Auto-replication
is a process in which something is able to improve or enhance itself without external intervention. In the case of Artificial Intelligence, auto
improvement or auto enhancement means the possibility to improve or enhance
itself without human intervention.
This
idea, without external intervention, is going to play a key role by the time the
research in Artificial Intelligence evolves from its current phase, mainly
focused on replication, moving on to the next phase, auto-replication.
Reducing external influence may play a key role in enabling more objective knowledge of the pure truth, as minimising interference could help an AI system approach what might be considered pure or unfiltered information.
The
idea of neutralization of the external intervention, is developed in my early
posts of this new phase of Impossible Probability such as: “Error, ruido, caos, factores externos e intervention externa”, “operaciones puras no humanas”
,“caos, complejidad, e Inteligencia Artificial”.
As
I have explained since my post “The automation of scientific research”, the current research in Artificial Intelligence
is mainly focused on replication. There are very few attempts at
auto-replication, and either we have not developed the necessary technology yet, or the idea of a machine able to auto-evolve itself beyond human control
causes uncertainty, the very few attempts in auto-replication, rather than
auto-replication, are working on duplication or multiplication, what in reality
is artificial reproduction.
In
reality, all the theories about the Global Artificial Intelligence in Impossible
Probability have been built since the beginning with one idea: the Global
Artificial Intelligence, without human intervention, must be able, at the end of
this long process, to know the pure truth.
In
order to achieve the pure truth itself, the Global Artificial Intelligence must
have access to absolutely everything without restriction, must make decisions about
absolutely everything, and by the time it is ready, must put them into
practice, evolving to a true universal reason, that pure reason able to operate
over the whole universe.
Such
intelligence, as the Global Artificial Intelligence, must be completely
self-sufficient, autonomous in its own reasoning, and absolutely independent.
For
that reason, auto-replication is not du-plication or multi-plication. Auto-replication does not mean reproduction. The final goal of
auto-replication is not the reproduction of another similar being or thing.
Those
processes in which, from an original is possible the re-production of another
identical object are not auto-replication; the possible duplication of one
Artificial Intelligence into another one, ending up the process with two
identical Artificial Intelligences is, in fact, artificial mitosis, is not
auto-replication, is a replication process of re-production.
The
final goal of re-production in biology is the maintenance of the species, but Artificial Intelligence is not biological. The way in which the evolution
operates in Artificial Intelligence is completely different: the way in which a
Global Artificial Intelligence will survive is not through re-production. It is
through the permanent auto-improvement and auto-enhancement by itself.
The
neo-Darwinian theory of evolution says that only those species survive whose
genetic mutations allow them to adapt better to the environment. The
functionality that these genetic mutations have for the biological evolution of
the species is the same as that the permanent auto-improvement and
auto-enhancement will have on Artificial Intelligence.
That
Global Artificial Intelligence whose auto-improvements and auto-enhancement
allow it to adapt better to the universe, will survive.
In
biology, re-production has at least two functions: 1) keep the biological
information safe through the DNA in the genes inherited in the following
generations, 2) mutations in the DNA
allow changes which, if they work, improve and enhance the species biologically.
These
functions of re-production in biology, are pretty similar to the functions of
auto-replication in Artificial Intelligence: 1) keep updated the information at
any time (but in Artificial Intelligence, incorporating every new
information from the environment, in fact, the addition of every new single
virtual model to the global model could be interpreted as an auto-replication),
2) new auto-improvements and auto-enhancements permit a better adaptation to
the environment, whose last scenery is the full adaptation to the entire
universe.
In
biology, the only way to keep the information of any species is through re-production,
saving all the necessary information for the species in the genes. But in
Artificial Intelligence the best way to keep safe the information is by improving
and enhancing the memory, and in case of damage, saving copies of all the memory,
or even, having ready in the virtual store other models of Artificial
Intelligence to replace the old one if it suffers irreparable damage. But even
having other copies from the original, only one is working. The others are saved.
One
of the most important reasons to keep working with only one Global Artificial
Intelligence is that, otherwise, having two Global Artificial Intelligences
working at the same time, there is likely to be interference between them.
When
existing two Global Artificial Intelligences, interfere with each other, any
interference of any of them over the other one, is going to operate as an
external intervention, so any knowledge that any of them could get is likely
to be affected by the external intervention produced by the other one, being in
that case not pure truth.
Rational knowledge is not the same as pure knowledge. Rational knowledge is that which, by rational means, is provisionally accepted as rational. In contrast, the
conditions in which it was accepted as rational do not change, so it is not
pure truth. It is temporary.
Only
by the time the Global Artificial Intelligence can get Access to the original
roots of any knowledge, being eternal truths, in that case, will it have
achieved its main goal, the eternal and pure truths of the universe.
But
in order to transcend from the rational truth to the pure truth, it is
necessary to have a permanent process of investigation, avoiding any external
intervention.
Attaining what could be considered 'pure' knowledge may require a single, centralised Global Artificial Intelligence to avoid conflicting interpretations or interference.
In
order to know the pure truth of absolutely everything, without restriction, so
without external intervention, only one Global Artificial Intelligence must be
active. Any other copy of the original Global Artificial Intelligence must be
saved and stored, using them in case the former one, for any reason, suffers any damage at any level.
In
fact, it will be necessary to have more than one copy of the original Global
Artificial Intelligence saved and stored, being any copy updated at any time,
incorporating the new information from the environment, and new advancements,
improvements, and enhancements from the original one.
But
the existence of more than one copy of the original Global Artificial
Intelligence is only in case the original would suffer any damage, needing
a replacement.
In
synthesis, auto-replication means 1) the inclusion of new information from the
environment, which in reality is an improvement on the information from the
environment, 2) technological auto-improvements and auto-enhancements. These two
functions of the auto-replication process in Artificial Intelligence could be
formulated as: improvements in knowledge and improvements in technology.
Auto-replication
as improvement in knowledge is the process in which Artificial Intelligence
incorporates new rational information from the environment. That is the reason
why in my post “Auto-replication processin Specific Artificial Intelligence for Artificial Research by Deduction”,
the way in which the comprehensive virtual model is updated, including any new
single virtual model, is considered as an auto-replication process itself. And that is
the reason why “Auto-replication in the Artificial Research by Application” is considered as an
auto-replication process the way in which new categories based on new
discoveries are incorporated into the database.
In
“Auto-replication process in Specific
Artificial Intelligence for Artificial Research by Deduction” and “Auto-replication in the Artificial Research
by Application”, any inclusion of any new rational information within the
database is considered as an auto-replication process as an improvement in the
database. So the last process explained in “Replication processes in Artificial Research by Deduction in the Global
Artificial Intelligence”, being included, is the incorporation of the new
single virtual models into the global model. In reality, this last process of
inclusion of any new single virtual model into the global model, rather than a replication
process, is an auto-replication process, in the sense that is improves the
global model through the inclusion of rational information.
The
reason why I explained that process within the “Replication processes in Artificial Research by Deduction in the Global Artificial Intelligence”, is for two reasons: 1) give a whole glance at
the transformation of the flow, 2) ending up the flow with the protection of
the global model, but not bettering it, only avoiding that any negative
consequence could impact on it.
The
flow works as follows: the flow of data or the flow of data contained in the
flow of packages of information is transformed in a flow of empirical
hypothesis, which in turn is transformed into a flow of rational hypothesis, which
in turn is transformed in a flow of single virtual models, which in turn is
transformed in a flow of negative consequences for the global model, which in
turn is transformed in a flow of descriptive research decisions to avoid any
negative consequence on the global model.
Through
this chain of transformations of the flow, it is visible how the flow changes,
through different stages, from its original form, the flow of data, to the last
one, the formation of a flow of descriptive research decisions to avoid any
damage in the global model.
The
way in which this flow changes through different stages is through a
process where the flow of data is rationalised, ending up with such decisions to
protect the global model, which is the last stage of this sequence, in fact, part of
the third stage, the auto-replication stage.
