|Homepage / Publications & Opinion / Articles, Lectures, Preprints & Reprints (post 2002)
Intelligence & Life - Real or Artificial?
Peter Cochrane ConceptLabs CA, Co-Founder
It is interesting to reflect that Mother Nature builds systems bottom up, an atom and molecule at a time, to a create complex structures and organism that generally come pre-programmed to perform specific functions. Over millions of years of evolutionary progress mammalian creatures became both cognisant and dominant, as they have learned to adapt to their environment. We on the other hand have built things top down, from large pieces of material that are cut, machined and processed to create machines that are inherently dumb and unthinking. But with the advent of low cost computing, information storage, and sensors, we can now create machines that rival and even surpass us in specific tasks such as landing an aircraft or playing chess. In Mother Nature's case evolution is the key steering and deciding mechanism for the growth of intelligence of experience, whilst in the case of our designed machines it is our intervention is that is key. So far we have afforded the gift of intelligence and ability to our machines, they have learned from us. But we have embarked on a new trajectory where our silicon-children are beginning to learn for themselves as we increasingly gift them small elements of an evolutionary tract.
With all our knowledge and computing power, our attempts to emulate our own abilities in silicon, should we be surprised that we find it difficult to understand the inner workings of the human brain or indeed the brain of any living organism? Teams of scientists have dissected and analysed the structure of nematode brains (worms) to create a road maps of the interconnection of just 22 neurons. At the end of such exercises they have realised perfect maps of connectivity, and zero understanding of how this so apparently simple machine works. So what chance we, with our brain of (an estimated) 1010 neurons, connected by some 103 - 105 dendrites per neuron, being able to understand our own brain? If we cannot decode the workings of the simplest a worms, then without some help we are unlikely to master the human brain. Unless we augment of our brainpower by some artificial machinery that we create, it is unlikely that our depth of understanding will grow significantly.
If we look back at the history of our species it appears that the encephalisation of our brain, and indeed our intelligence has been down to two principle factors. First a need to communicate and the creation of language caused a very rapid growth in the size of the human brain. Second, the need to be able to throw projectiles with great accuracy to kill prey, along with the ability to manipulate materials for making tools, also engendered very rapid growth. No such evolutionary forces have yet imposed themselves on the artificial brains that we have crafted from silicon. Indeed these artificial entities are designed to perform specific tasks and are allowed to adapt within the some limited framework that we dictate. They have never been given free reign, and they have never been in competition with any other life form or intelligence. Moreover, we have denied them a full sensory and ambulatory ability that is necessary for them to independently explore, learn and adapt. And of course, we are the DNA of machines, we dictate their rate of change and direction of evolution, and we will do so until we are forced out of the loop by their progress. Micro-machines and nano-technology will most likely create the next major advance that sees are partial, and then full, exclusion from machine life. Somewhere around 2015 computers will have more computing power and storage capacity than their human creators, and they may even become sentient. Certainly by 2025 they will have surpassed their makers in most things. So what use are such machines and what might they do for us today and in the future, and will they enhance human life and existence?
As a species we are tuned to a world of three physical, and a single time dimension, and our abilities are more or less strictly limited to visualizing things in a four dimensional framework. But in business, education, medicine, engineering et a most of our problems seems to involve far more dimensions and increasingly the problems we face are inherently dynamic and chaotic. Interestingly our military play games continuously and occasionally have a war, whilst in business we are continually at war and never play games. Probably one of the most significant contributions of Artificial Life (AL) and intelligence (AI) to military and business life at the present time is in the area of gaming and building business scenarios for the future. It is well beyond the human ability to think through all the possible moves on the chess board of war or business, whilst our machines are easily able to do just that and present us with scenarios that we may not have contemplated. It is increasingly important for executives of major corporations to be in a position to experiment with decisions prior to invoking them in the real world. The artificial intelligence and AL environments created on silicon are ideal for that purpose. It is worthy of note that the military have escalated these developments to a high level of refinement. At the beginning of the Gulf War the American military had played hundreds of cycles of games in the Middle East on machines, and tested out all conceivable scenarios that allowed them to estimate the size of force they required, the decisions needed at any one point, and the final outcome in terms of cost. I know of no similar enterprise undertaken today by any corporation on the planet, but the threats to corporate survival are equally important in many cases.
On a less visible and possibly significant level, AI and AL systems are currently solving problems in control, administration and logistics for individual manufacturing plants and transport operations. Many of these problems are classified as the Travelling Salesman Problem (TSP), which in human terms involves the compiling of an itinerary to visit a number of cities whilst minimising the distance travelled. Whilst we are capable of solving this problem for five or possibly ten cities we are totally incapable of solving the problem for 20, 200, 2,000 and even higher. It turns out that AI and AL systems can solve such problems, and moreover can solve them in many dimensions. Take for example the man in a van problem, being despatched to repair some piece of high tech equipment. In the case of a telephone company you are looking for a member of your workforce who is a reasonable distance from the customer to minimise travel expense but they also have to have the right skill set and the right components on their van. Such a problem is well beyond human ability to optimise and for those companies that have used AL despatch solutions there have been savings of up to 25% pa on maintenance crew support. At a deeper level such systems are also being employed in chemical and manufacturing plants where the logistics of piece part delivery or fluid flow is again extremely difficult for human beings to oversee and control.
