What is Artificial Intelligence?

- I'd like to ask you a question. Have you used AI in the last 24 hours? If your answer is no, let me ask you, if you have used face recognition to unlock your smartphone or interact with autocorrect or voice-to-text features? If so, then you're already a regular AI user. We may not realize it but we are already surrounded by AI. It's in our smartphones, it's in YouTube's recommendation engine, it's in Google Search. It's in many of the driver assist features in Tesla cars. It's in many hospitals where AI is starting to help radiologists diagnose MRI images. It's in our policing systems where facial recognition is used increasingly to identify suspects. So what exactly is artificial intelligence? The key to answering this question lies in first defining the term natural intelligence. That is intelligence as exhibited by humans and other artificial. According to the American Psychological Association, intelligence refers to the ability to understand complex ideas, to adapt effectively to the environment, to learn from experience and to engage in various forms of reasoning. Using this definition as a starting point, we can define AI as a computer system that exhibits these traits to at least some degree. By this definition, it's easy to conclude that a calculator would not qualify as an example of artificial intelligence. While it can do complex calculations, it does not adapt to the environment and it certainly does not learn from experience. That said, not everybody in the AI community dismisses the calculator as not being an example of very rudimentary AI. Nearly 400 years back when the French mathematician Blaise Pascal invented the first mechanical calculator, many a journalist ascribed human-like intelligence to this simple machine. After all, what differentiated humans from animals was that humans could calculate, animals couldn't. So if a machine can calculate, it must be called intelligent. Looking back, if we now conclude that no, a calculator is certainly not intelligent, it's worth noting that we humans have long suffered from hubris. Once a machine starts doing cognitive tasks as well or better than humans, we have a tendency to denigrate it as merely calculative ability not worthy of being called intelligence. Even today, a small minority of AI scientists dismiss facial recognition, natural language understanding, and driver assist features as not really true AI. To me, such values reflect, mold the hubris of these scientists, then how we should define AI. As may be obvious by now, AI is not a binary black-or-white concept. That is a computer system is either AI or not AI. Rather, like human IQ, computer systems also vary in their degree of intelligence. Merely 15 years back, the accuracy of image recognition systems to classify photos that did or did not include the image of a cat hovered around 50%. No better than tossing a coin. Today it's in the high 90s. Similarly, almost by the day, AI systems are getting smarter at interpreting MRI scans, understanding and generating speech and autonomous navigation. Perhaps in 10 years' time, we would look back and wonder why we ever called present-era systems as AI, similar to how we view calculators today. Present-day AI systems are extremely domain specific. An AI that is good at facial recognition is useless at understanding speech and vice versa. Domain specificity also implies that some types of AI are already far smarter than humans could ever be. For example, playing chess or the Chinese game of Go. On the other hand, in most other domains, humans remain much more intelligent than computer systems. The biggest difference currently between human and artificial intelligence is that human intelligence is multifaceted. The same brain can read, write, speak, calculate, interpret emotions and do 1,000 other cognitive tasks effortlessly. Even the best AI systems today reflect narrow, artificial intelligence. Now, think about the many aspects of your work context that are already or could soon be powered by AI.