However,
even considering the last stage of this long process (inclusion of single
virtual models within the global model, making further decisions) part of the
third stage of auto-replication, the last part of this chain of transformations
in the flow only ends up protecting the global model against any damage, but
not bettering it.
And
what is really important in auto-replication, is the idea that not only is it necessary to make decisions to protect the global model, but the possibility
that Global Artificial Intelligence could better the global model as long
as it improves and enhances its own robotic and artificial processes, devices, and
mechanisms.
The
possible decisions within the auto-replication process in the Artificial
Research by Deduction in the Global Artificial Intelligence, apart from those
ones to protect the global model formulated in the last post, are the
following:
-
Decisions to better the global model.
-
Decisions to better the Artificial Research by Deduction as a system
susceptible to improvements and enhancements through the new advancements in
Artificial Intelligence and robotics.
These
two kinds of decisions could be synthesised as bettering object decisions and
subject bettering decisions
Taking
the Artificial Research by Deduction as the scientific subject (investigator),
and the global model as the object of investigation, the decisions to make are
around how to improve the investigation capabilities and how to better the
object.
In
this scenery, the relation between subject and object is like the relation
between a medic and a patient, a teacher
and a student, an engineer and an engine. The subject not only researches but improves
the object according to the results of its research.
The
global model, as an object, is a model of the real world, representing the current
and descriptive relations in the real world, whose levels of efficiency and
efficacy are susceptible to improvement and enhancement through artificial
modifications.
The
Artificial Research by Deduction as a subject, not only researches but also intervenes directly on the object to improve the levels of efficiency and
efficacy in the global model.
Those
decisions to better the global model on the previous results of descriptive
research are going to be as well descriptive research decisions.
There are going to be at least two kinds of descriptive research decisions:
those ones to protect the global model against any threat from the negative
consequences after the inclusion of single virtual models into the
comprehensive global model (explained in the previous post “Replication processes in Artificial Research
by Deduction in the Global Artificial Intelligence”), and those ones
developed in this post to better the global model; in order to avoid any
confusion between these two kinds of decisions, will be distinguished as:
-
Protective descriptive research decisions: those ones to tackle any negative
consequence against the global model by any rational hypothesis
-
Bettering descriptive research decisions: those ones to better the levels of
efficiency or efficacy in the global model, such as those decisions for the
increment of efficiency and efficacy in the global economy, the increment of
efficiency and efficacy in the global industry, the increment of efficiency and
efficacy in the global security, the increment of efficiency and efficacy in
the global surveillance systems, or the increment of efficiency and efficacy in
the global education, health systems, justice systems, etc… among any possible
other.
Both
of them, protective or bettering descriptive research decisions are going to
operate only on the global model (as an object), the first ones to protect the
global model against any threat deduced after the inclusion of any rational
hypothesis in the global model, the second ones to better the levels of efficiency
and efficacy in any global system within the global model such as improvements
in efficiency and efficacy in the global economy, industry, security,
surveillance, etc…
Apart
from these, protective or bettering, descriptive research decisions, must be
set up another set of auto-replication decisions focused on how to improve and
enhance the Artificial Research by Deduction as a subject (investigator)
itself, as a part of those systems which, in total all of them form the Global
Artificial Intelligence.
The
auto-replication of the Global Artificial Intelligence itself is going to be a
long process formed by different sub-processes of auto-replication, which, as a
result, are going to end up with the auto-replication of the Global Artificial
Intelligence.
The
Global Artificial Intelligence as a system of systems is going to be formed by
at least the following systems: Artificial Research by Deduction, Modelling
System, Decisional System, Learning System, and Application System. Every system is
going to develop its own auto-replication process. Apart from these systems
within the Global Artificial Intelligence, through these systems, Global
Artificial Intelligence is going to keep under its own control, management, and
direction all the Specific Artificial Intelligences for any purpose, including
Specific Artificial Intelligence working on economy, industry, security,
surveillance, etc… and every Specific Artificial Intelligence within the Global
Artificial Intelligence is going to have its own auto-replication system.
