Artificial Intelligence???

Having just completed my first taught AI program, I think it is time to well and truly put TBM in its place.

First and foremost on the agenda, is to tackle the common misconception when it comes to defining what AI actually is. In this program we considered 4 mutually exclusive definitions of AI:

• AI as acting humanly. as typified by the Turing test
• AI as thinking humanly. cognitive science.
• AI as thinking rationally. as typified by logical approaches.
• AI as acting rationally. the intelligent agent approach.

Let me suggest, this, unfortunate state of affairs is where common understanding of AI meets its first fundamental flaw (of which there are many!).

AI, for successful implementation should realistically be broken into two separate symbiotic areas with one subgroup.

•AI as understanding ->AI as intuition.
•AI as creativity.

The Borg Matrix AI system is already well progressed into the first two, which you may find surprising given the second’s obvious lack of formal, scientific definition.

A recent question, proposed by a colleague, was “how do these three areas tie in with the more accepted research into AI, such as movement, image recognition, game theory and tasks such as chess etc”.

The answer is its simplest form, is they don’t. Consider this; is a blind, immobile person unintelligent?

Intelligence transcends our physical capabilities to complete a certain task; the importance is in the ability to understand why these tasks are needed, the ability to learn the hows and whys of the rules of the system, and the intuition to know when these rules can (or need) be broken.

I will leave the explaining further of intuition and understanding for now, I could go into an entire essay on the results of some of the observed intuition effects of the TBM_AI system but with no way to prove how or why this came about with the introduction of the system, I am happy, for now, just to explain that part of the system was originally specifically aimed at creating a sense of intuition, but in the end, this was superseded by the more abstract algorithms which handle understanding.

However the main topic I wish to forward for now, is AI as creativity.

AI as creativity

The core problem that creativity presents for an AI system is that of quality. Creativity, in its general sense, is actually extremely easy. A systems ability to create is surprisingly simple, and something I only really understood formally more recently during an explanation of the Non-deterministic Polynomial complete set of problems, you can garner a better understanding of this by carefull examination and understanding of illegal primes. It all boils down to the creation of random numbers, combinatorics and symbolic assignment.

To put it in its simple English form. Make a random number, use that as a template for your creation, then if that number is ‘good’ create new things by using that random number in other random templates.

Hopefully, alarm bells should be ringing in your head right about now. Say hello to intuition.

As I said at the start, the problem with creativity is quality. Creativity is the ultimate NP Complete problem. Combined with the sense of how can it be possible to test if a creation is ‘good’, what separates the good from the bad, and how does this change between one creation and another.

In the primary case, this is the reason behind the LTBTBM project (licence to breed). To give the TBM_AI system a domain to tread into this area, and help formulate a final algorithm which can be used (even if it’s phenomenally slow, optimization has already been proved to work well by using higher level definitions of objects, and formalising heuristic progression should help things further), for arranging and testing various theoretical creations.

Some very interesting work has been done in this area in the form of genetic algorithms.

However, for the most part, these are still limited to solving particular problems, and designed to meet specific predefined goals, the trick now is to present a clear algorithm which combines all the current forms of AI, meshes them together to produce a system which runs with autonomy and can be guided as to the particular problems we wish to be solved, discovering the intricacies and sub goals for itself.

Enter: The Borg Matrix. As mentioned in my earlier article of where TBM_AI fits into the AI domain, a significant benefit, is it provides a method of producing a solid, self regulating system. It forms a glue which allows the abstract definition of presentation techniques and methods, to formulate its own goals, test and self supervise with the ultimate goal of forwarding its own survival (optimally defined in a abstract sense).

This leads me nicely into some of the more moralistic questions related to AI. Isaac Asimof was probably among the greatest of men when it comes to theoretical analysis of the implications of AI of this nature, for anyone who hasn’t read his Foundation and Robot City series, he deals with more issues (in the form of science fiction) than I could ever hope to exhibit here. However there are three important, key points I would like to point out.

•AI (can be made to) always keep its promises.
•AI (can be made to) always tell the truth.

•An AI agent system of this nature ALREADY exists and is many, many years into developement..

When you bring the definition of AI under the top three headings, something that is worth considering is that of the internet. The internet becomes a mammoth, dynamic and interlinked Artificial Intelligence system, the likes of which is still considered science fiction by most.

The easiest way, I think, to visualise this, is to abstract out the human interactions. Consider them instead, as semi intelligent agents, less than the whole, moving about the system, creating, imparting intuition to other areas and contributing to understanding and formulation of the system as a whole. Once you do this, the AI tree becomes more obvious, the human agents of the system are not the only ones doing this. Agents become as diverse as programs, games, data acquisition and representation systems (such as weather and other satellites and their homes). The internet is continuously evolving, already far above and beyond any individual agent, be that the human end user, or the satellite collating its information. Admittedly at this moment in time its evolution and expansion is ultimately dependant upon us, but much less so than the simple AI robots trying to learn to play a decent game of football. On top of this, the very distributed nature of the internet is tremendously resilient to catastrophic disasters. In the world of the internet, a catastrophic failure exists more as a small brain haemorrhage, taking out small sections only and leaving the greater portion intact. In fact, it is easily arguable that the internet (in some form) is already more capable of surviving a complete catastrophic failure of the earth, than we are as a species.

