Dado un conjunto N tendente a infinito es inevitable que absolutamente todo suceda, siempre que se disponga de tiempo suficiente o infinito , y he ahí donde está el verdadero problema irresoluble o quid de la cuestión de la existencia ¿quién nos garantiza que dispongamos del tiempo necesario para que ocurra lo que debe o deseamos que suceda?


domingo, 4 de febrero de 2018

Artificial Research by Application


Artificial Research By Application is a possible model, among others, whose main purpose is the complete automation of scientific research, developing Specific Artificial Intelligences for every kind of scientific or academic investigation. Starting by the automation of empirical sciences, owing to they are easier to automatized rather than maths and logic, which will take more time. Although at the end of the automation process, all sciences empirical and analytical, including maths and logic, must be completely automatized.
In long term, one of the last goals of the development of artificial research, the automation of scientific research in all sciences, disciplines, academic or investigation field, is the knowledge of the pure truth of the universe, beyond the human understanding. What is going to need the creation of an authentic artificial pure reason with a wide range of tools that include artificial research tools in empirical sciences and academic fields as well as artificial research tools in maths and logic, for the creation a non-human mathematical logical model, a non-human science, and a non-human technology. Non-human as they could be generated only by an Artificial Intelligence itself, without human intervention, a true non-human science and technology beyond the error and human limitations.
For the achievement of this purpose the current models of Specific Artificial Intelligence are not going to be sufficient, so it is necessary the creation of a Global Artificial Intelligence, whose early stages, as an experiment, could be applied over a nation, a continent, or the whole planet, extending later its limits towards a Global Artificial Intelligence that could integrate information, and make decisions, over the whole universe.
For the creation of a Global Artificial Intelligence, previously, is really important the design of models of Specific Artificial Intelligence in any scientific research, that later could be linked or applied within the Global Artificial Intelligence. Only one Global Artificial Intelligence able to have access to every information without restriction, carrying on all kind of scientific research, and making decisions that could put into practice by itself.
In the previous stages for the creation of the Global Artificial Intelligence is really important the automation of the research process in science, creating tools of artificial research. The very first two possible  models of artificial research (artificial research by application, artificial research by deduction) that are going to be developed within the theory of Impossible Probability in this and coming posts on this blog, are only some examples about how is possible the automation of scientific research by Artificial Intelligence.
The science automation also contributes for the completion of the economy automation, through the continuous investigation and non-stop experimentation in Specific Artificial Intelligence, whose results could give new applications to manage automatically any economic problems at any level (national, continental, or global), making automatic decisions based on the artificial research of the economy, that could later put into practice.
The automation of the economy will have a lot of benefits. A production system based on a real model of production just on time, producing only what is going to be strictly consumed without unnecessary waste, saving natural resources. The automation of the extraction of natural resources, transportation, fabrication, distribution, and delivery, will reduce error and cost.  Cheap prices in more efficient economy, something really useful by the time the global warming starts generating serious economic and social crisis.
By the time the investigation in Artificial Research would have achieved the science automation, the way in which this models of artificial research are going to contribute in the creation of a Global Artificial Intelligence, after the completion of its first stage, a global application with a global data base including absolutely all possible information without restriction, would be through different strategies. 1): linking these Specific Artificial Intelligences in scientific research (as any other in economy, industrial management security, surveillance, etc.) to the Global Artificial Intelligence, so any Specific Artificial Intelligence could share information, investigations, results, and possible decisions to the Global Artificial Intelligence, and the Global Artificial Intelligence could make decisions that could be put into practice by the Specific Artificial Intelligences linked, or by robotic means under the only direction of the Global Artificial Intelligence; 2): a second option, and maybe the best one, after successful experimentation in artificial research, the inclusion of all these  applications  for artificial research within the Global Artificial Intelligence (along with all kind of Specific Artificial Intelligences, regardless of its purpose: economy, industrial management, security, surveillance, etc.), so the Global Artificial Intelligence could have all the possible and necessary applications to carry on any kind of scientific investigation, make any decision, and put into practice its own decisions by itself through its own robotic means. The second option is the best one because the Global Artificial Intelligence becomes a real singularity.
Perhaps, the way in which finally the Global Artificial Intelligence is going to be created is through, after the creation of a global application with a global data base including absolutely all possible information without restriction, the combination of these two options. Firstly, linking all possible Specific Artificial Intelligences (in science, economy, industrial management, security, surveillance,etc.), used by public or private agencies or institutions in any field or activity. Secondly, integration of all Specific Artificial Intelligences, among others for artificial research, within the Global Artificial Intelligence. However, the different possibilities of the integration process, will be developed in other posts, in this one I want to focus on artificial research by application.
