Perhaps we should start with asking ourselves, has AI lived up to our expectations? From the general public’s point of view, the answer would be a resounding ‘no’. This is not new; frustrations with AI, and its apparent lack of ability to ‘just do what it’s supposed to do’, go as far back as its inception. The general perception of AI, and in what regard it is held by the general public, researchers and developers alike, has always fluctuated. The answer to the question posed earlier would probably be quite different if we had a better understanding of what artificial intelligence is, and subsequently more realistic expectations of its capabilities.
The notion of intelligent machines that can mimic human behaviour and exhibit what humans deem to be ‘intelligence’, has long been surrounded by a lot of optimism and aspirations for the future. However it did not enjoy a smooth ride. The field of AI research was born in the 1950s, and in its early years was the recipient of many generous state-backed investments, in recognition of what AI was thought to be capable of achieving in the ‘near future’. A natural consequence of sweeping predictions, and unfounded claims, was the failure to meet expectations, leading to the whole field of AI suffering many setbacks in its years of infancy, in the form of funding cuts and general disrepute. This pattern repeated itself multiple times, up until the past decade, where increasing computational power, and the ever-growing size of data being mined and needing to be processed, and machine learning methods and algorithms coming to the forefront, revived the field of AI again. Though still generating a lot of hype, specifically surrounding big data, machine learning, and their applications, one can say our outlook and understanding of AI’s potential is a lot more mature than it used to be. But going back to the question earlier, is it mature and realistic enough?
Recently, I have had the opportunity to work on developing a HR chatbot, that is meant to handle the repeated questions and mundane tasks that a business HR department is typically presented with. This included questions such as ‘when is my payday?’, ‘how many days off do i get?’, and ‘when is the next holiday?’. I suspect the urge to develop said chatbot arose from the fact that chatbots seem to be all the hype nowadays. There doesn’t seem to be a site for any major business that doesn’t give you pop-up when you’ve been on their website for a minute or two, offering to help you through a live chat.. with their very own chatbot of course. Admittedly, clever machine learning algorithms allow such chatbots to function surprisingly well. But despite that, they really only give off the illusion of intelligence. This manifested itself clearly when queries like ‘I’m expecting a baby’ and ‘I’m expecting a donkey’, both yielded the same reply congratulating the user.
This gives rise to the question of what exactly it is we’re looking for in the field of artificial intelligence. What do we deem to be ‘intelligence’? Is it very clever classification algorithms that can separate data and classify with high accuracy? Maybe from a business’ point of view, that is sufficient at this point in time, and it provides the value they are looking for. But we can’t really call it artificial intelligence, not by a long shot. I postulate that in order for us to really advance artificial intelligence, we have to develop a deep understanding of natural intelligence. We have to make better sense of human intelligence and how a conscious person operates, and what goes through their mind, and what drives the decision process, and how emotions play a part, et cetera, et cetera. Once we do, we can try and implement this newfound understanding in the form of algorithms, machines, robots, and software applications, like chatbots and virtual assistants. Only then can we really call it ‘intelligence’.
I recently finished reading the book Incognito—The Secret Lives of the Brain by the American writer and neuroscientist, David Eagleman. As the title suggests, in his book he explores the ‘lives’ going on inside the brain. And though it’s impossible to summarise the book in a paragraph or two and do it justice, I will mention some of the ideas and theories presented in the book, which I believe can form a basis from which we can learn about human intelligence in more depth. It’s a superb book by the way, and I do recommend reading it, regardless of whether or not you’re interested in neuroscience or artificial intelligence.
One of the main messages of the book, is that consciousness and what we perceive to be under our control, constitutes only a minuscule part of everything else that’s going on in the brain, over which we have no control. The underlying processes and systems to which we have no access are responsible for nearly everything that shapes who we are, how we behave, and what defines each one of us as a person. What we have been doing in the field of AI, is trying to replicate this conscious layer of the brain, in algorithms and machines, which will only get us so far. Another very important concept introduced in the book, is how we as humans learn. How our senses collaborate to build a complete picture about something. And how once this thing is learned, it is relegated to what the author calls ‘zombie’ or ‘alien’ procedures, to which our conscience no longer has access. But one thing that sets our biological learning systems apart from our attempts at artificial learning is the fact that biology continues to invent new variations in circuitry, trying to solve problems in unexpected and creative new ways. This is in contrast to artificial learning, where once a problem is solved, we move onto the next one.
There are a lot more concepts the book touches on, like competing systems in our brain. Why it is that we develop multiple ways of doing the same thing, and how that helps in the grand scheme of things. Why it is that behaviour changes, and things said may seem outrageous when someone is drunk. Who is the real you, the drunk or sober one? How do you debate things in your head, who is talking to who? Why do you change over time, and why do your perceptions and convictions change? Many other questions like these, are very pertinent in developing a better understanding of human intelligence, and how it can be applied to artificial intelligence.
I appreciate that this blog post has raised more questions than it has provided answers, but that is its purpose. There are a lot of predictions and perspectives floating around regarding artificial intelligence. Even the biggest names in tech have disagreed, and have different understandings and perceptions of what artificial intelligence is capable of, and in which direction it’s heading. Perhaps the takeaway message can be summed up as follows: I strongly believe that before we’re able to make a leap in artificial intelligence, we have to make multiple leaps in understanding natural intelligence. We have to look back at how human intelligence evolved over centuries and millennia. Only once we’re firmly grounded in that understanding, we can hope to meaningfully advance the field of artificial intelligence.