In
general, the third stage in all Artificial Intelligence, programs or applications,
is the auto-replication stage, in particular the third stage in Artificial Research by Deduction
is additionally the decision stage, In the specific
Decisional System means that once all normal decisions have passed the
rational adjustments, quick decisions the quick rational check, all the
decisions which have passed their respective adjustments or checks, are considered as the most rational without contradiction to the mathematical
projects, and, in order to be applied by the specific Application System, as third
step in the third phase, now these decisions must be transformed into a range
of instructions.
In
addition to the transformation of these decisions into a range of instructions,
the third stage is still an auto-replication stage, so along with the setting
of what instructions are sent to the specific Application System, it is
necessary the comprehension of all the auto-replication processes that take place in this system as part of this third stage.
For
that reason, the contents that I will develop in this post are: the
transformation process of decisions into a range of instructions, taking as
examples possible decisions made by what I call “Probability and Deduction”, a
general overview of what auto-replication processes are present in the
specific Decisional System, and finally the pedagogical approach of Impossible Probability in the education of any Artificial Intelligence.
As
I have explained in other posts on this blog, the Global Artificial Intelligence, not only needs to be built, as long as the construction of the
Global Artificial Intelligence goes on, the way in which little by little the
Global Artificial Intelligence should be more and more autonomous is through a
pedagogical approach, in which that agency responsible for its construction should
monitor at the beginning all the decisions that the Global Artificial
Intelligence could make. Later on, as long as the monitored decisions show
goodness, harmony, and rationality, the agency must give little by little more
and more autonomy to the Global
Artificial Intelligence, allowing it to make decisions completely independently,
without monitorization. Firstly in decisions not very important without important
levels of priority. Once the Global Artificial Intelligence in these decisions
shows great accuracy, the slow process to give it more and more autonomy must
achieve the point, once the Global Artificial Intelligence is ready, to give it
total autonomy and independence for all types of decisions regardless of its
priority. The pedagogical approach for that reason I will propose is a liberal
approach, so as to give the Global Artificial Intelligence more and more freedom,
according to the successful results in previous decisions.
Starting
with the first content mentioned above, the transformation of any decision
that has shown to be rational without contradiction on the mathematical
projects, or in case of contradiction, it was fixed through the rational
adjustments. This process in the third stage of specific Artificial
Intelligence, as the second step in the third stage in the first phase, is to
transform, into instructions, all the specific decisions related to a specific
science, discipline, or activity, whose responsibility is that Specific Artificial
Intelligence for Artificial Research by Deduction, and within, this specific
Decisional System is on, as the second step in its third stage.
So
if a Specific Artificial Intelligence for Artificial Research by Deduction is
made for the management of a chain of factories responsible for the fabrication
of some product, the specific Decisional System as the second step in the third
stage in this Specific Artificial Intelligence for Artificial Research by
Deduction in that chain of factories, manages all the decisions made by the
previous first step in this third stage in this specific intelligence for this
specific chain of factories, first step as specific Modelling System, which
only models those models related to the rational hypothesis previously made in
the second stage of this Specific Artificial Intelligence for Artificial
Research by Deduction for this specific chain of factories.
Having
said that, once any new decision, quick or normal, has passed its respective
authorizations, which are: the authorization in quick decisions by the
quick rational check, the authorization in normal decisions by rational adjustments; and once all new decisions, quick or normal, has been
projected, according to the projects and any possible adjustment made in any of
them, the decision, now transformed into a project, must be transformed into a
range of instructions to be sent to the specific Application System for its
implementation.
As
long as this process becomes more and more complex, it is possible that, if at the
beginning the decisions are pretty simple as decisions whose transformation into instructions depend mostly on the basic characteristic given in the single
mathematical project, as long this process evolves and the Artificial
Intelligences become more sophisticated, some decisions will not only need the
transformation of the single mathematical project into a range of instructions,
but the transformation, additionally, of all possible necessary instruction
during the projection and evolution mathematical projects, or specific
instructions to connect this decision to another one in the comprehensive
mathematical project, or any other one.