On a more humane level the design of our buildings, roads, trains, aircraft and automobiles is increasingly governed by our enhanced awareness of safety issues. In the past our estimates on the survivability of evacuation procedures could only be estimated on the basis of past tragedies and simple probabilistic assessments. AL simulations have accelerated our understanding and enhanced our ability to design safer environments for living, working and travelling. The techniques involve the creation of software creatures that emulate a range of human reactions to fire or accident, and the motion of individuals and groups herding through doors and corridors. In some cases it has been discovered that the widening of a door or corridor by a mere 20cm means the difference between a 100% successful evacuation and the loss of human life. This broad class of application has significant implications for the logistics of people and goods transportation, and the design of electronic networks. Again significant advances in capability have been realised across a broad front using AL across the design of integrated circuits to the aerodynamics of automobiles.
Another area that has seen the rapid deployment of AI and AL technology has been the stock markets where the trading on long-term trends is still the forte of human beings, but on the short-term chaos it turns out that machines are far superior to humans. Here we see simple algorithms being used by those convinced that mathematics holds the key whilst the most sophisticated of applications employs evolutionary techniques based on past histories and all possible futures using AL.
So what of the future, where will these technologies take us? Our machines have so far been deprived of any sensory capability or indeed connection to the outside world other than by human hands and eyes. Their environment has been devoid of machine-machine competition, and in the strict sense they compete against human beings for any particular task. In the next phase of development we will see sensory capabilities attached to these machines that will give them immediate feedback resulting from their actions and the actions of others. Moreover, competitions between machines is only just starting to be used a technique to aid learning and adaptation. What we have to contemplate is the future business life where we have a symbiotic relationship with silicon that helps us in every aspect of what we do. This will become an essential feature of the future office and home. We have created a networked world of unbelievable speed and depth, where we can create information faster than it can be subsumed, where we have access to everything, but can't find what we need. AL and AI Agent Technologies that track our working practices, to become a facsimile of our thinking, wants and desires, that anticipate and take pro-active action on our behalf, will transform all aspects of business, education and living. When augmented my new simulation and display technologies they will also speed our understanding of the complex, and render decision support an everyday feature,
Imagine for a moment the rather slow and plodding progress of the genetic pool over the last 500M years since the Cambrian Explosion and the last 100,000 years of human presence. Genetics as a means of modifying a species, retaining information, and a learning library, is a very slow and cumbersome process. What we have learnt from AI and AL is that we can speed up evolution by millions of times and see effective lifetimes of individual entities pass in seconds as one new generation after another is replaced and enhanced. Most people on this planet do not understand or can conceive of the power of an exponential function. To illustrate this power I would parallel the impact of bank interest in a compound formula of the type: y = (1 + x)n. If we were to invest $1, for 10 years at an interest rate of 10% we would achieve a final value of $2.59. But on the technology front our interest rate is over 100% and if we take a technology that is advancing at over 100% per annum then our formula reveals an outcome of at least 1024 gain in a 10 year period which implies a gain of over 1M in 20 years and over 1Bn in 30 years. The exponential function that describes this progression is merely a special case of the general compound function: y = an.
An analogy for this would be to sit on the seashore and notice something small on the horizon. By the time we have realised that this is tsunami it is too late to run. Technology operates in exactly the same way, it seems insignificant at first but by the time we realise that something significant is happening it often too late for us to react. AI and AL will do and is doing exactly that. As managers, entrepreneurs, business people and the dominant species on this planet we have to harness this technology to help us keep ahead of the game and understand the implications of our decisions and actions. The recent crash of markets and the dot.com bust were all too predictable had we had the courage to sit down and think and model what the outcome might be. When a market sees 50% of its entire value created in less than nine months we should be prompted to ask the question - is this realistic and is it sustainable? With an increasing amount of telecommunications and IT linking all businesses and all parts of the planet, the prospects for failure that is cataclysmic is accelerating. The world is no longer a well behaved and linear place, it is now highly non-linear, unstable and chaotic. It will increasingly take us by surprise, and it will create outcomes that are counter intuitive. AI and AL offer us a means of modelling the possible outcomes and making good decisions in advance of any cataclysm. The mathematicians and physicist won't thank me for saying so, but as far as I can see AI and AL are the only means we have of solving many of the business engineering and scientific problems that now face us. To be blunt mathematics and physics just won't do, they are fundamentally incapable of modelling the sophisticated and complex situations that we now face. And worse, all our past experiences do not equip us for what is to come. In the same way the power of our eyes are amplified by a microscope or telescope, or our fingers by a spanner, then we have to look to artificial (silicon) brains to magnify ours. If we do not we will be outflanked and outclassed by coming waves of technology. And for business it is essential that we stay ahead, and on top of the game.
Word count = 2220