Monthly Archives: May 2014

On Intelligence

OnIntelligenceCoverLast week I went to Singapore for business and endured some dang long flights with a lot of time on my hands.  I tried watching the “Anchorman 2” movie, but it was just too off the wall (unintelligent?) for me, so I gave up on it.  Some of the other movies, like “Jack Ryan: Shadow Recruit”, were better.  But I still had a lot of time, so I turned to reading “On Intelligence” by Jeff Hawkins.

In the early chapters, where he was poking holes at a number of established approaches to intelligence, I was a bit skeptical.  But then, as he settled into his memory-prediction framework, he started to win me over with his different view.

Hawkins talks in depth about the biological processes in the human neocortex, which was interesting.  But the most interesting idea to me was his description of a “memory-prediction framework”.  Basically, this framework includes the obvious case of signals flowing from sensors up to higher cognitive levels, plus the less obvious case of signals flowing down from higher to lower cognitive levels.  Each cognitive level remembers what has previously occurred and predicts what is likely to occur next.  These cognitive levels detect and predict “sequences of sequences” and “structures of structures”.  Predictions allow for missing or garbled sensor data to be filled in automatically. There is also an exception case, where the prediction from above is at odds with the sensor data from below.  Exceptions also flow up the cognitive hierarchy until they are handled at some level.  If they flow high enough, we become aware of them as “something’s not right”.

What I find most intriguing is how this memory-prediction framework might be implemented artificially.  While Hawkins does address this, layered Hidden Markov Models (HMMs) would seem to be a useful direction.  Jim Spohrer tells me that Kurzweil’s book “How to Create a Mind” suggests exactly this, so I’m adding that to my reading list.

I wonder how much training data it would take to train such a model.  I can’t help but think of a baby randomly jerking and flexing its arms and legs; boys endlessly throwing and catching balls; and kids riding their bikes for hours. All these activities would generate a lot of training data.

I also pondered the implications for service science.  Do service systems have a hierarchy of concepts similar to lower and higher cognitive functions?  What kind of “memories” and “predictions” do service systems have?  Service systems always have documents and policies, but that is not the kind of “active memory” Hawkins thinks is important.  Service employees clearly have internal memories, but are there active memories between small groups of employees?  Do departments or entire organizations have memories?  What are the important “invariant representations” of different service systems?  Should we focus on the differences between Person arrives at front desk vs. Guest checks-in vs. Guest is on a week long vacation vs. Guest is satisfied with service? What are the common sequences (or even sequences of sequences) in an evolving customer encounter?  If we knew them, could we can predict next events.  “Be prepared” seems like a more modest and achievable goal for a service system than the kind of moment-by-moment predication Hawkins envisions.

If you’re particularly interested in bio-inspired intelligence, there is a lot of meat in this book to keep you busy and fascinated.  If you’re more interested in the artificial mechanisms for intelligence, like I am, focus on the memory-prediction framework.  Either way, I recommend this book.

Complex Adaptive Systems

ComplexAdaptiveSystemsOn some recent plane trips, I have been reading Complex Adaptive Systems by John H. Miller and Scott E. Page.

The authors discuss a number of topics we all presume we know, but that are seldom explicitly discussed.  For example, they have really nice descriptions of modeling and emergence.  I certainly can’t recall any specific sources for this material.

They talk about “computation as theory”.  Formal mathematical methods  are certainly a valuable means to understand some theories.  But some social systems must have a certain minimal level of complexity before important behavioral phenomena appear.  Those models are just too complex for formal methods.  Computation is the only way forward. This book is all about how to approach building and understanding computational (agent-based) models for social systems.

This book recaps and integrates some material from Wolfram’s A New Kind of Science.  That was also a good readI clearly remember reading that book a decade ago on my backyard patio, bathed in the physical light of the sun and the intellectual light of intriguing ideas.

I also thoroughly enjoyed Scott Page’s (same author) Understanding Complexity lectures which address the same topic, in an abbreviated fashion, in a video format.

The book is a particularly easy read because the examples are simple and clear, and because they use such interesting expressions and metaphors.  (For example, when talking about pixelated pictures, they have this footnote:  “For the romantic among you, assume a stained glass window”.  I loved that).  I also hope to learn a few things about better technical writing by examining the writing style of this book.


Using Commitments: Delegating and Assigning

I promise

I Promise

This time we look at two, additional commitment operations: delegate and assign.  Commitments relate debtors (the agents making the commitment and the agents who are responsible for satisfying it) and creditors (the beneficiaries of the commitment).    But sometimes, one agent makes the commitment, but another agent becomes responsible for satisfying it.  At other times, one agent negotiates the benefits of a commitment, but wants to transfer those benefits to another agent.   So after the commitment has been created, we may want to change either the debtors or the creditors or both.

For example, my Manager makes a commitment to his Executive to do some work and give a presentation on the results of the analysis.

CC1 = C(Manager, Executive, receiveRequirements, performTask and givePresentation)

Then he delegates this commitment to me.


Now, I am on the hook to satisfy the commitment because I’m now the debtor, not my Manager.  Note that the identity of the commitment (CC1) is not changed by these operations; it is still CC1 even after being delegated.

CC1 = C(Scott, Executive, receiveRequirements, performTask and givePresentation)

Now let’s say, the original Executive creditor decides that her manager (SeniorExecutive) really needs to see the resulting presentation instead.  So she can assign the commitment, which changes the creditor for commitment CC1.

CC1 = C(Scott, SeniorExecutive, receiveRequirements, performTask and givePresentation)

Many papers use the operation delegate to change debtors and assign to change creditors, as I’ve described here.  However, in my work, I use the more generic transfer operation to do either delegate or assign.  This is just a low-level detail, since both have the same effect.

Delegate and assign further demonstrates the flexibility of commitments.  Commitments explicitly capture an important type of relationship between agents; they capture something that debtors should do in the future.

Using Commitments: A Legal Contract

I promise

I Promise

OrderPayShip (the “workhorse“) showed one way to use commitments.  In today’s example, I’m going to show a different commitment pattern, which more directly models a legal contract.

Let’s assume Bob sells cars and Alice wants to buy one.  Expensive items like this (much more expensive than a pizza) often involve an explicit legal contract.  In this case, Alice and Bob want binding signatures on a legal contract.  We still need two commitments: one from Alice to Bob and one from Bob to Alice.  The trick is to make both commitments become active exactly when BOTH Alice and Bob have signed the contract.  We can model this legal contract as

C(Alice, Bob,  AliceSigns and BobSigns,  AlicePays)
C(Bob, Alice, AliceSigns and BobSigns, BobGivesKey)

Both commitments SIMULTANEOUSLY become unconditional only after both Alice and Bob sign the contract.  Then both parties have to satisfy their individual commitments.

The Workhorse pattern involved only two agent actions:  pay and ship.  What happens when each agent must perform lots of actions?   The legal contract pattern enables you to collect all of a single agent’s commitments to others into a single commitment.  So a more complex example, where Bob provides continuing service to keep Alice’s car running, might look like this

C(Alice, Bob,  AliceSigns and BobSigns,  AlicePays and AliceBuysGas and AliceMaintains)
C(Bob, Alice, AliceSigns and BobSigns, BobGivesKey and BobServices and BobRepairs)

The difference between Workhorse and Legal Contract is more style than substance.  This means you can approach problems from multiple directions.