Keith E. McInnis commented on both my last posts implying that ... I don't know what...; yet, I do know why AI fails and why science fiction writers were simply wrong.
The science fiction author Isaac Asimov. introduced (and later expanded upon) three laws of robotics in his 1942 short story "Runaround", although they were foreshadowed in a few earlier stories. The Three Laws are:
Science fiction often turns into science fact. But as of today, there are no actual robots with a capability anywhere close to those depicted in recent movies like I, Robot, and Transformers, or foretold in earlier ones; such as 2001: a Space Odyssey. In this case, one is forced to conclude that futurists and producers have missed the mark. This is mainly because it turns out to be not so easy to fulfill even the First of the Three Laws above; for that alone requires “knowing when to take action and then knowing also what action will likely be successful”. This need is met by the Causal Explanation Methodology or CEM, the diagram in my earlier blog post.
To bring this first law into more relevant terms we might say: "A father may not injure his family, or through inaction, allow his family to come to harm." Would we charge a father with breaking the law if he invested in the market and lost all the family's savings for retirement? Of course not. In this day and age, with the volatility of the stock market and the failure of banks, can any father, any trading program or anyone at all, exactly know which patterns indicate which actions one should take and when? Many believe that to be a matter of experience and almost entirely a matter of chance if not fortune or blessing.
It is not possible to establish a hard law that asserts what is right in a monologue according to the practice of precisely asserting true language where each label has an ascribed meaning. People do not think this way. It is just wrong-headed to think that labels can serve in any semantic sense of true agreement with present context. For this type of agreement is not and cannot be legislated.
As the attempts to date have shown (conclusively, imho) it breaks down. It does this all the time without fail. Why? It is not resilient. This is not a glib answer. All ascribed agreement breaks under the weight of the very act of asserting and ascribing. When an agreements relies on labels it eventually breaks down. It does this because it cuts off the investigation for resonance, it not only observes, it creates and uses complexity as an excuse to assert or ascribe — to state as fact an incoherent relation or misguided notion.
Exponents of this school of thought make claims (stating as fact, speaking in monologue with mechanical and economic efficiency) the agreement to which observes no further affordance. it is not a dialogue: it is not even an agreement but with a label. That is no agreement at all. It is a ‘transmission of information.’ There is no exchange. It does not listen, it translates, it represents. It computes and outputs. It is not resilient; not creative. Questioning is not allowed. Correction is the order of the day. Course correction is commonplace. It does not add to or improve the environment: it restricts the neighborhood and controls, frames and compartmentalizes instead and rejects as noise all that is disagreeable.
It is a reduction to machine.
To make reasonable financial or macro-economic decisions at any interpersonal scale the size of a family or larger; ought one seek patterns, be prudent and have the skills and technical knowledge to trade securities in the chosen sectors and sub-fields? Networking and ‘neural-network’ or machine learning algorithms can sift through the data with the greatest of ease and efficiency. Statistical probabilities have been well-covered and have achieved impressive results. There are mature Bayesian networks as well. Though none of them solve the problem AI set out to solve. They failed due to the incoherence of representation and category errors. They also failed because they did not have the abstract science or the doctrinal function necessary to do any other than the most imperceptive reasoning.
Further, Bayes networks are hobbled by the independence assumption. Any research of human understanding and awareness based upon Bayes theorem is doomed before it begins. People do not have independent stories or narratives. That is false! On the face of it! I am not interested in Bayes theorem when what I need to know is if the abstract value of a trade-up or trade-off is equitable or fair. I am interested in the abstract postulates of geometry applied not to shapes but to forms of abstract currency or commodity.
In light of the good Count Korzybski’s program of general semantics, the mathematician R.D. Carmichael gave a partial definition of a “doctrinal function” when writing about the structure of Exact Thought in his book “The Logic of Discovery” (1937, pp 110-146). A Doctrinal Function is "a body of propositions made up of a consistent set of postulates by such processes as compel assent to the conclusions reached." (Ibid. p 122-123).
Doesn't that make more sense? Think of personal or motivated actions as composed of not just one disconnected action, but as a “pattern of actions” –like those involved in the exchange of cards for cards of different value in a poker game. Just like in the song, "you have to know when to hold them and know when to fold them" over the course of the game.
Which pattern of actions leads to success? In any given situation this has often been a bit of a holy grail --a mystical or mysterious object-- so many gambler's play, instead, with chance. Many people take actions for granted and still more are all preoccupied by chance. The words “you have to know when” may be grammar, may be letters, may be representations but that is not all.
The words speak to observing the complexity and affordance of the moment. They suggest one be vigilant and explore for an opportunity, investigate for resonance. The words suggest playing an active role in “you have to know when to fold them.” They encourage one watch and listen; ask questions. The words mark off a frame and environment where a dialogue takes place. It does not have to be one of talking or conversing: only a dialogue of play actions and exchanges. There is agreement between all players in the environment of the poker game. This is called agreement with present context and it is a characteristic of resilience. Poker is an aged and honored game of chance as are many others. It does not “break down” no matter who plays.
Modern AI or even university or government research has been working on ascribed coherence to labels. Can the entirety of human experience be anticipated, captured in a lable, grasped and understood? It does not have to be; if one has a doctrinal function. We cannot anticipate what kinds of geometric shapes we may be faced with in the future. Though no mathematician is the least uncertain that they can discover everything that can be known about the dimension, area and mass when they are encountered.
I did not go that way and I cannot speak to why anyone ever thought AI would work. I have a new abstract science with a doctrinal functions that marks and grades the resilience of a given or present context. That is, I have an algorithm that seeks out and discovers the amount of resilience between a compression or other symbolic representation and a given or present context. It is fully tested and proven effective. I want to organize efforts to build more useful appliances. That may be the wrong word, though what I have in mind is an appliance for processing exact thought according to a well-known doctrinal function of discovery.
I have conceived a fantasy game for teaching, abstract sciences (geometry, algebra and exact thinking) and how to recognize and use doctrinal functions. I want to target dropouts and hard-case children already at risk or lost. This 'game of life" has a mission to save lost boys and girls who have (against their nature) been caught in a whirling spiral and spectrum of dissonance caused by their nurturing in a system of ascribed coherence. Anyone in the business of making games could score a big win here. I will begin looking elsewhere for developers if no one nearby has interest in this sort of opportunity.