Abstract
Cognitive, ˈkɒɡnɪtɪvˈ, is
derived from the word, cognition. Cognate means being relative and continuity
is the cognition. As humans, we have an added advantage of self-learning.
Although, the concept of learning is used as a synonym for understanding (or)
interpreting by individual thru perception, verbal, semantic or syntactical
procedure. On the other side, machines, being brain less, need program (or)
instruction towards learning. Thus, the need of cognitive programming.
Jean Piaget, is the
popular psychologist known for his theory on cognitive development. His theory
is based on the learning principles of kids, though. The same is suitable for
all programming the machines as well. This can be considered as the 3rd
generation of computing. In the beginning we have abacus model with the help of
tabular systems. Accounting is the best example of this model. Then came the
programming to automate these models for a shorter and quicker intervals. All
programming languages fall under the 2nd generation of the
computing. Now that the maturity of this model is at a stable state, it is time
to induce the intelligence to these programming models, such that, they become
suitable for self-awareness and knowing becomes a practical approach for
learning on the happening around the past and present. Hence, cognitive
development is an approach for programming these machines.
Problem
Two decades back, storing
the data was a challenge, but, not now. With the advent technology, within the
current world, we can store a TB of data in a one centimeter hologram. It is
not far from now that the world will see Zettabyte as well as Yottabyte sized
data. The problem is not in storing the data, but, in retrieving the same
within a given interval. It is calculated by the scientists that an average
brain of human can store 3 terabytes of information. It is understood that
anything between 1 to 10 TBs of data can be accommodated. With such high
quantity of data, it is found that most of the humans find challenge in
recollecting the past situations at any given situation.
The processing power of
the human starts after the ½ second on instruction. For example, we met a
person and we wanted to recollect who that person is, when we have last met,
what are the details of the person, etc., When the brain gets instructions to
recollect, it starts functioning only after 300 mills seconds to 500 milli
seconds, which is about ½ of a second. It is takes less than 50 milli seconds
for a machine.
But, it is surprised to
see that both human and machine struggle to retrieve data from the archived
sources between 2 seconds to 5 seconds. The battle is won by machine after 5
seconds, whereas humans, if fail to recollect the information with 5 seconds,
they will continue to fail until 17 mins, at times, they will never as well.
For humans, it is proved that if they don’t recollect within 5 seconds, their
failure rate in recollecting will deteriorate with a faster pace and then their
brains ends in dead lock. The data in the current world is doubling for every
12 to 18 months. And the truth is that this data is unstructured.
So the need of the current
situation is the need of the better technologies and matured systems to adapt
the changes that are not programmed and result with better actionable
decisions. Thus, the need of the situation is to have a nature relation between
the human and machine within any given domain and be able to advice services in
that domain. This system must be in a position to motivate on own performance
and relates other same situations. This help to acquire the data for difference
situations and generate the probabilistic memory solution hidden from the
unstructured knowledge.
Most of the industry
leaders are now providing their research on this field by means of APIs and
giving access to their neuromorphic architectures. These APIs will help various
industry leaders to take advantage of the highly complex neuron model based
systems to predict their business behavior. These APIs also help as a
functional equivalence to natural cognitive process with the training provided
from various consumers. The learnings from the consumers will be useful with
wide sharing, thus, one industry adaption of these APIs will help the other
industry as well.
The entire foundation for
this kind of programming should be the deduction of the information, rather
than the deterministic system. This can only be achieved if the reflection of
the data is inferenced with experience based learning. The entire process
should be based on the nature language interaction with the perception process.
These technology programming models should be self-reinforced systems. It is
too hard to predict the natural and intuitive behavior towards each problem or
situation. It needs deep reasoning and judge the reasoning with the evidence
that is found from the augmented data along with the augment of human
perception.
It is all about making
patterns and port them for the given situation. The challenge in these
programming models is sequencing the natural processing in the context of the
collaboration human with advanced machinery. Personification of the
business from the context of the user is the success of any cognitive programming. The learning of these machines would be obsolete over
the period of 5 to 8 years. Thus, the shifting of the learnings of the archaic
machines has to be shifted to newer infrastructure. Not only should the
hardware of the future systems change, but the software as well. It is also a
challenge towards this technology shift is about the privacy of individual
context is applied on the similar other contexts. The context of perceptual
intelligence is the way to populate the data from the unstructured source to a
variety of presentation mechanisms.
Industry leaders
Microsoft Cortana
Microsoft’s cognitive computing services are formerly known
as project Oxford. They named these services as Cortana and is reachable at https://www.microsoft.com/cognitive-services.
Microsoft Cognitive Services expands on Microsoft’s evolving portfolio of
machine learning APIs and enables developers to easily add intelligent features
– such as emotion and video detection; facial, speech and vision recognition;
and speech and language understanding – into their applications. Cortana’s
vision is for more personal computing experiences and enhanced productivity
aided by systems that increasingly can see, hear, speak, understand and even
begin to reason.
Cortana comes with natural language process with multi model
interfaces. It serves with all 3 styles of interaction with human. It supports
speech recognition, vision recognition as well as language understanding with
grammar to colloquial. Its neural networks can learn through training and can
identify difficulties.
IBM Watson
IBM Watson is a technology platform that uses natural language processing and machine learning to reveal insights from large amounts of unstructured data. Watson is initially put into learning mode with various phases. In the beginning of inducing the cognitive intelligence in any field, Watson is betrothed with the digital data in all varieties. The second phase is to sanitize the data with a human interacted language, in the form of spoken English. This phase is put into practice with a means of questions and answers. During this phase, Watson learns the jargon as well as the pattern of the questions.
As this is progressing in conjunction with the digitization
of the data, there is a need for proofing the data that is accumulated from
voice as well as text. At this juncture, Human intervention is highly important
and the brightest mind helps Watson to distinguish the unwanted information as
well as highly classified points.
IBM states that the era of computing has evolved from
tabulating systems of 1900s to Programmable systems era by 1950s. Later that,
all the advancement of the machinery made lot of changes into the programmable
system and finally IBM could reach to the current state as cognitive systems
with the result of Watson. After this, they envision to cognitive computing
(or) brain cube as anticipated by 2020
Google’s Deepmind
Google’s acquisition of DeepMind’s AI platform, http://www.deepmind.com, it made history as
first computer program to beat a professional player at the game of GO, which
is a googol times more complex than chess. Demis Hassabis, CEO of Deepmind,
criticizes that the other vendors in this field are just as a ‘Narrow’ AI, Narrow Artificial
Intelligence and they are designed for one and only one purpose. Whereas,
Deepmind’s vision is to build an AGI, Artificial ‘General’ Intelligence. Demis also thinks that, their vision is to
convert the unstructured information into actionable knowledge. Their vision is
to build machines such that they become general purpose learning machines.
These systems should learn automatically from raw inputs and are not
pre-programmed. These systems are made for general, such that, the same system
can operate across of a wide range of tasks.
Conclusion
Understanding the systems neuroscience and making the
machine to adapt (or) learn from the data feed, it if highly difficult to have
a set of programming lines to remain to function as time progresses. They have
to learn directly from their experiences and without any retraining.
---------------------------------------
This is just out of my collection, if you have any objections, drop me a mail at dskcheck@gmail.com
No comments:
Post a Comment