Você está na página 1de 8

Artificial Intelligence (AI) Methodology for Definitions and for Hypotheses

Celal Korkut VATA

Abstract:

Current Artificial Intelligence (AI) Algorithms are mainly based on the facts as defined and stored on
big databases and are often inefficient on assumption based thinking.

This conceptual work aims to fulfill the need for a basic, universal data mining approach utilizing
brute force of calculation power of today's technology on big definition databases based on
postulated hypotheses using ever advancing algorithms.
Introduction:

Mathematical cognition and problem solving abilities are though to earn assets requiring years of
training and inborn abilities for humans. Also, mathematics requires a garden variety of skills of
complicated formula including derivatives, integrals, trigonometric functions, as well as fourier and
lapacian transformations with all the possible defined constants. Simulation algorithms working on
powerful computers are real life examples of these formula taking the heavy load off the human
shoulders. The fact that world heritage assets of i) mathematical, physical, and chemical formula, ii)
the text based literature along with iii) the internet usage statistics of the last decade are in global
databases has a tremendous potential for novel Artificial Intelligence (AI) Algorithms.

However, these formula and text are based on former definitions as well as on former constants
obtained as a results of collaborative global scientific and literature work of the last century and do
not function on undefined bases in an intuitive way.

Although, the algorithms of today are various and classifed as i) clustering, ii) supervised, iii)
unsupervised, iV) fuzzy logic, v) neural networks, vi) SMOTE, they are mainly based on "the find
function" that can be followed by a "replace function" as exemplified by word processors.
Fundementtally, term frequencies for text mining or term assigned measurements for sensor
readings are processed as statistical values by these algorithms to produce decisive outputs. A whole
industry is founded on these simple algorithms. Google Translate (1) is based on predefined statistics
of term-co-occurrences build by the "find -term1 and term2- function".

As such, "find and replace" based functions can also be used in formula solving. "Energy(E)" can be
found on one side of the formula and can be replaced with the value on the other side of the equal
sign as defined by the equation E = mc 2.

This is the basics of mathematical formula solution. The equal sign is restrictive in the sense that it is
a "when condition".
In other words, there is no sense of intuition in the databases and in the algorithms.

On the other hand, philosophically, thinking not only involves analyzing the defined options with
great speeds and accuracies but also putting forward novel hypotheses, postulations and analyzing
and verifying or not verifying their outcome. In other words, "When this happens, that will happen.."
is not a question, it is a function. "What would happen if this happens? "is the right kind of question
for an Artificial Intelligence (AI) Algorithm to work on using big data.

In this paper, we are describing an "if function" on top of the well known, powerful "term frequency
or term value" algorithms of data mining classified as i) clustering, ii) supervised, iii) unsupervised, iv)
fuzzy logic, v) neural networks, vi) SMOTE. Such an "if function" stands at the boundary of
imagination and reality and is an important step towards human level of cognitive function as
examplified by the phrase; " If I were You....". In alternative terms, we are letting Artificial
Intelligence (AI) Algorithms dream about a postulation of interest and remember every statistics of
the dream to dissect the reality part out of the dream for further learning.
Methodology:

A question is asked to the algorithm as such :

"What would happen if Energy (E) was equal to Time (T)?

1. The algorithm checks the database of definitions for Energy (E) = Time (T).

2. If the above definition is not verfied in the database, the algorithm applies the
hypothesis mode and solves all the formula in the database using find/replace function
with respect to Energy (E) = Time (T).

3. The algorithm compares the results with defined equations in the database. If no
previously defined equation is verified, the algorithm confirms and stores the fact that E
is NOT equal to Time. Energy (E ) ≠ Time (T).. And records that information.

4. If solving Energy for Time, (E) = Time (T) verifies a formula that is already defined in the
database, the algorithm verifies, Energy (E) = Time (T), based on the defined
formula in the database and stores and updates the information. The algorithm can also
work conditionally and answer as "Energy (E) = Time (T), provided that π = 1.75321"

*The "if questions" can specifically be asked by a person as well as they can be
performed automatically in an repetetive setting based on a preprogrammed
learning scheme.

*Term co-occurrence statistics with defined threshold values can be used for
verification of text based replacement hypotheses as such "what would happen if dogs
were cats...?"
Potential Results:

Although, such an approach seems to be pretty straightforward, it has tremendous implications and
garden variety of applications. As such, the question "what would happen if π = 1.75321 ?" can be
analyzed, verified or not verified in a fraction of a second on all the formula databases, revealing all
the implications at every level of such a case and all the log information of the analysis can be kept
for future reference with the class information " verified" or "unverified". For example, if Pi is equal
to 1.75321 (π = 1.75321) then the geometry we know today will be completely different, the gravity
and orbits of planets will be redefined, and there won't be cars with wheels.