Within the theory of Impossible Probability I will develop, at least at theoretical level, what it would be: artificial research by application, and artificial research by deduction. The first one replicates the process of hypothesis formation within the application of a specific science, discipline, or an academic field. The second one replicates the deduction process in order to get hypothesis for any science, discipline, or academic field.
The first one, artificial research by application, the hypothesis formation is replicated within the application for a specific kind of investigation. The second one, artificial research by deduction, the application itself would be a replication of the deduction process that could later be put into practice in any scientific investigation.
In my last post The artificial empirical hypothesis, I developed examples of each kind of artificial research, by application and by deduction, and specifically in the first one, artificial research by application, I developed examples of artificial research by application in medicine and astronomy.
In both of them, I developed, at least at theoretical level, the main lines about how would be the characteristics of this Specific Artificial Intelligence in every stage: application, replication, and auto-replication. Examples that I do not have any doubt they could be improved in fuure investigations in this field. What I leave on this blog is only some contributions. In fact, I am completely sure that Specific Artificial Intelligence such as these ones, and much better, are going to be built in coming years.
For instance, in my last post The artificial empirical hypothesis in the stage of replication in the artificial research by application in medicine, I only proposed replications for the deduction process, but I did not add other enhancements, as the inclusion of a memory about the medical problems of every patient, something really useful doing diagnoses much more personalized depending on the medical history.
Imagine a Specific Artificial Intelligence for artificial research by application in medicine, able to make good hypothesis (diagnoses) about any medical problem in any individual person in the population of a country, continent, or the planet, ordering the production of medicines just on time (production also automatized by Specific Artificial Intelligence for the industrial management), improving itself by auto-replication. While at the same time, it could have a memory with the medical history of every person in a country, continent, or the planet, making predictions in short, medium and long term, about individual, national, continental, global, medical problems, making decisions on how to prevent them.
The creation of such a Specific Artificial Intelligence in artificial research researching at individual, national, continental, global level, making decisions, and, if we have developed sufficient robotic tools, put it into practice, it looks like a Global Artificial Intelligence.  With the difference that, an Specific Artificial Intelligence operating in only one science, or  only one discipline, or only one academic field of investigation, it is only an Specific Artificial Intelligence with a very vast field of action ( from individual to national, continental or global level), while a true and real Global Artificial Intelligence must be able to carry on all kind of scientific research in any science, discipline, academic field, in any place, at the same time that it should have access to absolutely all possible information without restriction, managing all kind of Specific Artificial Intelligences, from education and health systems, and justice to economy, industrial management, security, surveillance, among all others, having sufficient robotic tools to put into practice all kind of decisions simultaneously.
It is necessary a clear idea about what is a Specific Artificial Intelligence, even if it works at planetary level, and what is a Global Artificial Intelligence, that should  be linked or integrate all Specific Artificial Intelligences, wherever they are and whatever they do.
It is possible the creation of an Specific Artificial Intelligence for artificial research by application in medicine which could operate around the world, as any other Specific Artificial Intelligence, such as a global security system or a global surveillance system. But they are going to operate, even globally, in only one science, field, or activity. While a real Global Artificial Intelligence would be global because it would be able to take a whole planet or even the universe as a field of investigation or action, and do researches globally in more than one science, field, or activity, so it could be able to do multidisciplinary studies. So any decision that a Global Artificial Intelligence could make, could be based on the results in more than one research in more than one science, field, or activity, making decisions about practically everything with huge accuracy.
In coming years the race for the creation of a Global Artificial Intelligence is about to start, and it must start as soon as possible.
In the early stages for the creation of a Global Artificial Intelligence, is necessary the creation of Specific Artificial Intelligence for artificial research. One possible model, mong others, could be through artificial research by application.
In general, artificial research by application follows the three general stages of any other Artificial Intelligence: application, replication, auto-replication.  The first one, application, is not really an Artificial Intelligence stage, but is the base for the following stages. Normally the application in one science, discipline, academic or investigation field, or activity, consist mainly of a data base about this specific science, this specific discipline, this specific academic or investigation field , or this specific activity. In this data base, all the information, data, taxonomies, classifications, files, categorizations, events, facts, phenomenon, characteristics of individuals and populations (physically, behavioural, psychologically), or any other thing, even any necessary tool to use later, must be stored and described in quantitative terms.