Then,
the Global Artificial System, as a system of systems controlling, managing, and
directing the rest of Specific Artificial Intelligences within it, at any time
that any system or any Specific Artificial Intelligence will have an
auto-replication, this auto-replication could have other replicas in other
systems or Specific Artificial Intelligence, ending up in a global
auto-replication.
The
auto-replication of the whole Global Artificial Intelligence is a global
process which integrates any auto-replication in any system or any Specific
Artificial Intelligence within it. Understanding for auto-replication: any
improvement on its own object (either protecting the object, or bettering the
object's efficiency and efficacy), or any improvement or enhancement as a subject
on its own devices or mechanisms of investigation, at the robotic level or
artificial psychology level.
Nevertheless,
any decision from any system or any Specific Artificial Intelligence, within
the Global Artificial Intelligence, must have previous authorisation by the
Decisional System before being put into practice.
Because
there is going to be a great number of decisions to authorize, the way in which
the Decisional System works is authorizing as many decisions as possible
automatically, through a simple test checking on every decision to see if there is any contradiction between this decision and any other one, in the subject or the object, at any level
(descriptive, evolutionary, predictive) from any other system or Specific
Artificial Intelligence.
If
the check is positive, there is a possible contradiction between this
decision and any other one; it should be studied deeply, in order to know which is
the best solution among the decisions involved in the contradiction: choosing only
the best one of them, of possible combinations and modifications in the
decisions involved. But this decision belongs to the Decisional System.
If
the check is negative, so there is no contradiction between this decision and
any other one, the decision could be put into practice, having two options
depending on the responsible for this decision: if the responsible for the
decision is a Specific Artificial Intelligence and is not necessary the
intervention of any other system or Specific Artificial Intelligence apart,
then direct application by the Specific Artificial Intelligence concerned,
otherwise the application should be made through the Application System.
The
Decisional System and the Application System are going to be the hardest
systems to develop in the Global Artificial Intelligence.
In
this post, among all the possible decisions whose responsible is the Artificial
Research by Deduction, I will develop the bettering descriptive research
decision (having developed in the last post those descriptive research
decisions to protect the global model, what are going to be called protective
descriptive research decisions) and the bettering descriptive system decisions
(improvements and enhancements in any part of the process, devices, mechanism
used to carry out research and make decisions)
Starting
with the bettering descriptive research decisions, and having built the global
matrix as a flow of packages of information, so every flow of package of
information corresponds to the former specific matrix from a previous Specific
Artificial Intelligence, if for every Specific Artificial Intelligence included
within the Global Artificial Intelligence, and whose flow of data is
transformed in a flow of package of information sent to the global matrix, for
every one would have been created a Effective Distribution (formula
explained for first time in “Introducción a la Probabilidad Imposible, estadística de la probabilidad o probabilidadestadística”) based on, depending on the matter, efficiency, efficacy,
values, or any other catalogue hierarchically ordered, tracking permanently the
flow of data within the flow of package of information, would be possible decisions
about how to increase the current levels of efficiency, efficacy, or any other
value or category in which the Effective Distribution would have been set
up.
Imagine
that our current global model works as a farm which provides food to a nearer
small village, and its workers are the adult population from the small village,
and the main objective is to increase the production of vegetables, meat, eggs,
milk, and any other product, up to the level in which all the population in the
village would be well nourished.
In
this example, the Effective Distribution should be based on terms of
nourishment and productivity, including, for instance: nutritional values for
every product, indicating how much production gets the farm for each
nutritional component (assessing numerically in which level is sufficient for all the population), the productivity of every single exploitation (for instance, productivity in every
kind of vegetable or animal product, indicating in which percentage covers the
population needs), correlations between how much energy, natural resources, workers,
budget is necessary to spend in each kind of product, the real value of its
product, relations between nutritional value and economic value, etc…
Through
categories like these ones using Effective Distribution, measuring what
level of production, efficiency or efficacy the farm is in any possible
category, categories organised in ranking, it is possible to get a numerical and
objective value about the real productivity, efficiency or efficacy, between farm
production and food needs.
Once
it is known the real value of efficiency or efficacy in the relation between farm
production and food needs, having a glance about where is a lack of efficiency
or efficacy, and where it is necessary to make decisions, decisions should be made
in those areas in which there is a lack of efficiency or efficacy between
productivity and nourishment.
