EVOLUTION OF AI
It’s the golden age of artificial intelligence and what we once thought was impossible is now pushing new limits.
There is so much to discover and so much that is already within our reach. From customizing our playlist for us to performing operations, they have become more critically involved than ever so let’s look back at how they evolved to become the indispensable part they are today.
The history of it all...
The idea of an Artificial Intelligence existed for a time longer than we can imagine with olden myths and fictional stories having entities that could imitate human nature and was proposed in the 1930s to the scientific community in an executable manner. That, someone, was the great mathematician, Alan Turing, who had a simple idea with an infinite scope of expansion. The idea was that if humans can execute day to day activities with a pre-learned set of knowledge then so can machines!
It was from that moment onwards that scientists from a variety of fields sought to mechanize human thoughts. The American-Hungarian mathematician John von Neumann worked with Turing to establish that idea through electrical machines. Neumann proceeded to generate the concept of cellular automata which is widely used as the basis of artificial life today.
There were many naysayers due to the lack of hard proof but around 70 years later and we sure have made those involved proud.
Turing then introduced the Turing test in 1950 by publishing it on his paper and to this date the Turing test is used as the benchmark test of how close to replicating the human mind a machine is.
Known as the Imitation Game, it involves secluding a man and woman from an interrogator who has to guess which is which by asking questions and guessing from the written replies. The man aims to fool the interrogator, while the woman tries to help him.
The imitation part in the Turing Test comes when a computer program replaces one of the participants, either the man or the woman. Thus, the test observes whether the interrogator can determine which is a computer and which is a human. The idea was that if the questioner could not tell the difference between human and machine, the computer would be considered to be thinking and could imitate human behavior.
For a long time, no machines were able to pass the Turing test but in recent times some machines have successfully passed the Turing test and it is questioned because the environment and time factors always seem to act as an additional factor in these tests. Which means every test when conducted could have different criteria and levels. Furthermore advanced AIs are said to have a disadvantage during the Turing test as the machine is not required to do any logical or anything that requires smart thinking, they just have to pass as humans. Here a question that hits everyone is, Is it time to change our long set standardized test? To which many scientists reply with ‘that function apart, for an AI to completely integrate with our society, no matter how smart or miraculous its features are it ultimately has to pass on as a human because that is how they will perfectly fit in our lives’. Even with the different factors the Turing test still helps understand how smoothly the AI functions, which at any date would be an important thing for machines.
Even with all the advancements, the lack of funding and people involved in research was holding the potential of it all back. During the last quarter of the 20th century, AI garnered lots of attention with it being seen as a potential part of businesses and replacing human work which was dangerous.
As we move forward we can now see AIs evolving at a speed faster than ever but what is even more amazing is that AIs have been seen to be evolving themselves as well.
We all know neural networks, which is an integral part of developing an AI, loosely mimics the structure of the brain, and learns from training data by altering the strength of connections between artificial neurons. Scientists have sought to speed up the process and have done so by automating some steps but these systems still stitched already made circuits to create a function that limited its expansion.
To overcome this issue, softwares like AutoML were developed where AI programs could be developed with nearly zero human input using just school level knowledge. It randomly collects and combines algorithms to test out whether a task is possible or not. If not, it moves to another combination which creates a cycle of different algorithms, and all the while it compares to some human-designed codes and mixes and mutates all these codes to arrive at an answer. It exchanges parts of algorithms or adds in some part to avoid dead-end or any other problems. While this is not a fully developed system, it has a lot of scope in the future. Adding basic algorithms or functions could lead to faster solutions for more complex and larger problems. Starting with small problems and then moving on to bigger ones is the approach that scientists have sought, and hoping all goes well we will soon have state of the art AI features.
Scientists back in the day might have had bigger better hopes for what AI would have been today but something this important has to be devised with patience and careful consideration. With all that we have seen, a future with both humans and machines now seems possible.
