What is AI ? Breakthroughs that paved way to the bright future of AI

What is Artificial Intelligence?

Area of Computer Science which simulates intelligence in machines and systems, similar to Natural Intelligence in humans is known as Artificial Intelligence. Intelligent agents learn, plan, solve problems with the valid reasoning, accumulate knowledge and perception to gain decision-making capabilities of their own. AI is omnipresent in today’s world from recommending you the next thing to watch on Netflix, virtual assistants Alexa and Siri performing simple day to day tasks to Fraud Detection in financial markets.

History of Artificial Intelligence

Visualising the AI Dream / Chimera of AI

For centuries, renowned philosophers have dreamt of machines that have a mind of their own. In 1872, Author Samuel Butler in his fictional novel, ‘Erewhon’, wrote about a machine that has a consciousness of its own. Fictional novels in the 19th century like ‘Frankenstein’ and ‘Darwin among the Machines’ described intelligent machines that decided like humans.

In 1920s and 30s, David Hilbert asked the top mathematicians a fundamental question: ‘Can all mathematical reasoning be formalized?’ This was later answered by Godel’s incompleteness proof, Turing’s machine and Church’s Lambda calculus.

1939-1945, the Second World War used modern computers as massive code breaking machines. During the same time, Norbert Wiener laid the foundation to Cybernetics and Neural Networks with ‘Feedback Theory’. He theorized that intelligent behavior was a result of feedback mechanisms based on his research on feedback loops. He argued that feedbacks could be simulated to make machines intelligent.

Birth of AI, 1952-56

During this time, scientists and researchers from various areas discussed the possibility of a machine replicating a human brain.

In 1950, Alan Turing published his historic paper which explored the possibility of creating intelligent machines with true intelligence.

The famous ‘Turing Test’ went on to become a foundational proposal in the philosophy of Artificial Intelligence. It stated that, if the machine was capable of carrying out a conversation over a teletype, which was indistinguishable from a human conversation, the machine is called ‘intelligent’.

In 1955, Allen Newell and Herbert Simon developed the first AI program, ‘Logic Theorist’. It followed a tree model where the most accurate conclusion for the problem statement was reached selecting branches.

In 1956, John McCarthy along with other senior scientists organized Dartmouth Conference in Vermont. The term ‘Artificial Intelligence’ was coined at this conference and the researchers defined the mission and celebrated the early successes in the field of AI.

Budding years of AI, 1956-1974

During this time, computers were capable of solving algebra word problems, prove geometry theorems and learning to speak English.

In cases where the possible paths for a problem just too large, scientists reduced the search space by using ‘Rules of thumb’ or heuristics. This would drop the paths unlikely to lead to a solution. Newell and Simon captured the general version of the algorithm called ‘General Problem Solver’.

Stanford developed the STRIP system, which searched through the goals and sub-goals to plan actions and control the behavior of robot Shakey subsequently.

Semantic networks which represent concepts/entities as nodes and relations among them as links between nodes was used in many AI programs. MIT AI Laboratory researchers, Marvin Minsky and Seymour Papert suggested using situations like micro-worlds which are artificially simple situations.

Defense Advanced Research Projects Agency provided three million dollars every year to AI research. Newell and Simon’s projects at CMU and Stanford AI project of John McCarthy received similar grants.

AI Winter, 1974-1980

In the 70’s, AI scientists had raised the expectations very high and when their ideas failed to materialize, funding for AI and Neural Networks was affected. The problems faced by the AI researchers were :

The processing power and memory of the systems were not powerful enough to boost results. For instance, Ross Quillian’s work on NLP used only 20 words due to low memory on the system.
Many of the trivial solutions developed failed to scale up and create useful systems.
AI applications required a huge amount of data for the program to learn and pick patterns from. Building such large databases in the 1970s was an impossible task.

As a result, government backed agencies like DARPA, NRC, ALPAC along with the top universities stopped funding for most of AI specific research.

Revival of AI, 1980-1987

In the 1980s, corporations focussed on gaining AI related knowledge to create Expert Systems in future. During the same time, the Government of Japan invested in the fifth generation computer project.

In 1982, John Hopfield proved that ‘Hopfield Net’, a neural network was capable of processing information and learn in a new way. During this time, David Rumelhart’s new way to train neural networks, ‘backpropagation’, gained momentum.

Second AI Winter, 1987-1993

Researchers saw another AI winter where many AI projects lost funding.

Hans Moravec and Rodney Brooks proposed an idea suggesting that intelligent behavior machines need a body. This laid the foundation for Robotics.

Resurgence of AI, 1993-Present

In 1996, IBM’s ‘Deep Blue’ defeated the world chess champion, Gary Kasparov. Around the same time, Sony unveiled AIBO robot pets and Honda introduces ASIMO, world’s most advanced humanoid robot.

In 2011, Virtual Assistants likes Siri, Now and Cortana emerged from the tech leaders Apple, Google, and Microsoft.
In the same year, IBM Watson, an AI system developed by IBM defeated Brad Rutter, the biggest winner of all time with a streak of 75 wins, in the game of ‘Jeopardy’.

In 2016, IBM ROSS became the first AI Attorney in the world, landing a job in a New York-based law firm.

Speech recognition, information about landmarks and points of interest are automatically marked on the cars GPS and various new inclusions are making the smart car experience better.

Manufacturing and Medical Sciences extensively uses robotics. Major automobile manufacturers are testing out their self driving cars. In a few years time, self driving cars may be a reality on the highways. Japan is planning to send a humanoid robot to the moon.

US military is developing robots to replace or help soldiers on the front line with minimum or no supervision. Flight simulations and virtual environments help train over 50000 soldiers.

As you can see, Artificial Intelligence is not new. It has been around for more than 50 years. But we are making advances at an exponential pace in the last two decades. It is progressing fast by leveraging improved processing power, terabytes of memory, explosion of digital data and the speed of communication infrastructure. In some ways, the commercialization era of AI has kicked off recently and it will profoundly affect the world as we know it, more profoundly than the ways internet and mobile phones have done so before.