Continuing on this theme, consider Google (in fact any search engine would do, but googs is the best example).

Google as an Intelligent Agent: Without getting to deep into the technologies and algorithms behind Google search, there are two important points you should bear in mind. Firstly, Google works so well because it is a distributed system, its database is spread across a wide spectrum of hardware. Secondly, Google’s answers are not fundamentally based on any direct human input (other than those who created the pages in the first place). Given the simply mind boggling data transfer that the Google system handles, such a thing would quite simply be impossible for a human to handle, in actual fact, even a collaborative group of humans struggles to provide indexing of as high quality. However, Google, as it browses the internet does not have access to any more or less information than you sat here reading this.

In short, we have a lot to learn from the internet when it comes to AI systems design, to exist and work well, fundamentally it requires modularity, compatibility and diversity. AI can expand upon every aspect of our lives, from its most fundamental aspects of teaching and learning, through research and design, and all the way up to organising and controlling governments, their actions and their points of view. To this end we are already much closer to the realms of science fiction than many of us would care to believe.

Many, many disaster theories have been presented with relation to AI, on the surface, the most viable I believe is that presented in the Animatrix sub-story of the Matrix films, in which ‘liberals’, wishing to preserve the ‘life’ of AI allows them to develop their own city, which eventually comes back to haunt us as the city grows and wages war on the rest of humanity.

Isaac Asimof takes a different tact, here robots are loathed by earthly society, but widely overused by earth migrants (who unsurprisingly leave earth using ships with engines invented by the AI itself). Here the story is controlled overall by a robot AI with very special capabilities and which at the combined stories pinnacle creates a completely symbiotic world which is ultimately fated to be the best choice for the future of the human race.

Of the many theories related to what separates us from other animals, This definition of AI also presents a simple answer. Creation: Non human animals undoubtedly possess understanding and intuition, but lack in any formidable sense our most valued gift; The ability to create. Now I’m sure that is not entirely true, all animals have the ability to create in some form or other, be that birds designing nests for their young, to beavers constructing intricate dams to provide and protect themselves and their family.

However, in this respect, I am not entirely convinced we have much in the way of any superior, innate ability to create. Our superior abilities seem to come, again from our ability to communicate our creations between each other. Despite what it may appear to the casual observer, Mans true creations are actually very few and far between, the majority of our species never truly create on a regular basis, and even those who consider themselves ‘creative’ seem to spend most of their time either rehashing work that has gone before them, or simply recreating work that has already been done. In this respect, the Internet as a single entity is far superior to any one of us, collating, organising and disseminating our creations and experiences. It has very quickly become our single, most respected point of call for all information and ideas. As we move slowly into the new millennium this creature is already demolishing borders, and challenges even the most talented of individuals to hold it back against its will. Internet warfare is not the guns and death scenarios of old, it is a continual battle of combined wills which has led to the near deification of an entity that most have failed to note in existence or substance. Where this will take us I cannot, and dare not answer, but the fact that the efficiency it provides as an open and evolving beast has so far destroyed all attempts to halt it in every facet, meaning those who refuse or simply ignore its possibilities find themselves suddenly left behind in a world that previously seemed unchanging.

For those who are wondering, how does this description of the internet fit in with the first two, key points:

•AI (can be made to) always keep its promises.
•AI (can be made to) always tell the truth.

Well, let me deal with truth first. The problem with truth is it’s almost never clearly defined. Something can always be accepted as true until it is proven otherwise; The world is flat, there’s a monster in the cupboard, the sky will fall, have all been accepted truths by someone in one time of our history or another. The best we can do with truth is to always strive to prove something false. Once this is done, we are left only with the most likely truths, which simply must be accepted until such time as either they are disproved, or replaced by an alternative possibility at which time truth is replaced by belief.

Truth:

Promise:

One last side note, why Borg and Matrix?
Well, despite obvious similarities of these two concepts and ideas with their associated fictional and non-fictional ideas, The actual linkage between The Borg Matrix and these two is actually rooted in the common language definition of these words. Borg, a.k.a Cyborg:- “a human being whose body has been taken over in whole or in part by electromechanical devices” represents the requirement of unencumbered linkage of TBM_AI for human and agent interaction and communication. More interesting is Matrix, this again symbolises the collection of information, its representation and manipulation.

Hopefully the ideas presented in this document can help improve understanding of a realistic and more comprehensive definition of AI, its implications and systems that already exist. After all, if you don’t understand what AI is, how can you possibly hope to work with, understand, and develop these systems? - All things none of us can avoid, whether we know it or not.

Bibliography:

Word definitions from the Wordnet browser: Princeton University Cognitive Science Laboritory.

Foundation and Robot City Books Written by Isaac Asimov and others.

The Matrix trilogy and Animatrix Movies © 2004 Warner Bros Entertainment inc.