The Building Blocks of AI: Neural Networks and Machine Learning

AI today is a topic of conversation, a subject for corporate presentations, a space for technological debates; in general, a term uttered by many who are familiar with a machine. AI is a concept that is being adopted by many sectors; indeed the latest and hottest buzz word in the world of technology.

However, the fact is Artificial Intelligence has remained in the nascent stage of development for quite a while. When one hears the word `artificial intelligence’ the visuals of different characters from science fiction come to mind. What was a mythical concept of time travel or spacecraft with extra-terrestrial creatures is not considered a futuristic science anymore? Though still primitive in form, AI can solve unique business problems, provided one knows and decides how to leverage the technology to its potential.

James Goodnight, Co-Founder and CEO of SAS says: `AI is certainly being hyped as another buzzword, and there are many companies rushing to jump on the AI bandwagon’. Despite the hype, AI is a significant emerging technology that will continue to grow in importance. Years ago the term `cloning’ became very popular; and in course of time, was included in the general context and usage, where anything which is replicated was described as `cloned’. Dolly, the sheep (the first mammal cloned from an adult somatic cell) became a widely talked about subject. Along the same lines, people have started using AI to describe any analytics problem, be it faulty equipment or credit card fraud. The catchword now is AI, and Sophia, the Robot has replaced Dolly sheep. With everyone using the buzzword AI, one needs to see whether reality aligns with this hype.

There is so much information on the internet stored in the last few years. Wall Street Journal recently stated that the internet is filling up with good morning messages, and largely by Indians, many of whom are yet new users of the internet! On a more serious note, there are huge quantities of data collected and stored in the recent times. Almost all information is online and accessible to net browsers. A tool to statistically filter, analyze and process large data sets, not merely as a process but with intelligence, is the primary need of industries, investors, for that matter, any establishment across sectors and geographies. A growing technology like AI can unlock the value of data, and eventually help transform businesses, industries, and above all, people’s experiences. It is no surprise that AI being such a powerful concept has become not just a household term, but a word across the technology world.

Though the research in Artificial Intelligence dates back to the 1940s, it is only around the 1980s that AI progressed from research to application. In recent years, however, interest in the field of artificial intelligence has sky-rocketed as evidenced by the amount of funding invested in researching and adopting it.

To go back to sci-fi, we have seen in many robots featuring stories about how the machine ultimately overpowers the man to the point of bringing about total annihilation. This threat could be as real as the creation of artificially intelligent robots. Well, known personalities like physicist Stephen Hawking and Tesla CEO Elon Musk have been stressing on the importance of regulating artificial intelligence.

There is merit in the thought that AI technology will help society in a range of areas like transportation, education, health, manufacturing, retail, arts, medicine and so on. It can be a danger though if used to build weaponry. Leaders across the world have found this worrisome, even terrifying.

One has to admit, turning a buzzword into business is exciting, but like communications specialist Lindsay Kwan says: The danger of artificial intelligence is that we can both overestimate its power today and underestimate its capabilities tomorrow.

Which, between the two, is riskier, is a matter of conjecture.

In the mid-eighties, Small Wonder, a sci-fi TV series entertained adult and child alike. The story revolved around an android in the form of a 10-year-old girl. Looking back, the series was indeed very amateurish and loosely structured. The point is it introduced for the first time, the machine made man to the lay person. Thirty odd years later we hear of Saudi Arabia’s decision to grant citizenship to a robot. Endowed with artificial intelligence, visual data processing and facial recognition, the robot, Sophia is designed to get smarter over time.  

The citizenship as such is of hardly any consequence; what it signifies is that Artificial Intelligence (AI) has come of age. Critics have spoken at length about the absurdity of awarding citizenship and an array of rights to a robot, while Saudi ironically is yet highly restrictive about women’s rights. However, many see this as a positive move towards making the country tech savvy and progressive.

In the computing history, man and machine interfaces have only improved and enhanced over a period of time – hence both are not mutually exclusive. Satya Nadella, CEO of Microsoft, said AI, mixed reality, and quantum computing will shape the world in the coming years. The crux of the matter is that the technology world will have a blend of the real and virtual. While current technology enables in us the ability to collect data, Nadella referring to AI says, what is more important is to reason over the data to create intelligence. This is most critical as a differentiator between mechanical equipment and an AI enabled robot. In the mid-fifties the term GIGO (Garbage In Garbage Out) became a descriptive term for computers, primarily because the machines lacked the ability to think for themselves. Whatever output emerged was only as accurate as the information entered. Deep down, computers still are improvised calculators. Recent development in AI comes under the aegis of `machine learning’ and involves programming computers to learn from example data or past experiences and process information based on that.

So how does machine learning impact tasks and get independent of human influence (in terms of execution of task)? Let’s take an example: Automatic transmission in cars has been prevalent for quite some time and in India, they are getting popular by the day. With increasing traffic bottlenecks people are finding it more convenient to drive a car which does not require changing gears. The driver needs to navigate by keeping the steering in control. With AI, autonomous driving, a wide step ahead of automatic transmission, the technology required for a car to accelerate, turn, and brake on real time roads without human intervention is not anymore impossible or far into the future. The recognition that autonomous driving was a prediction problem solvable with machine learning meant that autonomous vehicles could start to become a reality in the marketplace years earlier than has been anticipated.

By breaking down tasks into components, we can begin to see how AI will affect the workplace. AI is usually framed in terms of machines versus humans, but gradually we see it more in terms of understanding the level of judgment necessary to pursue actions. In cases where whole decisions can be clearly defined with an algorithm (for example, image recognition and autonomous driving), we expect to see computers replace humans. This will take longer in areas where judgment cannot be easily described – where there is high level of ambiguity in arriving at decisions or solutions. In other words, where the task is not a mere process with anticipated outcome and the pros and cons fail to fall into previous experience; AI may not accomplish the same without human intervention.

Product companies are all uniformly using AI during their product launches. This has caused excitement as well as anxiety amongst the people. Will AI assist and enable humans to perform better and with ease, or will AI replace humans to the point of redundancy?

With AI moving into the cerebral workspace, it cannot be denied that it will soon take over many tasks hitherto performed by humans. Just as the advent of machines had a significant impact on how people performed physical labor and dispensed with many jobs handled by humans, AI is soon to be responsible for transitioning many of the tasks performed by humans to robot responsibilities.

In the medical field, AI can help improve diagnostics to a very high level, treatment and care will still require human attention and action. Online shopping has rendered many counter sales jobs redundant; but, at the same time courier and door delivery jobs have increased. Apart from these examples, AI is already playing a large role in the areas of music, architecture, education, manufacturing, gaming and just about any sector or industry. In the corporate world (as a quick example) AI has gotten into recruitment – a very important task in any company. Automatic process was only used before to sort resumes based on keywords. Now, many organizations like Ola or Accenture are using bots to streamline their recruitment processes.

As Artificial Intelligence systems evolve and become more and more sophisticated, job displacement is bound to happen. However, interestingly, many new jobs very different from the ones that are currently in existence will be created. Jobs that involve creativity and those that require advanced degrees and highly specialized skill sets will be available alongside the continuous technology transformation.

A span of 50 years in technology’s perspective can be deemed a very long period; however, if one goes through the article called ‘The Electronic Highway’ published by IEEE spectrum in 1969, it feels like the 2 engineers Robert E. Fenton and Karl W. Olson, traveled half a century ahead to come up with specific solutions for vehicle autonomy.  What is happening now in the automotive industry is what Fenton & Olson had offered as solutions. Though obsolete today, the solutions look at the future with predictions that are coming true now.

In a fast world, such as the one we live in today, commute needs to be swift, smooth with minimal effort by the commuter. Just as the autopilot system in airplanes was designed to relieve the pilot of some tasks so that he could sit back a bit, specially on long haul flights, the emerging self-driving cars require little or no effort from the driver to navigate the vehicle.

For a long time autonomous vehicles, though talked about in articles and technology forums remained a distant dream. The Electronic Highway notwithstanding, suitable technology with adequate software/hardware was a challenge. However, many automotive companies have created ready to hit the road self-driving cars that can drive, park, obey traffic signals, negotiate an intersection and navigate through busy roads. In his book `Rise of the Robot’, Martin Ford says Google’s autonomous fleet in 2012 had driven over 300,000 accident-free miles on roads ranging from freeways jammed with stop-and-go traffic to San Francisco’s famously convoluted Lombard Street. The next year, Google released data showing that its cars consistently outperformed the typical human driver not only in acceleration and braking, but as well in general defensive driving practices.

Almost every car manufacturer across the world is actively into making fully automated, or partially automated cars. The conservative, or rather cautious approach of not going in for fully automated vehicles is because decision-making – a critical aspect in driving cannot be fully owned by machines. Technology is advancing in leaps and bounds and yet every designer, manufacturer or user is aware that no technology can be 100% reliable – at least not as yet. While roads remain consistent with good conditions maintained, the surrounding environment is always dynamic. Traffic diversions, construction activities, weather conditions, sudden road blocks are situations which no system is equipped to handle end to end. And above all, the crux of the issue here lies in the fact that an autonomous car may not be able to decide when to seek the human driver’s intervention.

Having said, autonomous vehicles have some to stay. Self-driving cars’ contribution to society would be a huge reduction in the number of traffic jams and all cars will drive with equal speed and in adherence to traffic rules and lane discipline, thus minimizing the chaos that can occur due to the diverse driving styles of humans.

As Martin Ford states, after electricity, the single innovation that is central to the development of society is the automobile. Very soon, we will look back wistfully to recall the long late evening drives, steering the wheel in sheer pleasure while listening to music and enjoying the landscape.

The other side of the coin no more road rage, driving tickets or penalties!