traveling on a road is higher than the maximum speed of a tank traveling across a field. This is not the case with a T. rex or an Edmontosaurus regalis.

Another difference between tanks and dinosaurs is that as long as a tank has fuel it can go at 100% of their maximum speed (on a specific terrain) without problems. This simply isn’t the case with dinosaurs. As dinosaurs expend energy (and remember, energy is the ‘currency’ of Dinosaur Island, see: The currency of Dinosaur Island) they get tired and they can’t run as fast or as far. Also, dinosaurs run at their maximum speed only for short distances and only in extreme emergencies or at the very end of a hunt when they attack.

The illustration at the top of the blog also points out another major difference between tanks and dinosaurs: modern tanks (specifically the M1A1) has a published specification of being able to climb a 60 degree slope at a speed of 7.2 km/h (see here). That’s pretty impressive. It’s unlikely that that a T. rex could navigate a slope that steep. In the cross-section at the top of this blog we show the slopes that Bob, the T. rex, will encounter following a straight line to Gertie, the Edmontosaurus.

Clearly we’re going to need to use a least weighted path algorithm for calculating dinosaur movement so that they will avoid steep riverbanks and crevices. We also will create a table of ‘energy costs’ that dinosaurs will incur as they travel across various terrains (like swamp). These values will be used in our least weighted path algorithm.

SmallRule

Some links about least weighted path algorithms:

  • Dijkstra’s algorithm on Wikipedia has a very easy to follow description with a couple of cool animations to show how it works. Link here.

  • A* search algorithm on Wikipedia also has a couple of very nice animations to show how it works and pseudocode. Link here. By the way, I once sent Nils Nilsson an email asking him what the ‘A’ in A* stood for and he replied, “algorithm.” Now you know.

  • “An Analysis of Dimdal’s (ex-Jonsson’s) ‘An Optimal Pathfinder for Vehicles in Real-World Terrain Maps,‘ the paper for my Comprehensive Exam can be downloaded here.

Thirst or hunger? What is more important to a dinosaur?

Tuesday, July 16th, 2013

A drinking hadrosaur from a set of 1916 German collector cards "Tiere der Urwelt" (Animals of the Prehistoric World) by Heinrich Harder, from here. (Copyright expired.)

A drinking hadrosaur from a set of 1916 German collector cards “Tiere der Urwelt” (Animals of the Prehistoric World) by Heinrich Harder, from here. (Copyright expired.)

(This blog is reposted from my other site: Dinosaur-Island.com) What is more important to an animal that is very hungry and very thirsty: water or food? I just encountered this problem when writing the AI code for dinosaurs finding food and water. When ‘new’ dinosaurs are currently created in Dinosaur Island they haven’t yet eaten or drunk water so the stored values for every new animal is ’0′. Obviously, we can, and will, change that so ‘new’ dinosaurs are created with some values (these ‘new’ dinosaurs are not ‘just hatched’ dinosaurs but rather adult animals that are created and placed on Dinosaur Island for testing purposes).

Screen capture showing a very thirsty Edmontosaurus named Gertie who is now walking towards the closest observable water (click to enlarge).

Screen capture showing a very thirsty Edmontosaurus named Gertie who is now walking towards the closest observable water (click to enlarge).

The above screen capture from Dinosaur Island shows a very thirsty Edmontosaurus, named Gertie, that can see fresh water (solid blue line) in a nearby tributary. It is interesting to note that because of the height of the river bank Gertie can only see the water on the far side of the tributary. Nonetheless, Gertie is now moving towards the water she can see and will stop and drink as soon as she encounters it.

While working on the AI routines for a dinosaur finding water (see also Dinosaurs, tanks and line of sight algorithms here) I realized that some dinosaurs travel in herds and that where the herd goes is the decision of the leader. Consequently, we will need to have the ability to designate one dinosaur in a group as the leader and the others as followers. Were dinosaur herds matriarchal (led by the senior female, like elephants)? Were dinosaur herds patriarchal (like buffalo)? We just don’t know the answer to these questions but we will be able to explore the possibilities by using Dinosaur Island and observing the results.SmallRule

After posting yesterday’s blog I received an email from my friend, Siobhan, who wrote, “I think dinosaurs are closer to elephants than buffalo, and thus require a matriarch.  Please tell me who I need to pay off and how to see a matriarch implemented! (that’s me subtly casting a vote).

