The How’s of our Machines answering like a Human.
The concept of Artificial Intelligence has been spreading its paws over each and every industry since the dawn of 1950s. Cognitive scientist Marvin Minsky coined the term “AI” and felt optimistic about its wide applications, growth and importance in our lives — possibly while sitting in his lab at the Dartmouth College. Prior to his predictions, a young British polymath named Alan Turing explored the mathematical possibility of artificial intelligence.
Turing suggested that humans use available information as well as reason in order to solve problems and make decisions, so why can’t machines do the same thing? — a logical and arguable question, which then led to scientists, mathematicians, statisticians, etc across the globe scratching their minds and exploring pages of Turing’s “Computing Machinery and Intelligence” — published in 1950.
What halted Turing and his prognosis about Machines predicting our meal choices for lunch then, basically to work on his thesis further? As per an article published by Harward edu on Machine Learning —
First, computers needed to fundamentally change. Before 1949 computers lacked a key prerequisite for intelligence: they couldn’t store commands, only execute them. In other words, computers could be told what to do but couldn’t remember what they did. Second, computing was extremely expensive. In the early 1950s, the cost of leasing a computer ran up to $200,000 a month.
Right from coining the term AI in 1950s while struggling to get hold of a computer system — to — it helping scientists and virologists in 2020 in researching and analyzing vast amount of data about the coronavirus pandemic — AI has come a long way into the modern world.
But what really is AI? and how does it work? — in the most simplistic terms — AI leverages self learning systems by using multiple tools like data mining, pattern recognition and natural language processing. Ok, these terms are simple, but the HOW is still far away from being answered. To get to the point of how really pattern recognition and learning works, one needs to understand how the human brain processes data, learns and remembers millions of GB of information, each second. AI operates similar to how a normal human brain functions during regular tasks like common-sense reasoning, forming an opinion or social behavior.
To understand the HOWs of AI or Machine learning, let us understand and decode the meaning of Artificial Intelligence first:
- AI is the capability of a machine to imitate human behavior.
- It is basically intelligence exhibited by machines — by processing input data provided with the help of complex mathematical calculations and processes to produce a refined output which can be easily understood by humans.
- Remember AI augments or imitates or mimics cognitive functions exhibited by human and not replaces it.
- The entire idea of AI is based on the premise that intelligence is not real or human.
Now let us understand the processes involved in Machine Learning —
- Machine learning leverages existing data (it maybe in the form of images, videos, calculations, processes, actions etc.) to train its algorithms and models. These models or algorithms are mathematical calculations fed to the machine learning code or system.
- Numerous sets of examples, training sets, data is fed into the system. More larger training set, more accurate the algorithm would be. This feature works eerily similar to the process of how humans learning or remember a certain data. When we are bombarded with numerous data with similar patterns or samples, we tend to remember it and possibly can reproduce the data after a certain period of time.
For example: Learning to play a musical instrument involves hours of practice and “Riyaz” which involves understanding the chords, practicing the ways of playing and producing desirable sounds. When subjected to hours of practice and same chords over and over for a period of time, we tend to remember and can produce the music without any mistake in future attempts. But for humans, if the practice or feeding of information is reduced or stopped, we may partially or entirely forget the piece of information, this fortunately does not happen in case of machines, which are fed with the mathematical calculations and processes to imitate data efficiently.
3. This process of practicing or training data to effectively imitate it in future relates back to the way human brain works. The idea of machine learning dates back and is similar to one of the best understood concept of neurophysiology, “myelination”. Myelin is regarded as static, inert insulation on axons.
A systems-level perspective on learning suggests how myelin might contribute a non-synaptic cellular mechanism of memory. Complex tasks — such as learning to ride a bicycle and evoking rich memories of past experiences with proper context, place, sequence, sights, sounds, smells, emotions and appropriate links to other stored memories — must require coupling between neuron populations from many different brain regions.
4. Each item in the training set to the AI is labelled either 1 or 0. The ML process is divided into two steps: Training and Testing.
Training Phase: In the training phase labelled training data is fed to ML algorithm, which is basically a combination of “Input data + Expected Output”. Machine Learning algorithm studies the data patterns and works out a logic based on the input and output and in the end a ML output is derived, which is a learned model. This learned model can then be used with test data sets.
Testing Phase: Test data contains input data and it generates output based on logic derived from training data. The system classifies test data based on the patterns learnt from training data. This uses patterns from test data along with logic of learned model to make predictions and desired output.
Patterns of test data + Logic learned model = Prediction and derive Output.
Each time an AI system runs a round of data processing, it tests and measures its own performance and develops additional expertise. Because AI never needs a break, it can run through hundreds, thousands, or even millions of tasks extremely quickly, learning a great deal in very little time, and becoming extremely capable at whatever it’s being trained to accomplish.
References:
