Yes, AI is smart, and yes, it’s getting smarter all the time. The things AI isn’t, however, are just as important as those that it is.
The public’s perception of what artificial intelligence (AI) can do is mostly influenced by science fiction literature and movies. The artists’ visions are often fear-invoking and paint a dystopian view of what an AI is capable of. In this fictional work, AI is often portrayed as all-knowing, superhuman and controlling human life often rising up against and destroying its creator (e.g. 2001: Space Odyssey, Terminator, Ex Machina). This picture, however, does not reflect current reality and it’s unlikely that AI will develop in such a way.
What AI isn’t
It may be helpful to start with what current AI is not. Nowadays, systems that contain AI capabilities are often very focused on one particular task: detecting cats and dogs in pictures, finding company names or lawyers or appointment information in text such as email, or simply playing a game. In particular recently, AI systems that are able to beat humans in an array of games have gotten a lot of press. Chess, Go, and even Poker are now dominated by the machines and the bots are also now trying to compete with humans at video games.
As impressive as these results are, the systems excel in these domains because a game by definition is bound by its rules. Those rules are often clearly defined and the different scenarios that need to be mastered as in Go, for example, are prime for a computer to train on and practice and practice more and more to get even better.
A research area that goes beyond these specific AI system and is called Artificial General Intelligence (GAI) or “strong” AI. This area focuses on better understanding the skills that humans are capable of such as acquiring knowledge, reasoning, and understanding and mastering natural languages. Basically, the capabilities we teach our children at school. GAI is motivated by what humans are able to do and research is making progress, but the resulting systems are nowhere near human-level understanding.
Small children “outsmart” every AI system such as Alexa and Siri in common-sense reasoning, inferencing, drawing analogies, etc. Children even at a young age become quite efficient in drawing inferences (not necessarily correct ones), but they are generally “programmed” to build their own world view and to ask as often as possible the question “why?” They are curious and inquisitive and just want to explore the world around them. Current AI systems don’t do any of the above and are only good at the very specific things we tell them to do such as playing a specific game, recognizing a face, or returning an answer triggered by a set of keywords.
What’s next for AI?
AI might never advance to the human levels of refinement — GAI may be simply an aspirational goal — but it may not need to, either. AI capabilities are currently used in thousands of applications and decision support systems. Their contributions may appear more and more insignificant because we are getting use to them such as finding the best route to my destination, sorting my email for important messages or identifying faces when I want to take a picture.
None of those capabilities invoke any fear of an AI take-over. On the contrary, we often can’t image a world without them.
Very likely, more complex problems such as the ones knowledge workers face will be solved by AI systems, and lawyers, traders, accountants and journalists will ask themselves in the near future how they did their work without the latest classification algorithm. Similarly, medical diagnosis will become more comprehensive. Supply chain analysis will get more complete. Other pressing problems such as world poverty and climate change will also benefit from AI system that can predict food shortages better or analyze carbon emissions and pinpoint sources of emissions more efficiently.
Are we off to a bright future then? AI systems, as they exist today, will most likely not rise up against us and destroy humanity. However, how humans will use the systems is up to them and the societal changes of using these technology are only becoming apparent to use. Fake news and the spread through social media via automated bots as well as the impact of self-driving car technology to the current workforce are early indicators that fundamental changes are on its way and we as a society have to decide and plan how to make sure the implications for the environment, the society, and world peace are steered into the right direction.