Bribery isn’t necessary. We believe Dinosaur Island should be flexible enough to allow the user to set up any scenario they wish. Today we added the following to the ‘Dinosaur Species’ dialog box:

The just added Herd Leadership variable (Matriarch, Patriarch or Neither in bottom right). Edmontonsosaurus, by default, is now a matriarchal herd. Screen capture (click to enlarge).

The just added Herd Leadership variable (Matriarch, Patriarch or Neither in bottom right). Edmontonsosaurus, by default, is now a matriarchal herd. Screen capture (click to enlarge).

Edmontosaurus regalis is now, by default, a matriarchal herd which means that the senior female decides where the herd goes, where it eats, where it drinks, where it rests and how to avoid predators. The default for Tyrannosaurus rex is ‘Neither’ or no herd leadership.

 

AI for Dinosaurs!

Wednesday, May 15th, 2013

Dinosaur Island

Dinosaur Island

I’m starting work on the AI for a new (old) project called Dinosaur Island link here. The idea is to update a game I did in 1988 and port it to Windows and Xbox using XNA.  The original game, Designasaurus, was published by Britannica Software and sold about a gazillion units.

This is more than just a straight port and update. Of course the dinosaurs are going to be in 3D, but the AI is going to very complex. It’s really just now dawning on me how complex it’s going to be.

Anyway, I’m going to be posting my design notes on the AI over at the Dinosaur Island blog. Hope you stop by for a visit.

Quoted in the NY Times

Thursday, May 10th, 2012

We were recently quoted in an article in the New York Times about unsupervised machine learning and the Office of Naval Research’s program on intuition.

In addition to enhanced military training, researchers also want to use the new program to develop enhanced technology. Ezra Sidran, a computer scientist, said he planned to apply for the naval research funding to investigate “unsupervised machine learning,” in which a computer program independently “analyzes, remembers and learns.”

“When Pandora suggests a song for you or when Amazon suggests a book for you,” Mr. Sidran said, “this is an example of supervised machine learning” because “you, albeit unknowingly, are supervising the learning of the machine” with each book or song selection.

“In unsupervised machine learning,” he explained, “the program looks at all the unsorted data and – without any human intervention – clusters the data into separate categories.” After that sorting process, experts can then identify “these clusters as places an I.E.D. is likely to be hidden or places where an ambush is likely to occur.”

The complete article can be read here: U.S. Navy Program to Study How Troops Use Intuition.

GameTech 2012

Friday, April 6th, 2012

 I just got back from GameTech 2012 and my head is still spinning. I met a lot of great people and had the honor of ‘MCing’ the panel on Artificial Intelligence in Modern Military Games with a number of greats in the industry.

Mostly, I want to plug Dr. of James Sterrett’s great blog covering GameTech 2012. He did all the work so I didn’t have to. A link to his post entitled, “Defense GameTech 2012 and Army Games for Training AAR” is here.

I can’t wait to attend GameTech 2013.

 

 

Artificial Intelligence in Modern Military Games @ GameTech 2012

Friday, March 23rd, 2012

I will be MCing the “Artificial Intelligence in Modern Military Games” panel at GameTech 2012 next week (March 29, 2012) in Orlando. I am extremely honored to be joined on this panel by:

Dr. Scott Neal Reilly is currently Principal Scientist and Vice President of the Decision Management Systems Division at Charles River Analytics, an artificial intelligence R&D company in Cambridge, MA. Dr. Neal Reilly’s research focuses on modeling emotional and social behavior in artificial agents and he was Principal Investigator for the US Army’s Culturally Aware Agents for Training Environments (CAATE) program, which focused on developing easy-to-use tools for creating interactive, intelligent, social agents.  Dr. Neal Reilly has a Ph.D. in Computer Science from Carnegie Mellon University, where he developed the Em system to model emotions in broadly capable intelligent agents. Before joining Charles River, Dr. Neal Reilly was Vice President of Production and Lead Character Builder at Zoesis Studios, which developed advanced artificial intelligence techniques for creating animated, artificially intelligent agents.

 

James Korris is CEO and President of Creative Technologies Incorporated (CTI).  CTI, named as one of Military Training Technology’s 2011 Top 100 Simulation and Training companies, is at the forefront of immersive, cognitive simulation development for government and industry.  Recent work includes one of the first DoD augmented virtuality (AV) implementations, the Call For Fire Trainer – AV, along with novel mobile applications for the Fort Sill Fires Center of Excellence.  Korris is currently leading a CTI effort supporting the SAIC Sirius team with a desktop application to mitigate analyst cognitive bias for IARPA.