The second stage, replication, consist of the replication of psychological processes. From collecting empirical information in quantitative terms, measurement, from the environment, including later the information in a memory, and making a deduction, an empirical hypothesis that must be contrasted. The deduction process within the application is not really difficult. Once the empirical information has been collected and measured, is compared with the data base in the application. Those information, data, taxonomies, classifications, files, categorizations, events, facts, phenomenon, characteristics of individuals and populations (physically, behavioural, psychologically), or any other thing, included in the date base, whose quantitative description is more similar to the empirical information, are selected from the data base as possible empirical hypothetical: descriptions, causes, effects, correlations (depending on the objective of each Specific Artificial Intelligence: descriptive studies, identification of probable causes or effects or both, correlational studies,  stochastic studies); empirical hypothesis about what is happening. Once we have a collection of empirical hypothesis about what is happening, by rational criticism all empirical hypothesis are rationally contrasted. That empirical hypothesis or those empirical hypothesis which by rational criticism pass or passes the rational contrast, within the less margin of error, is or are chosen as rational hypothesis, making further decisions upon the results.
The rational contrast process made by the Specific Artificial Intelligence must follow the general steps as in any other rational contrast. Once the empirical hypothesis, or group of empirical hypothesis, has been identified, then the Specific Artificial Intelligence must proceed to the selection of a sample of subjects and options, according to the nature of this or these empirical hypothesis, doing all the necessary calculations to get their empirical value, in order to contrast with a critical reason. If the empirical value is equal or superior to the critical reason, the empirical hypothesis, or group of empirical hypothesis within this criteria, are considered as rational hypothesis provisionally true, forming part of the rational truth.
The reason why in this second stage more than one empirical hypothesis could be true, is because in medicine for instance, we can have a patient with a collection of symptoms, and these symptoms could be produced for one or more causes. It is really important in artificial research that the application can attribute any effect or correlation to one or more variables at the same time, being able to do multicausal studies, integrating in one possible model different variables working and correlating at the same time.
In astronomy, if we want to explain the geological or atmospheric pattern in Jupiter or Saturn, or in our own Earth, is absolutely indispensable to work and make correlations with a lot of variables at the same time.
The real advantage that artificial research should take from humans, is the fact that humans only can work within a very limited amount of variables, for that reason we have the feeling that this world is complex. If we want to understand what is happening, the reality, it is absolutely necessary to build such models of artificial research able to work with ever more exponential amounts of data and variables at the same time, beyond human understanding.
A very developed Global Artificial Intelligence should be able to work with ever more exponential amounts of data and variables, beyond human science. Only then, when a Global Artificial Intelligence can make multidisciplinary and multicausal models integrating an exponential number of factors, factors collected from: its own global data base, and all its applications, and measurements from the environment; the Global Artificial Intelligence would have developed a really deep access to a more isomorphic knowledge, in direction to the pure truth.
An immediate advantage of the combination of different hypothesis from a single data base in a Specific Artificial Intelligence, is to improve itself its own data base, in case it finds out something from the environment that does not match at significant level with previous information in the data base, because it is a new kind of category, event, fact, phenomenon, characteristic, feature, not registered in the data base yet. For instance, in medicine, a new virus or bacteria not found out before, a first evidence ever to include in the data base.
If it finds out something which does not match at a significant percent with anything in the data base, it has to be included describing this new record in quantitative terms, measurements.
The formation of a collection of hypothesis, even though neither of them would match significantly with the new record, would give some clues. For instance, if it finds out a new bacteria or virus, the chemical composition of medicines used in other diseases caused by other virus of bacteria, even though at not at very high percentage of similarity with the new one, could give some clues about what chemicals must be combined for the creation of a new medicine, in addition to any other component which could fix the problem better and sooner, depending as well on the chemical composition of the virus or bacteria.
Through virtual simulations based on artificial models about the structure of this new bacteria o virus, testing every possible chemical combination up to find the most likely to get rid of it.
For the design of medicines, the data base should include a collection of chemicals and combinations, specifying possible use, properties, advantages and reactions. For anything we expect from the Specific Artificial Intelligence, the data base should include all what it could be necessary. The data base is the warehouse, storing information and any necessary tool.
