AI, Machine Learning, Deep Learning? XSENSE
ALTAMONTE SPRINGS, Florida – January 10th, 2017
XSENSE wants to put some order
into the world of AI.
Artificial Intelligence, Machine Learning and Deep Learning are three very hot topics right now, trending both on social networks and our daily lives.
They sound like synonyms and seem interchangeable. Is that true?
We frequently hear these concepts when we talk about , Analytics, and the technological changes better known as Industry 4.0.
Let’s make a first, short, statement:
- Artificial Intelligence is the broader concept of machines able to carry out tasks faster, smarter and more proficient than a human being
- Machine Learning is the application of Artificial Intelligence in a way that a machine is able to learn from a given training set. Allow machines to access data and they will learn for themselves
- Deep Learning is a branch of machine learning based on a set of algorithms that brings Machine Learning closer than ever to AI. Deep Learning enables the use of large neural networks to learn multiple levels of representation and abstraction to make sense of a huge amount of data such as images, sound, and text
Machine Learning, somehow,
emerged from Artificial Intelligence. Why?
There are two main reasons that led us to the rise of Machine and Deep Learning.
Number one: a pioneer in the field of computer gaming and artificial intelligenceCheckers-playing Program in 1959 that is the world's first self-learning program, probably the first demonstration of the fundamentals of artificial intelligenceHe stated that rather than teaching computers everything they need to know about the world and how to carry out tasks, we should better teach them how to learn for themselves.
The second and most recent
reason is the rise of the Internet, and the huge increase of the amount of
information being generated, stored, and made available for analysis.
A Neural Network is a computer system that classifies information like humans do. It can be taught to recognize, for example, images, and classify them according to elements they contain.
They are essentially based on a system of probability: thanks to data that were previously inserted, it is able to make statements, decisions or predictions with a high percentage of certainty. The addition of constant feedbacks enables its learning by sensing or being told whether its decisions are right or wrong, modifying the approach it takes in the future.
Here comes XSENSE
Our customers frequently ask
which algorithms XSENSE is based on.
This is not obvious, XSENSE is not inspired by anyone.
Let’s make some order:
1. Content Identification
Autonomous identification of arguments and subjects. Language
independence even on an empty database
b. The manual creation of a notion, implies the autonomous analysis of subjects and language. If these are not present yet, XSENSE will create and associate them to similar notions
c. If the user deletes arguments not reputed important, XSENSE will not give importance to them as well during next imports
2. Conceptual Networks
a. During the import of a
notion with one or more subjects, XSENSE looks for similar notions already
present in the database, creating relationships between them.
The more are imported, the more the subjects identification becomes accurate
b. When new notions are learnt, words that have been used are analized and associated to arguments. XSENSE is the ultimate problem solver when it comes to words disambiguation. A word used in different contexts, has different meanings
c. If two or more arguments, that were once separated, become synonyms, XSENSE relates them to a unique argument and consequently updates the knowledge base
d. If new subjects are
manually created, XSENSE looks for similar notions already present in the
database to create proper relationships
3. Dynamic Content Enrichment
a. Arguments might be enriched with notions from different sources. Remember to specify them (eg: Wikipedia)
b. These new arguments are automatically related to similar ones, in agreement with the user
notions might have different priorities if compared to imported or created notions.
The level of priority is set by the system integrator, it determines the way
XSENSE answers to questions
4. Logical Reading Simulation and Learning Data in a Structured Format
a. Mind Discovery allows the creation of
empty conceptual structures (es: main title , paragraph, subparagraph) that
will be filled by XSENSE, creating new data structures through massive imports
of documents and html pages.
Imported notions have the same reading logic.
Example: in a notion about the price of a service, the name of that service (specified in the title or in the paragraph) is automatically related to that price.
b. This data structure will be
replicated everytime that structure is entirely valorized. For each document
might exist several groups of notions with the same logical structure
5. Unalgorithm (adaptive AI algorithms)
UNXFLOW is a logical entity, based on arguments and notions, which allows XSENSE to create a conversational workflow, not predefined, in runtime mode
a. UNXFLOW is created by Mind Discovery and implies the definition of one or more variables (mandatory or not mandatory) that may assume a value with manual or massive imports (Logical Reading Simulation and Learning Data in a Structured Format)
b. If the system integrator sets a variable as "Mandatory" XSENSE asks him to edit an in-depth question to be asked in case that variable will be not correctly set with the first question
c. UNXFLOW is generated by the user’s questions, XSENSE will then ask in-depth questions to identify a unique, proper, answer
d. L'UNXFLOW has not any predefined steps. This allows the user to answer XSENSE with a new question or completely change argument. You can even call back a UNXFLOW previously opened
e. XSENSE performs a constant analysis of imported notions. If many of them are similar, XSENSE will suggest the creation of a precompiled UNXFLOW.
6. Dialogic Analysis
a. Maintenance of the context(s) of the conversation
b. Identification of a complete change of argument
c. Identification of a partial change of argument
d. Elaboration of answers in relation with argument, relevance and time frame
e. Elaboration of different
answers depending on the interlocutor
7. Deep Learning
XSENSE learns from conversations with users. It may ask them about the meaning of a word that it doesn't know yet. It learns new notions and may suggest the Trainer to apply some modifies to the Knowledge Base, even in relation with how the user reacts to XSENSE's answers. Each new notion is contextualized into a domain in complete autonomy, with no need of external intervention.
8. Subjects Hierarchy and Synonyms Identification
Thanks to UNXFLOW, XSENSE is
able to understand if two arguments are synonyms or in a hierarchical
relationship. For example, just by reading texts (written in any language) XSENSE
knows that the cat is related to the filine world and viceversa.
If we talk about an institutional role, XSENSE associates it with a person and, as a consequence, with a name.
Here i show it works:
Input A) The President of The
United States of America is Donald Trump
Input B) The President of The United States of America is married
User's request: is Donald
XSENSE Answer: The President of The United States of America is married
All the algorithms are language independent because they simulate the cognitive process of the human being in learning languages.
In 2017, XSENSE will be able to understand programming languages and learn it dynamically, in order to autonomously develop software and its own code.
Machine Learning, Deep Learning, Neural Networks.
We have guided you into some main concepts of the fourth Industrial Revolution.
Will you ride with XSENSE?