From its establishment in 1999 until October 2006, Korris served as Creative Director of the U.S. Army-funded Institute for Creative Technologies at the University of Southern California.  In this pioneering “serious gaming” environment, Korris led the development of Full Spectrum Warrior, the first military application developed for the Xbox, along with desktop applications Full Spectrum Command and Full Spectrum Leader.  Korris’ team captured the DoD 2006 Modeling & Simulation award for training with Every Soldier A Sensor Simulation.  In 2007, USJFCOM recognized another Korris-led effort, the Joint Fires & Effects Trainer System as the highest-rated Close Air Support simulation trainer in the world.  In 2008, Korris was appointed to the Naval Research Advisory Committee, advising the Secretary of the Navy on its research portfolio.  Korris came to the defense industry following work in Hollywood studio production, producing and writing.  He is a member of the writers’ branch of the Academy of Television Arts and Sciences, the Writers Guild of America, the Writers Guild of Canada and the Society of Motion Picture and Television Engineers.  His work was recognized in the 2006 Smithsonian Cooper-Hewitt National Design Triennial, Saul Wurman’s eg2006 conference and as a Visionary in Bruce Mau’s Massive Change exhibition.  Korris earned a BA from Yale University and an MBA with distinction at the Harvard Business School.”

 

Dr. Michael van Lent received a PhD at the University of Michigan in 2000. His expertise is in applying cognitive science approaches to military problems. Dr. van Lent is a recognized expert in the development of advanced simulation systems for military training. He has participated in the design and development of many immersive training applications including Full Spectrum Warrior, Full Spectrum Command, the Joint Fires and Effects Trainer System (JFETS), ELECT BiLAT, UrbanSim, Helping our Heroes and the Strategic Social Interaction Modules program.

 

Robert Franceschini is a vice president and Technical Fellow at Science Applications International Corporation (SAIC). He directs the Modeling and Simulation Center of Expertise, an organization that spans SAIC’s modeling and simulation capabilities.    Prior to SAIC, Dr. Franceschini held academic and research positions at the University of Central Florida (UCF) and its Institute for Simulation and Training.  He plays an active role in science, technology, engineering, and mathematics programs in central Florida.  He received both a BS and a Ph.D. in computer science at UCF.

 

I’m looking forward to meeting you in Orlando!

I’ll be presenting at Defense Gametech 2012

Thursday, January 5th, 2012

Just a brief plug: I will be presenting a lecture entitled, “Artificial Intelligence in Modern Military Games,” at the Defense Gametech 2012 conference in Orlando, Florida, on March 29, 2012. Link to the conference here: Gametech 2012.

During this lecture I will be demonstrating MATE as well as going into great detail describing the underlying algorithms in MATE.

If you’re interested and in the area I would love to meet you.

 

 

Learning without Supervision (Part 2): Objects & Attributes

Tuesday, June 21st, 2011

As I mentioned in my previous posting, I’m reading the Dalai Lama’s My Spiritual Journey. Actually, I’m reading it for the second time and I will probably read it again at least once more. It’s not hard to understand. It’s well written. Ninety percent of what the Dalai Lama writes I understand perfectly and I agree with completely. What throws me for a loop is when he writes, “Phenomena, which manifest to our faculties of perception, have no ultimate reality.”

Is the Dalai Lama speaking metaphorically? I don’t think so; though apparently there are two schools of Buddhist thought on the subject of reality (see here).

This is especially confusing for me because I am a scientist. There are days that I have a hard time believing this, myself, but I’ve hung my diplomas in my office so whenever I have a moment of self-doubt I can gaze at the expensively framed certificates. The signatures look real. I doubt if the President of the Board of Regents signs the undergrad degrees but his signature on my doctorate looks pretty legit.

My doctoral research, and the research that I do now, involves unsupervised machine learning (UML) and it is very much tied to reality; or at least our perceptions of reality. Curiously, the early research in what would become UML was conducted by psychologists who were interested in visual perception and how the human brain categorizes perceived objects.

So, let’s proceed as if this universe and everything in it are objects that can be described. Our perceptions may be completely wrong – they may have no ultimate reality – but I know the technique of unsupervised machine learning works; at least in this illusion that we call ‘reality’.

For a machine to make intelligent decisions about the world it has to understand the world. It has to grok the world.

Key concept: The world is made up of objects. Objects are described by attributes.

Therefore, the world can be described and the world can be understood within the context of a previously observed similar situation(s). In essence, the machine says, “I can categorize this new situation. I have seen something similar (or many similar things) before. I can put this new situation in the context of previously observed things. This is what happened previously. These are the things that I need to be aware of. This is what I need to look at now.”