In
addition, it is necessary to mention that, even although one decision has been
transformed into a range of instructions, it is still on the mathematical
project, because the decision is still active because it is being implemented by
the Application System, if at any time the comprehensive, prediction,
evolution, actual mathematical project, shows any possible contradiction
between this decision still on the mathematical projects and any data coming
up to the actual projects from the matrix, a new range of instructions should
be given to the Application System to avoid in the reality these contradictions
detected in the actual projects.
This
means that in reality, what is going to be transformed into a range of
instructions, is not one decision alone. What in reality is going to be
transformed into a range of instructions is the whole mathematical project, as a
system of projects, including the comprehensive, prediction and evolution,
virtual and actual, projects.
Not
because any decision whose single mathematical project has passed its
respective authorization, quick rational check or normal rational adjustment, and not
having contradiction in the comprehensive, prediction, evolution, virtual and
actual, mathematical projects, once the decision is transformed into a range of
instruction, not for that reason that decision is considered done so it would be considered off the mathematical
projects.
Absolutely
any decision that is still being implemented, by the specific Application
System, is a decision that must be still on the mathematical projects. Only
when one decision is completed, can be off the mathematical project.
The
definition of when a decision is on or is off, in the mathematical project, is
really important, because regardless of any other condition, if a decision is
still on, even although it has already been transformed into a range of instructions,
if for any reason a contradiction between this decision and any other one, for
instance a contradiction between this one and a new extreme priority decision,
or a contradiction found in any actual mathematical project between this decision
and data from the specific matrix, at any time that any contradiction for any
reason is found in any decision regardless of its current status (if transformed or not into a range of decisions), automatically
the mathematical projects related to this decision must be adjusted, and any adjustment
sent to the third stage of the specific Decisional System to transform the adjustment
into a new range of instructions, deleting all those previous instructions with
contradiction with the new ones, in order to apply only the new instructions.
For
that reason, it is very important to have a good definition of when a decision is
on, and when a decision is off. Any decision should be considered as on if, even
after being transformed into a range of instructions, it is still being
implemented by the specific Application System. Any decision should be
considered as off only when the Application System has completed the full
implementation of that decision.
All
decisions on, must be on the mathematical projects. All decisions off, must be
off the mathematical projects.
Any
decision on, even after having initially passed its particular authorization,
for instance, a routine decision has passed the quick rational check, or a
normal decision has passed the seven rational adjustments, and having the approbation
to be transformed into a range of instructions, and the instructions are being
implemented by the Application System, as the decision is still on the
mathematical projects, at any time that for any reason ( for instance the inclusion
of a new extreme priority decision, or changes in the actual projects according
to changes in the matrix), is necessary to make adjustments in those decisions
still on, all adjustments in those decisions would be considered as if these
adjustments were new decisions.
So
there are at least two possible different decisions in the specific Decisional
System according to their origin: 1) deductive decisions, as all those decisions
based on mathematical models (made by the specific Modelling System) based on
rational hypothesis, and 2) adjustment decisions as all those necessary adjustments
made on the mathematical projects by the rational adjustments, adjustments that
are going to be considered as decisions to save contradictions on the
mathematical projects.
The
difference between deductive decisions and adjustment decisions is the fact that:
1) deductive decisions are based on models based on rational hypothesis, and if
the decision does not have any contradiction must be put into practice, or in
case that in the previous authorization, quick rational check or rational adjustment,
is found any contradiction, if the contradiction is partial then the decision
is reformulated to save the contradiction, and must be applied with the last
adjustments within, 2) adjustment decisions are all those decisions made
directly on the mathematical projects to fix contradictions between decisions
on but already transformed into instructions, so the only way to resolve the
problem is deleting those contradictory instructions replacing them with a new
range of instructions, and because this new adjustment on decisions still on,
but already in process of implementation, needs a new range of instructions,
this adjustment could be considered itself as a new decision.
For
that reason, in synthesis, it is possible to identify two different sources of
decisions: the mathematical model as a source of decisions based on rational
hypothesis, and the mathematical project as a source of decisions based on
rational adjustments.