As stated above, the defined “if function” is not only confined to the formula databases. Also, any
text database can be analyzed using the same approach. The question "what would happen if cats
were dogs ?" will lead to replacement of all "cat" terms with the "dog" terms in the database and
lead to the analysis of the coexistence statistics of other terms such as "bark, bite, and jump" before
and after the replacement (for cat and dogs, separately) and can conclude that i) "cats are not
dogs!", ii) "cats would bark" if the coexistence statistics of "barking - dog" is above the defined
threshold limit with respect to "cat - barking".

Further, the question "what would happen if Trump was Atatürk?" would look up the coexistence of
terms such as "war, no war, peace, lying, courage" with respect to terms "Trump" and "Atatürk" and
the algorithm can conclude Trump is not and can not be Atatürk and there will be "Peace at Home
and Peace in the World" if Trump was Atatürk.

The defined “if function” can also be used for future prediction as such, the question "What will
happen if SDP is elected instead of CDU in Germany?". The algorithm will test all the coexistence
statistics of "immigration, exports, imports, good USA relations, bad USA relations, good Chinese
relations, bad Chinese relations” with respect to “CDU” and “SDP” on the newspaper database and
can conclude that "Germany will move away from USA and get closer to China and that would
happen faster.. "

In a creative fashion, the same "if question" can be asked for a musical database "What would
happen if Steve Vai was Yingwie J. Malmsteen ?", the algorithm can replace Steve Vai for Malmsteen
and can analyze and compare all the "musical note coexistence statistics, i.e. note touches" of Steve
Vai's works and Malmsteen's works" and conclude i), Steve Vai is not Malmsteen and ii) if Steve Vai
was Malmsteen he would play "Rising Force" as the following (2.).

Detailed description and evaluation of the mechanisms of various algorithms is beyond the scope of
this work as we are not experts in the field and are respecting the global hardwork. We are only
forseeing a constant improvement in the field of algorithms along with the global computer
processing power and speed.

From that perspective, the ever improving algorithms classifed as i) clustering, ii) supervised, iii)
unsupervised, iv) fuzzy logic, v) neural networks, are efficient relays of immense computational
power of today and are used for getting the most predictive outcome for the future, based on the
postulation of interest.
Conclusion:

Although, in this work we are not describing a definite data mining formula, an exact algorithm, we
are confidently describing a novel means of postulation analysis based on i) current defined
mathematical formula and constants, ii) text co-occurrence frequencies as well as their threshold
values for decision making and iii) mathematical musical analysis. .

The details of mathematical formula mining can easily be implemented and further build by the
expert mathematicians. The dynamics of text co-occurrence based context mining is well studied in
Google and forms the basis of Google Translate (1). Google can take that method one step further by
upgrading “term co-occurrence statistics thresholds" for better predictive outcomes. Musicians in
the data mining field can implement the basis and improve their algorithms in real time (3).

In fact, these algorithms are advancing at an unprecedent speed. We are just putting forward a
frame of hypothesis testing and learning for Artificial Intelligence (AI) by coordinating the ever
advancing algorithms under a defined “if function”.

In short, the described system of postulation learning using an “if function” is a realtime learning
approach for Artificial İntelligence (AI), recording all the verified and unverified information
regarding the outcomes of any hypotheses in the database for future reference. Once the novel, real
information (reality) comes in, the system can autonomously fine-tune the decision making threshold
values for "term co-occurrences" for more accurate predictions for the future. The fine-tunning step
for decision making treshold values can either be i) autonomous or else ii) involves advanced settings
as self writing codes. As said before, we are letting Artificial Intelligence (AI) Algorithms dream about
a postulation of interest and remember every statistics of the dream to dissect the reality part out of
the dream for further learning.
We firmly believe that when; i) an Artificial Intelligence (AI) Algorithm independently asks the
question “what would happen if robots were humans?", ii) analyzes the results on the database with
updated threshold values of "term co-occurrences", iii) records the results as verified or unverified
for the future reference and iv) recalls these results in the near future, they reach a certain level of
self awareness and start to having a basic consciousness as an important step towards Artificial
General intelligence (AGI) away from Artificial Narrow Intelligence (ANI) of today.

References:

1. Google Translate

2. www.youtube/watch?v= XsZbGxlmc6c

3. www.youtube.com/watch?vv=tLDSE28QAG0

Você também pode gostar