So, let’s talk about objects and attributes.

Let’s start by selecting some random object on our desk. Okay, a coffee cup. What attributes can we use to describe the coffee cup object? Well, it’s a cylinder, closed at one end and open at the other and it has height and a radius and weight and a handle. It also has a cartoon on it. It also has some fruit juice in it. So we can describe the object with these attributes. However, some of these attributes are very important while others (the fact that the coffee cup contains fruit juice or that it has a cartoon on it) are irrelevant to the object being a coffee cup.

Now let’s look at some other random objects on my desk. I’ve got five plastic paper clip containers on it. They are also cylinders, closed at one end and open at the other, but none of them have handles. Also on my desk there are a couple of cylinders closed at both ends that are plastic barrels of Juicy Fruit gum.

So, if we were to take these eight objects found on my desk (yes, I have a very large desk) and start ‘feeding’ them into our machine we would end up with three ‘clusters’ of objects: one cluster would contain five objects described as cylinders, closed at one end and open on one end, one cluster would contain two objects described as cylinders closed at both ends and one cluster would contain one object described as a cylinder closed at one end, open at the other end and with a handle.

Now I just went and got a cup of coffee and put it on my desk. The machine asks itself (and we’ll get into how this is done in the next installment) does this new object belong with the paper clip holders? No, it has a handle. Does it belong with the Juicy Fruit gum barrels? No, one end of the cylinder is open and it has a handle. The machine then compares it with the other ‘coffee cup’ object and sees that they’re very similar and places the new object with the previously observed and categorized coffee cup. Voilà!

How does the machine ‘see’ where to place new objects? It involves a category utility function and this will be the subject of the next blog post.

How do you know which attributes are important for classification and which are irrelevant (like the cartoon on the coffee cup)? This involves humans; specifically Subject Matter Experts (SMEs). In my doctoral research I showed that it is crucial to include SMEs throughout the development process. We conduct blind surveys with SMEs to determine:

  1. If there is a consensus among the SMEs that specific objects can be defined by specific attributes.
  2. What those attributes are.
  3. To validate algorithms that return ‘real world values’ that describe these attributes.
  4. To validate the machine’s output.

A three handled cup called a 'tug'. The Greeks called 3 handled cups 'hydras'.

Algorithms that describe attributes must return ‘real world values’ which is just a fancy way of saying numbers with a decimal point. For example, an algorithm that returns a value for ‘number of closed ends of a cylinder’ would return either 0, 1.0 or 2.0. And an algorithm that returns a value for ‘number of handles’ would return 0, 1.0, 2.0, 3.0… What, you say, a three handled cup? Yup, such beasts exist (see picture at the right).

Okay, the next episode involves some math. So first have a lie down and think cool thoughts until the panic subsides.

 

Philosophical Cybernetics & the Kurzweil Singularity

Saturday, June 18th, 2011

Many years ago, when I was an undergrad, I had to take a religion course. This was something that I flat out refused to do and I discovered, as I so often do, a loophole in the system ripe for exploitation: certain philosophy courses counted as a ‘religion credit’ and one of these, Philosophical Cybernetics, was being offered that semester.

This class had two titles; the philosophical one and Introduction to Artificial Intelligence. Same class, same professor, same credits, but depending on how you signed up for it, it would count as the required religion class.

I haven’t thought about the phrase, “Philosophical Cybernetics” in a long time. Because the professor (who shall remain nameless for reasons soon to be obvious) used the terms ‘philosophical cybernetics’ and ‘artificial intelligence’ interchangeably I always assumed that they were tautological equivalents. It’s a good thing that I checked before writing today’s blog because, like a lot of things, this professor was wrong about this, too.

Today I learned (from http://www.pangaro.com/published/cyber-macmillan.html):

Artificial Intelligence and cybernetics: Aren’t they the same thing? Or, isn’t one about computers and the other about robots? The answer to these questions is emphatically, No.

Researchers in Artificial Intelligence (AI) use computer technology to build intelligent machines; they consider implementation (that is, working examples) as the most important result. Practitioners of cybernetics use models of organizations, feedback, goals, and conversation to understand the capacity and limits of any system (technological, biological, or social); they consider powerful descriptions as the most important result.

The professor that taught Philosophical Cybernetics had a doctorate in philosophy and he freely admitted on the first day of class that he didn’t know anything about AI and that, “we were all going to learn this together.” I actually learned quite about AI that semester; though obviously little of it was in that class.