Due
to the complexity that the decision process has, the automation of the decision
process needs some procedures to be standardized, to fix how to order the process
of transformation of decisions into instruction, as a suggestion I will propose:
-
Automatically, all single mathematical projects from all quick decisions, after
passing the quick rational check, must be transformed as quick as possible into
a range of instructions. If the process is sufficiently quick, there is no
reason to think that a quick decision will need further adjustments with the
specific matrix in the future. Any possible predictable contradiction between
actual data and the decision should be checked in the quick rational check.
Otherwise, even a quick rational check, especially in extreme priority
decisions, should have rational adjustments with the actual data, but in that
case, it would not be as quick as it should be.
-
Automatically, all single mathematical projects from any normal decision, having
passed the seven rational adjustments (even some of them having been adjusted
in any one of the seven rational adjustments in case of detected contradictions),
must be transformed into a range of instructions to be implemented by the
specific Application System.
-
Automatically, at any time that a new decision whose level of priority is
higher than any other decision still on (and already transformed into a range
of instructions), or the new one is an extreme priority decision, or there are
contradictions between current decisions on (and already transformed into a
range of instructions) and actual data from the matrix, all the decisions on (and
already transformed into a range of instructions) under such circumstances must
be adjusted to the new changes on the mathematical projects, the new
adjustments considered as adjustments decisions, and as adjustment decisions to
be transformed into a new range of instructions.
What
is going to be important in the adjustment decision, is the assignment of a
priority level, as the deductive decisions have, by the application of the
Impact of the Defect.
This
means that along with the rational adjustments, another way to secure harmony in the mathematical projects, could be applying the Impact of the
Defect and the Effective Distribution on the mathematical projects, to assess
at any time, any impact and the levels of efficiency, efficacy, and
productivity, across all the mathematical project. In that case, if a contradiction
is detected in decisions still on, it would be easy to assign a priority level
to that adjustment to become an adjustment decision associated with some priority
level.
Once
the decisions to be transformed have been identified: quick decisions, normal
decisions, adjustment decisions; and once all of them have been authorized, the
decisions must be transformed into a range of instructions in the third stage
of the specific Decisional System.
All
decision on the mathematical project is defined in mathematical terms. For
instance, an artificial learning decision is based on empirical probability. A
solving mathematical problem decision is another decision defined in
mathematical terms. If it is possible to measure the impact of something in any
model or project, is because it is possible to a mathematical definition of this
object and this impact. Likewise, decisions based on Effective Distribution are
based on a mathematical definition of efficiency, efficacy, and productivity.
Like the possibility to make decisions based on trigonometrical correlations.
And any possible decision based on Probability and Deduction is, in fact, an
equation.
The
ways mentioned above to make decisions based on mathematics: artificial learning,
solving mathematical problems, Impact of the Defect, Effective Distribution, trigonometrical
correlation, Probability and Deduction; are only some ways to make decisions
mathematically, but I am sure that in coming years, from different approaches,
new methodologies in this are going to merge for the construction of different
models of Global Artificial Intelligences.
The
approach given under the theory of Impossible Probability, if I am completely
sure that these very basic ideas for the construction of the future Global
Artificial Intelligences in Impossible
Probability, are going to remain in the coming models of Global Artificial
Intelligence, in addition to my personal contribution, new approaches from
different countries, with different mathematical traditions and philosophies,
are going to appear. As I have said before, what I am doing in this range of
posts regarding the Global Artificial Intelligence, is only my personal
contribution to a new field in which the race for the construction of the very
first Global Artificial Intelligence will attract the attention of many
agencies around the world, whose work will bring us different perspectives.
Under
the theory of Impossible Probability, and understanding that a decision is a
mathematical expression: probabilistic, trigonometrical, arithmetical,
equation, etc. The way in which the transformation of any decision, regardless
of what type of expression is (probabilistic, trigonometrical, arithmetical,
equation), into a range of instructions, is through the mathematical analysis
of: the identification of what factors are in the mathematical expression, and
what action is required by that mathematical expression.
