I watched the movie “Terminator 2: judgment day” in 1991 in the era of video cassettes. Though being a ten-year-old kid and completely unaware of the scientific jargons of machine learning, computer vision and artificial intelligence, I was so fascinated by the concept that at times I began to consider myself a cyborg. The movie presents the idea of a self-aware, artificial intelligence based computer (Skynet) that spreads into millions of computers all over the world, finally outperforming human abilities and wisdom and taking over the world. The humans are left struggling for their survival. Skynet sends a terminator back in time to destroy the leader of the human resistance while he is still a boy. A second protector (Arnold Schwarzenegger) is sent back to the past by the human resistance to protect John Connor, their future leader. The cyborgs contain a neural net learning system with the abilities of self-awareness, learning from environment, adaptationto changing situations, arbitrarily interpreting the non-formalised tasks, speaking naturally, copying the voices of others, and understanding human handwriting, etc.
It was a sensational science fiction, though. Nobody ever thought about the potential of this fiction becoming reality.
Imagine a future where machines will be solving our routine problems, our smart phones will be suggesting diets and medication based on our health, our cars will be driving without human interaction, business consultants will be replaced by a machine learning based software consultant, people will no longer be hiring legal consultants as legal advices will be given by machines, millions of people will lose jobs as their tasks will be taken over by more efficient and intelligent machines, we will be using an intelligent software to plan our future and even to select our life partner. All this is not very far.
The idea of self-learning machines dates back to the invention of computing technology. Though the computing power of machines outperform humans’ computing abilities, computers largely depend on their programming and users’ input and are not able to make their own decisions. Machine learning was hindered by the huge computing power required to teach machines self-awareness, learning patterns from input data and making their own decisions. It was inspired by human learning which is based on experience and observation, with the only difference that machine learning takes all the possible observations simultaneously and tries to learn the underlying patterns in a much shorter time.
It wasn’t until the advancements in computing technology and development of more efficient machine learning algorithms when machine learning was given a serious thought to teach self-awareness and patterns from observations in the same way as humans do. We witnessed a major breakthrough in machine learning in 1997 when IBM chess computer Deep Blue beat the Russian grandmaster Garry Kasparov.
The continuous developments in the field of machine learning have transformed it to a much mature field and have led to rapid uptake by the tech industry’s largest companies, including Google, Facebook, Microsoft and IBM who recognised the power and future potential of machine learning and artificial intelligence long ago. A type of machine learning, the deep learning, is becoming increasingly popular these days and is considered to be the future of artificial intelligence. The difference between deep learning and other traditional machine learning algorithms is that deep learning is capable of learning patterns from raw data, adjusting its decisions on the fly, and improving its knowledge and capabilities over the course of time. By analysing vast amounts of digital data, deep learning can learn all sorts of useful tasks. In some cases, it can learn a task so well that it outperforms humans. It can do it better, faster and at a much larger scale.
Google recently purchased a deep learning based framework, called DeepMind, from a British company and has become a leader of machine learning innovation since then. In March 2016, machine learning took a giant leap when Google’s Go-playing software, Alpha Go,beat the world-class player Lee Sedol to win the five-game series 4-1 overall. The ancient Chinese board game long considered impossible for computers to play at a world-class level due to the presumed level of intuition required and is considered to have more positions than the total number of atoms in the universe. This is a huge success. The same technology can well be applied to different walks of life.
We use a number of applications based on deep learning and just take them for granted. Some of them worth mentioning include Google’s context-based and image-based search, speech recognition on Google’s android phones, automatic image tagging and understanding visual translation of languages, natural language processing, face recognition and automatic tagging in Facebook images, and recommendation systems, to name a few.
Google has open-sourced tools to allow the community to adapt them for their own uses, and improve them. It is just a matter of time that we will be able to see machines thinking, seeing and making decisions with much more efficiency than humans. The next couple of decades are going to be crucial and challenging for all the technology related companies to adapt this paradigm in order to survive. Those who lag behind will be simply wiped out by the storm of the technological advancements brought about by machine intelligence.
The idea of machines surpassing human intelligence and abilities is still far-fetched, though not beyond possibility. Some scientists do point to a technological singularity we are heading to, which will result in an intelligence explosion, resulting in a powerful super-intelligence whose cognitive abilities could be far above humans’. The current growth of machine intelligence apparently seems to be more rapid than the evolution of human brain. If the progress in machine learning keeps accelerating along with the increasing power of computers and other related technologies, we may witness machines in future which will outperform human abilities. At that time, we will have to address the ethical and social issues pertaining to machine intelligence and their conflict with human values.