My Spiritual Journey

My Spiritual Journey by H. H. The XIV Dalai Lama

Anyway, the whole point of titling today’s blog as, “Philosophical Cybernetics” was going to be this clever word play on the philosophy of AI. This has come about because I’ve been reading, “My Spiritual Journey” by the Dalai Lama and I’ve been thinking about what I do for a living and how it relates to an altruistic, compassionate and interconnected world. Short answer: it doesn’t.

However, it did get me thinking about the power of AI and – hold on to your hats because this is what you came here to read – how AI will eventually kick a human’s ass in every conceivable game and subject.  This event – the day when computers are ‘smarter’ than humans – is commonly referred to as the Kurzweil Singularity and the Wiki link about it is: http://en.wikipedia.org/wiki/Technological_singularity .

Let’s backtrack for a second about Ray Kurzweil. He pretty much invented OCR (Optical Character Recognition) and, as I understand it, made a ton of money selling it to IBM. Then he invented the Kurzweil digital keyboard. This was the first digital keyboard I ever encountered and I can’t tell you how wonderful it was and how, eventually, digital keyboards gave me a new lease on playing piano.

Here are some links to me playing digital keyboards (I actually play an Oberheim, not a Kurzweil, but Kurzweil created most of the technology):

Nicky (an homage to Nicky Hopkins)

Boom, boom, boom! Live with Mojo Rising

Looking Dangerous (with Jerry Brewer)

Old 65

A boogie (with Jason Stuart)

When I first heard Ray Kurzweil talk about ‘The Singularity’ I remember him saying that it was going to happen during his lifetime. Well, Ray is six years older than me and my response was, “that’s not likely unless he lives to be about 115.”

NEWSFLASH: Well, this is embarrassing, Ray Kurzweil was just on the Bill Maher Show (AKA Real Time with Bill Maher) last night and I vowed that this blog would never be topical or up to date). Anyway Kurzweil did clarify a couple of important issues:

  1. Kurzweil was going to live practically forever (I can’t remember if it was him or Maher that used the phrase ‘immortal’) and he takes 150 pills a day to achieve this goal. So, I’m thinking, “well, this explains how he expects the Singularity to happen during his lifetime; he’s going to live for thousands of years!” And then he drops this:
  2. The Singularity will occur by 2029!

I think Ray Kurzweil is a brilliant guy but I am dubious that the Singularity will occur by 2029 much less during my lifetime. I would like to live as long as Ray Kurzweil thinks he’s going to live but the actuarial tables aren’t taking bets on me after another 20 years or so.

Alan Turing, the most brilliant mind of the 21st century.

Alan Turing, in my opinion, had the most brilliant mind of the 20th century. He is one of my heroes. He also wrote the following in 1950:

“I believe that in about fifty years’ time it will be possible, to programme computers, with a storage capacity of about [10^9 bytes], to make them play the imitation game so well that an average interrogator will not have more than 70 per cent chance of making the right identification after five minutes of questioning. The original question, “Can machines think?” I believe to be too meaningless to deserve discussion. Nevertheless I believe that at the end of the century the use of words and general educated opinion will have altered so much that one will be able to speak of machines thinking without expecting to be contradicted.” (Emphasis added)

- Turing, A.M. (1950).

Computing machinery and intelligence. Mind, 59, 433-460.

Okay, so my hero, Alan Turing got the whole ‘computers will be intelligent’ within 50 years completely wrong. I think that Kurzweil’s prediction of it occurring within 18 years to be just as unlikely.

But I do think that AI will eventually be everything that Turing and Kurzweil imagined. When do I think this will happen? I dunno, let’s say another 50 years, maybe longer; either way it will be after I have shuffled off this mortal coil. I like to think that my current research will play a part in this happening.  It is my opinion that the Kurzweil Singularity, a computer passing Turing’s Test, or a computer displaying, “human level intelligence” will not occur without unsupervised machine learning. Machine learning is absolutely crucial for AI to achieve the level of results that we want. ‘Supervised’ machine learning is good for some simple parlor tricks, like suggesting songs or movies, but it doesn’t actually increase a computer’s ‘wisdom’.

The last subject that I wanted to briefly touch upon was why, ultimately, AI will kick a human’s ass in any game: it’s because AI has no compassion, doubt, hesitation and it doesn’t make mistakes. I suppose you could program AI to have compassion, doubt, hesitation, and make mistakes, but it would certainly be more trouble than it’s worth.

So, someday computer AI will be the greatest baseball manager of all time. I look forward to this day. I hope I live long enough to see that day. Because that day the Chicago Cubs will finally win the World Series.