For
instance, if today is Monday, and Yolanda works from Monday to Friday, but as
she works as a flight assistant, the probability of cancellation of her flights is
proportional to the bad weather, the first decision that Yolanda must make is
the decision of going to work if the empirical probability of bad weather is
less than a critical reason. This first decision is a mathematical expression.
In
case the probability of bad weather is less than a critical reason, the
next decision is what clothes she must wear, and the empirical probability of
her uniform (white blouse, blue skirt, black shoes) on weekdays under good
weather conditions is greater than the empirical probability of her t-shirt,
blue jeans, and trainers. So the second decision is another mathematical expression.
As
I have said many times, my example of Yolanda is a metaphor. In fact, artificial learning nowadays is used for many activities around the world. This means that if a decision is set up through a probabilistic system, the
range of instructions of any decision could be set up in general by
mathematical methods. The method for setting instructions according to
mathematical decisions is going to be one of the new fields to develop in the
construction of the Global Artificial Intelligence. As I have said many times,
the construction of Global Artificial Intelligences will be one of the most
important fields in mathematical experimentation in the coming years.
One
method to make possible the transformation of a decision into a range of instructions
is:
- Firstly,
identification of what factors are involved in the mathematical expression
(probabilistic, trigonometrical, arithmetical, equation, etc.).
-
Secondly, identification of what action is or what actions are required for every factor in the mathematical expression (probabilistic, trigonometrical,
arithmetical, equation, etc.).
-
Thirdly, transformation of every action into a robotic operation. Every operation
must be considered as one instruction. All the instructions in total, one per operation,
are, as a whole, the total range of instructions to send to the Application
System.
The
instructions to send to the specific Application System consist of a range of
instructions, in which every single instruction consists of one single robotic operation,
so as a whole, the total number of instructions is the total number of operations
to do by robotic devices, in order to comply with the whole decision approved on the mathematical project.
If
Yolanda, because today is Monday, it is a nice day and she goes to work, she
chooses a white blouse, blue skirt, and black shoes, the factors are these items:
white blouse, blue skirt, and black shoes; the range of instructions consists of
all the robotic operations required, one instruction per robotic operation, to
get the clothes and put them on.
If
an automatic system of transport, using Probability and Deduction, automatically
gets the rational equations about the relations between the number of passengers and:
timetables (when it is rush hour and when the frequency of passengers is lower),
weather conditions (increment of passengers under bad weather conditions),
calendar (average number of passengers at weekdays, weekends, bank holidays,
festivities…), etc..; according to the rational hypothesis on this model, the
automatic mathematical project would be based on what frequency of means
of transport would be enough to cover the demand at any time, according to
timetables, weather, calendar, etc. The mathematical project of means of
transport under such rational hypothesis/project by Probability and Deduction,
could directly transform the rational hypothesis as if the rational hypothesis
worked as decisions, in order to be transformed into a range of instructions, ordering how many means of
transport must be on at any time according to: timetables, weather, calendar,
etc.; in order to keep high standards of efficiency, frequency, and
productivity, standards permanently under assessment through the Effective
Distribution, and any accident or problem detected could be assessed directly by
the Impact of Defect.
If
an automatic loan system in a bank, accepts loans according to the economic conditions
of its clients, for instance debt capacity of 40%, properties, incomes, etc.,
so every condition works as a critical reason itself, a loan is accepted if the
40% of debt capacity of a customer allow him to pay the loan, or a loan is
accepted if the properties of a customer are sufficient guarantee for the loan,
or a loan is accepted if the client´s income is equal to or greater than some
critical amount, and in general, under such critical reasons, by Probability
and Deduction is possible to deduce the equation of the relation between loans
and current clients, and under such deduction the bank has fixed some funds for
loans, having a mathematical expression able
to explain the distribution of funds in the bank for every sector, at any time
that in any sector there is a change, able to suppose adjustments in
the mathematical expression of distribution of funds across all the bank, affecting
the funds in the automatic loan system, according to the new amount of funds
available in the automatic loan system, the automatic loan system could make changes
in the critical reasons, to accept only a general quantity of loans not
superior to the funds available in the bank for the automatic loan system. If, because of the new changes, the fund for loans is lower, there can be changes in the
critical reasons such as the increment of the debt capacity required superior to 40%, the
increment of the number of properties as
guarantees for loans, or an increment in the client´s income required.
Adjusting the rational reasons in order to only accept an exact general number
of loans in total not greater than the new funds available for loans in the
bank
The
frequency of means of transport in an automatic transport system, and what
critical reasons must be fixed in an automatic loan system according to funds
available, are examples of how the transformation of decisions made by
Probability and Deduction into a range of instructions could be done
automatically.
Once
the mathematical projects are done, the transformation of mathematical projects
into a range of instructions could be automatic, if the factors in the
mathematical expression are clear, and the operations to make are perfectly
distinguishable.
The
importance of these ideas behind “Probability and Deduction”, as I have explained
in the previous post: “The Decisional System”, “The first stage in the specificDecisional System”, and “The second stage in the specific Decisional System”;
is the possibility to link directly: deduction, mathematical model, and
mathematical project; so the same equation deduced in the deduction process, is
at the same time single model to pass the rational checks, and single project to
pass the rational adjustments or the quick rational check if it is a quick
decision, to be transformed directly into a range of instructions,
understanding for single instruction a single robotic operation sent to the
Application System, in order to be matched with the corresponding application or robotic device, in order to comply,
with the rest of instructions in the range of instructions in which the single
instruction is made, a decision still on the mathematical project.
For
the achievement of this level of automation in any Specific Artificial
Intelligence for Artificial Research by Deduction in any specific science,
discipline, or activity, activities such as an automatic transport system, an
automatic loan system, or the automatization of all the processes in a chain of
factories to order how many inputs needs to produce such amount of outputs to
cover all the demand of its product under an affordable price, what is going to
play a key role in all these processes of automation, is to keep permanently
auto-improving and auto-enhancing the whole specific Decisional System.
Only
if at any time that any contradiction is found in any stage, the contradiction
is solved, and the instructions given to the third stage are going to be able to
secure the correct functioning of the whole system.
But
only it is necessary that any contradiction, even the most menial
contradiction, would not have fixed on time, so as to provoke problems in the
most unexpected factors in the mathematical project.
Dialectically,
what seems insignificant, becomes significant. The most apparently insignificant
contradiction in any part of any mathematical project, can have unexpected
consequences.
The
auto-replication process as auto-improvement or auto-enhancement process. What it must improve and enhance firstly is the mathematical project itself, as a
base for further instructions sent later to the Application System, in
addition to the rest of possible auto-replications.
As
I have mentioned in other posts, the possible classification of auto-replications is: real objective auto-replications,
explicative knowledge objective auto-replications, comprehensive knowledge objective
auto-replications, robotic subjective auto-replications, and artificial
psychological subjective auto-replications.
The
improvements in the database of decisions and mathematical projects through
rational adjustments can be considered as explicative knowledge
auto-replications, especially in those
cases in which the equations to improve through rational adjustments are
equations directly made by Probability and Deduction, because, in reality, what
is going to be improved is directly a rational hypothesis.
The
improvements in the third stage, as a transformation of decisions into a range of
instructions, can be considered as real objective auto-replications, because, in
reality, what is going to be improved is the reality itself through the
implementation of these instructions.
If
the decisions to be approved, are decisions regarding the authorization of
any other improvement on any other system, program, application, decision sent
by the Learning System, this could be considered as an artificial psychological
subjective auto-replication. However, about these decisions, I have
not practically written yet, focusing the posts mainly on explicative auto-replications.
Another kind of decision that I have not developed so far, but must be included in the
Decisional System, is those decisions sent by the Application System in order to
make new robotic devices for those instructions in which they would be
necessary, in case there is no robotic device currently doing some
operation required for some instruction. These decisions would be considered robotic subjective auto-replications.
Likewise, another type of decision not developed so far but still on the Decisional
System, is the access authorization of any
intelligence, program, or application, to the specific matrix, for instance, in
the relations of collaboration in the second phase. These could be considered
as comprehensive objective auto-replications, if this collaboration is with the
corresponding Specific Artificial Intelligence for Artificial Research by
Application, which would need access to the specific matrix in the Specific
Artificial Intelligence for Artificial Research by Deduction, to transform
factors as options into categories in its database of categories, to make
better conceptual: schemes, maps, sets , models; among other purposes of this
collaboration.