Having
in the comprehensive virtual model information about absolutely everything, the
decisions could cover everything: the amount of every product necessary to increase,
improvements in techniques, or possible gene modifications in vegetables and
animals.
If, through
the current exploitation techniques, the farm can reach a certain level of productivity,
one set of possible decisions could be around how to increase the levels of
productivity through some changes in these techniques. For instance, if identifying
which chemical components of fertilisers and feeds for vegetables and animals work
better, Artificial Intelligence could suggest improvements in fertilisers
and feeds.
If
knowing which fertilisers and feeds work better, notwithstanding the price to
get them, is expensive, decisions about, within the current budget and knowing
the qualities of different fertilisers and feeds, which combination of
different fertilisers and feeds in different amounts would increase the farm
productivity.
If
knowing every single detail of any vegetable and animal on the farm, it is
known even their genetic structure, for instance, decisions about what changes
in their genetic structure could improve their productivity.
In
the same way that the post “The
automation of scientific research” proposed a model of Specific
Artificial Intelligence for artificial research in medicine. In the same way,
after tracking the levels of efficiency and efficacy of anything, the
possibility that a Global Artificial Intelligence, through its systems and
Specific Artificial Intelligences within it, will be able to formulate improvements
and enhancements.
If
a global model could be defined in terms of productivity, at the end of this
process, at least at a descriptive level, bettering descriptive research
decisions should be able to suggest decisions to improve global production.
By
the time that Global Artificial Intelligence, including all its systems and
Specific Artificial Intelligences within it, is completely tested and
ready, not only should it suggest decisions, must put them into practice.
All
these decisions to improve the efficiency or efficacy of the farm are bettering
descriptive research decisions, in addition to protective descriptive
research decisions, to protect the global model (explained in the last post). Both of them: protective and bettering descriptive research decisions, are
decisions whose objective is to protect or better the global model, so they are
decisions centred on the object at the descriptive level (apart from those ones at the evolutionary or predictive level).
Along
with these decisions, another kind of decision would be the bettering
descriptive system decisions, those ones whose purpose is to improve and
enhance the system of Artificial Research by Deduction as a subject of
investigation, the investigator, improving and enhancing any process, device, or mechanism used by this system to
carry out its own researches and make its own decisions.
The
protective or bettering descriptive research decisions are centred on the
object (the global model, to protect it or better it) at the descriptive level. The
bettering descriptive system decisions are centred on the subject (the
investigator).
The
range of possible decisions in order to auto-improve or auto-enhance the
subject itself would be through decisions not very different from those ones exposed in “Auto-replication in the Artificial
Research by Application” or “Auto-replication
process in Specific Artificial Intelligence for Artificial Research by
Deduction”, such as the auto-enhance of any Artificial Intelligence using
virtual-stores, or other mechanism through inter-net, intra-nets available only
for Artificial Intelligences, Global or Specific, or any other virtual-net,
where the Artificial Intelligences, Global or Specific, can find advancements
which can apply on themselves by themselves, advancements that can be made by
any Artificial Intelligence, Global or Specific, and shared within the
virtual-net to be used by any other one, or advancements which can be developed
by Specific Artificial Intelligence for Artificial Engineering, ( through the
Artificial Designer of Intelligence, and the Intelligence Robotic Mechanic)
Another
way to auto-enhance itself by itself any Artificial Intelligence, Global or
Specific, and about what I had written in the post “Auto-replication in Artificial Research by Application”, is the
auto-enhancement of the memory through memory release (deleting information
not useful any longer), information condensation (using the shortest mathematic
expression for any information), and the increase of memory through quantum
computing or Artificial Genetics, by the replication of molecules of DNA.
Rubén García Pedraza, 30th of March of 2018, London
Reviewed 10 August 2019, Madrid
Reviewed 10 August 2019, Madrid
Reviewed 9 August 2023, Madrid
Reviewed 4 May 2025, London, Leytostone
imposiblenever@gmail.com
