The intuitive algorithm
Roger Penrose considered it impossible. Thinking could never imitate a computer process. He said as much in his book, The Emperor's New Mind. But, a new book, The Intuitive Algorithm, (IA), suggested that intuition was a pattern recognition process. Intuition propelled information through many neural regions like a lightning streak. Data moved from input to output in a reported 20 milliseconds. The mind saw, recognized, interpreted and acted. In blink of an eye. Myriad processes converted light, sound, touch and smell instantly into your nerve impulses. A dedicated region recognized those impulses as objects and events. The limbic system, another region, interpreted those events to generate emotions. A fourth region responded to those emotions with actions. The mind perceived, identified, evaluated and acted. Intuition got you off hot stove in a fraction of a second. And it could be using a simple algorithm.
Is instant holistic evaluation impossible?
The system, with over a hundred billion neurons, processed information from input to output in just half a second. All your knowledge was evaluated. Walter Freeman, famous neurobiologist, defined this amazing ability. "The cognitive guys think it's just impossible to keep throwing everything you've got into computation every time. But, that is exactly what brain does. Consciousness is about bringing your entire history to bear on your next step, your next breath, your next moment." The mind was holistic. It evaluated all its knowledge for next activity. How could so much information be processed so quickly? Where could such knowledge be stored?
Exponential growth of search path
Unfortunately, recognition of subtle patterns posed formidable problems for computers. The difficulty was an exponential growth of recognition search path. The problems in diagnosis of diseases was typical. Normally, many shared symptoms were presented by a multitude of diseases. For example, pain, or fever could be indicated for many diseases. Each symptom pointed to several diseases. The problem was to recognize a single pattern among many overlapping patterns. When searching for target disease, first selected ailment with first presented symptom could lack second symptom. This meant back and forth searches, which expanded exponentially as database of diseases increased in size. That made process absurdly long drawn – theoretically, even years of search, for extensive databases. So, in spite of their incredible speed, rapid pattern recognition on computers could never be imagined.
The Intuitive Algorithm
But, industry strength pattern recognition was feasible. IA introduced an algorithm, which could instantly recognize patterns in extended databases. The relationship of each member of whole database was coded for each question.
(Is pain a symptom of disease?)
Disease1Y, Disease2N, Disease3Y, Disease 4Y, Disease5N, Disease6N, Disease7Y, Disease8N, Disease9N, Disease10N, Disease11Y, Disease12Y, Disease13N, Disease14U, Disease15Y, Disease16N, Disease17Y, Disease18N, Disease19N, Disease20N, Disease21N, Disease22Y, Disease23N, Disease24N, Disease25U, Disease26N, Disease27N, Disease28U, Disease27Y, Disease30N, Disease31U, Disease32Y, Disease33Y, Disease34U, Disease35N, Disease36U, Disease37Y, Disease38Y, Disease39U, Disease40Y, Disease41Y, Disease42U, Disease43N, Disease44U, Disease45Y, Disease46N, Disease47N, Disease48Y,
(Y = Yes: N = No: U = Uncertain)
The key was to use elimination to evaluate database, not selection. Every member of database was individually coded for elimination in context of each answer.
(Is pain a symptom of disease? Answer: YES)
Disease1Y, xxxxxxN, Disease3Y, Disease4Y, xxxxxx5N, xxxxxx6N, Disease7Y, xxxxxx8N, xxxxxx9N, xxxxxx0N, Disease11Y, Disease12Y, xxxxxx13N, Disease14U, Disease15Y, xxxxxx16N, Disease17Y, xxxxxx18N, xxxxxx19N, xxxxxx20N, xxxxxx21N, Disease22Y, xxxxxx23N, xxxxxx24N, Disease25U, xxxxxx26N, xxxxxx27N, Disease28U, Disease27Y, xxxxxx30N, Disease31U, Disease32Y, Disease33Y, Disease34U, xxxxxx35N, Disease36U, Disease37Y, Disease38Y, Disease39U, Disease40Y, Disease41Y, Disease42U, xxxxxx43N, Disease 44U, Disease45Y, xxxxxx46N, xxxxxx47N, Disease 48Y,
(All "N" Diseases eliminated.)
For disease recognition, if an answer indicated a symptom, IA eliminated all diseases devoid of symptom. Every answer eliminated, narrowing search to reach diagnosis.
(Is pain a symptom of disease? Answer: NO)
xxxxxx1Y, Disease2N, xxxxxx3Y, xxxxxx4Y, Disease5N, Disease6N, xxxxxx7Y, Disease8N, Disease9N, Disease10N, xxxxxx11Y, xxxxx12Y, Disease13N, Disease14U, xxxxxx15Y, Disease16N, xxxxxx17Y, Disease18N, Disease19N, Disease20N, Disease21N, xxxxxx22Y, Disease23N, Disease24N, Disease25U, Disease26N, Disease27N, Disease28U, xxxxxx27Y, Disease30N, Disease31U, xxxxxx32Y, xxxxxx33Y, Disease34U, Disease35N, Disease36U, xxxxxx37Y, xxxxxx38Y, Disease39U, xxxxxx40Y, xxxxxx41Y, Disease42U, Disease43N, Disease 44U, xxxxxx45Y, Disease46N, Disease47N, xxxxxx48Y,
(All "Y" Diseases eliminated.)
If symptom was absent, IA eliminated all diseases which always exhibited symptom. Diseases, which randomly presented symptom were retained in both cases. So process handled uncertainty – “Maybe” answer, which normal computer programs could not handle.
(A sequence of questions narrows down to Disease27 - answer.)
xxxxxx1Y, xxxxxx2N, xxxxxx3Y, xxxxxx4Y, xxxxxx5N, xxxxxx6N, xxxxxx7Y, xxxxxx8N, xxxxxx9N, xxxxxx10N, xxxxxx11Y, xxxxxx12Y, xxxxxx13N, xxxxxx14U, xxxxxx15Y, xxxxxx16N, xxxxxx17Y,xxxxxx18N, xxxxxx19N, xxxxxx20N, xxxxxx21N, xxxxxx22Y, xxxxxx23N, xxxxxx24N, xxxxxx25U, xxxxxx26N, xxxxxx27N, xxxxxx28U, Disease27Y, xxxxxx30N, xxxxxx31U, xxxxxx32Y, xxxxxx33Y, xxxxxx34U, xxxxxx35N, xxxxxx36U, xxxxxx37Y, xxxxxx38Y, xxxxxx39U, xxxxxx40Y, xxxxxx41Y, xxxxxx42U, xxxxxx43N, xxxxxx44U, xxxxxx45Y, xxxxxx46N, xxxxxx47N, xxxxxx48Y.