In other cases, the formation of a group of hypothesis, even though neither of them individually fit perfectly, a possible combination of them could be more successful, or at least, give it some clues about possible causes and effects, which combine with some other artificial deductions might have a good grasp about what is happening, the reality.
In this way, artificial research by application could work in combination with artificial research by deduction, so it would be advisable the creation of channels of co-working between them. After the completion of the first experiments of each kind of artificial research, after successful experiments in artificial research by application and successful experiments in artificial research by deduction, a following step would be the experimentation of ways in which they can co-work together. And, after successful, the inclusion of these results in the Global Artificial Intelligence.
Finally, auto-replication. Auto-replication means the artificial ability to improve and enhance itself at any level, from the robotic level to software level. In case of an Specific Artificial Intelligence for artificial research, these improvements and enhancements could be oriented to the improvement of its own application (improvements and enhancements in its data base), improvements or enhancements in its psychological replication processes (modifying them to bettering, substituting some processes for other ones much better, or incorporating new ones depending on new advancements in artificial science, artificial engineering, robotics, etc.).
Nowadays, the artificial psychology is more focused on the psychological replication processes, and the results on auto-replication, if some of them are hopeful and useful, are not so developed as we are going to need in coming years, for the creation of the first models of Specific Artificial Intelligences in artificial research able to auto-replicate itself at all levels, towards a real and complete Global Artificial Intelligence at all levels, from its global data base to its ability to improve and enhance any of its artificial psychological processes and its own applications, at robotic or software level, without human intervention .
Actually, there are some experiments of auto-replication, but much more than replication they are experiments of duplication, but duplication it is not strictly the same than auto-replication. Duplication is only the creation of something identical to other thing, a copy, when auto-replication means the artificial ability to improve and enhance itself.
Artificial Intelligence must not be understood as a copy, is the next stage in the evolution.
The good thing in the current developments in artificial duplication (the fact that one Specific Artificial Intelligence can create another Specific Artificial Intelligence as a copy, duplicate) is the fact that at the same time one Specific Artificial Intelligence can duplicate itself, the original Specific Artificial Intelligence could make some changes in the new duplication, changes that can include, but not necessarily, improvements and enhancements.
Duplication as replication in strict terms is not a strict auto-replication process. But, if something is able to duplicate itself, the original model could be improved and enhanced in the duplicate model.
When the Global Artificial Intelligence will be a reality, its main purpose is not the creation of copies of itself, but one reason, among others, for duplicating itself would be if it was at risk of disappearance, so creating a copy would be one way to survive, or in case it would have decided to reconstruct entirely itself by creating a duplication but with improvements and enhacements. In this last situation, duplication in order to make utter changes across all its components, from robotic to software level, introducing improvements and enhancements, would be one of the most perfect scenes of auto-replication.
Auto-replication as duplication only, without changes, would be like biological mitosis. But, instead of the replication of a single cell, the replication of an entire Artificial Intelligence (Specific or Global)
. If the biological mitosis is the conservation of the entire information (memories and skills) stored in our genes, the artificial mitosis (duplication) of an Artificial Intelligence would be the conservation of all the information (memories and skills) accumulated in one Artificial intelligence (Specific or Global).
But biological mitosis or artificial duplication, are mechanisms to keep save information, not to change them for new and better ones.
So, artificial duplication without any change from the original model, is not strictly auto-replication. Only if the original model made some changes into the new copy, so the copy would not be an exact copy from the original, due to the original would have integrated some improvements and enhancements into the copy, this process of making some improvements and enhancements into the copy during the duplication process by the original, would be an auto-replication process.
But, in general, the auto-replication process not necessary needs to take the duplication process as the only way of auto-replication.
Auto-replication in Artificial Intelligence (Specific or Global) is all kind of improvements and enhancements done by an Artificial Intelligence itself (Specific or Global), within or without a duplication, but always without human intervention.
This auto-replication could be done over its original systems (from robotic systems to software, including improvements and enhancements in its data base, artificial psychology including all its replications, and improving and enhancing its own ways to improve and enhance itself) or the auto-replication process could be done over an improved and enhanced duplication of itself, duplication made at the same time by itself, without any human intervention.
In order to achieve this level in Artificial Intelligence (Specific or Global) is going to be necessary a huge development in robotics, which controlled by an Artificial Intelligence, allow it to operate in the real world, having robotic tools to take information from the real world, upon the information the researches, making decisions that it could put it into practice directly by robotic means.
Rubén García Pedraza, London 4 February 2018