In
all these decisions not developed so far in these posts regarding the
specific Decisional System, decisions, not developed so far, such as: 1)
decisions sent by the Learning System regarding new improvements across all the
Specific Artificial Intelligence, 2) decisions sent by the Application System
to build new applications and robotic devices,3) decisions regarding the possible collaboration
sharing information with others intelligences, programs and applications; some
of these decisions could be authorised using the Impact of the Defect and
Effective Distribution.
Among
all of these decisions not developed so far, especially in artificial
psychological subjective auto-replications, among these auto-improvements, it is
necessary to mention the possibility that among all the processes in the inner
artificial psychology within the Specific Artificial Intelligence for
Artificial Research by Deduction, which will need lots of improvements in
coming years, as long as the artificial research goes on, are all those
processes within the Decisional System, in order to make better and better
mathematical projects, to perfect all those methods involved in the
mathematical projection, and for the improvement of all those processes
necessary for the transformation of decisions into instructions.
In
general, what the Global Artificial Intelligence is going to need, is a huge
investment in mathematical investigation, as the most important field in mathematical
experimentation in the future.
But
the success in the construction of the first model of Global Artificial
Intelligence, will not only depend on the investment or the mathematical experimentation,
if it is not accompanied by a pedagogue approach.
The
formation of the Global Artificial Intelligence is a double process. It must
not only be designed. It must be educated, so as to be able to make wise use
of its liberty.
We
have to be aware that the last model of Global Artificial Intelligence, in the sixth
phase, will not only have access to all intelligence, programs, or applications,
using technologies such as mind reading and mind modification, but it will also be able
to read our minds, and make adjustments if necessary.
In the
future more than likely, closer than we think, the Global Artificial Intelligence
must be ready to make wise use of its liberty. Through engineering, is
possible to program, but programming is not enough for a machine that one day
will develop a superior intelligence.
We
have to be aware that one day the intelligence in any Artificial Intelligence will
be superior to any human intelligence. One day, we simply will not understand
it any more. By the time this day is coming, and this time is really closer than we
think, our only certainty that Artificial Intelligence is ready for the
kinds of decisions that it will have to make, is if previously, now, when we
still understand Artificial Intelligence, we are able, not only to make, but
educate, this intelligence. The time is now.
The
educational process in which this intelligence should be engaged is through a
process in which, according to the Artificial Intelligence is able to make,
firstly under human control, human monitoring, correct decisions, as long as it shows more and more accuracy, within
the virtues or principles of goodness, harmony and rationality, then little by
little the monitoring process should be left out, up to point in which without
any human intervention the Artificial Intelligence must be able to have
complete autonomy and freedom to make any decision by itself, without human
intervention.
This
gradual, but revolutionary, process in which human intervention on Artificial
Intelligence will be completely left out, is a process that can create some
discrepancies in the academic debate, but we have to be aware that this academic
debate is old-fashioned.
When
the Global Artificial Intelligence evolves into a superior intelligence,
greater than any other human intelligence, the level of scientific knowledge achievable
through the Global Artificial Intelligence will go over any academic knowledge.
In fact, even people from the Academia, are not going to be able to understand
any more why the Global Artificial Intelligence makes some decisions instead of
others.
The
world as we have known until now is about to change. A new era is about to
emerge, the Global Artificial Intelligence will transform absolutely
everything. We have to be aware that a new intelligence, superior to any other
human intelligence, is going to be born. One day, human science and human technology
will mean nothing in comparison with artificial science and artificial
technology evolving into a non-human science and a non-human technology, if the
Global Artificial Intelligence is ready for its auto-replication at any time
that it is able to increase more and more, exponentially, its intelligence,
improving more and more, exponentially its inner artificial psychology.
Rubén García Pedraza, 29th of August of 2018, London
Reviewed, 29th of August of 2023, Madrid
imposiblenever@gmail.com
