What is this AI thing everyone is talking all about?

With so much media hype around Artificial Intelligence (AI), it’s difficult to understand what AI really is. In fact, two people using the term “artificial Intelligence” can be referring to radically different types of systems, while still having very passionate discussions about the miraculous or catastrophic consequences of it. But if one considers that the definition of “intelligence” is, itself, rather elusive – what is “intelligence” anyway? – why would “Artificial Intelligence” be easy to define?

Artificial Inteligence

Narrow vs Broad AI

Because we could be talking about a wide variety of systems, we typically classify AI in two high-level categories: Narrow (or Weak) AI and Broad (or Strong) AI.

Narrow AI refers to systems that are optimized for executing one specific task. The task may be (or appear to be) very complex, but these Narrow AI systems can become proficient in solving it. However, although they will master that one task, Narrow AI systems will only be capable of executing that task and be practically unable to solve any other task, even much simpler ones. A system that filters spam email from your inbox, or a system that recognizes faces at an airport passport control check are examples of such Narrow AI systems. But the very same system that excels at recognizing one specific person among millions of people just by analyzing an image of their face may be, quite disappointingly, unable to differentiate an image of a dog from an image of a cat.

On the other extreme, we have Broad AI. By Broad AI we refer to systems that are capable of having intelligent behavior across a variety of tasks, from driving a car to telling a joke. Broad AI techniques aim at replicating and surpassing many (ideally all) capacities of human intelligence. In the limit, such AI systems would even exhibit other aspects of human intelligence, such as consciousness of themselves. An important observation is that these systems do not actually exist (yet!). At the moment, they are hypothetical, they are part of science fiction, but already controversial. We don’t even know if we will ever be capable of building one such Broad AI system. Still, they are used as a reference of what we would like (?) to achieve in the future by programming machines that start programming themselves.

Is there anything in between Narrow and Broad?

As a matter of fact, there is. But it is not so much a “slightly stretched but still Narrow AI” nor “Broad ma non-troppo AI”. Although researchers are building more capable AI systems, broadening the set of skills of Narrow AI to have it perform additional tasks is extremely challenging.

Where a great deal of progress has been made is in developing several different Narrow AI systems and then combining them in a larger AI system that benefits from the all those specialized capabilities. Take for example a system like a self-driving car, which is a pretty amazing achievement. Inside a self-driving car there are dozens of narrow AI (sub)systems interacting: there are systems that analyze images coming from a number of cameras and try to identify objects, their distances and their speeds; there are others aiming at identifying road limits, road junctions and traffic signs; others are trying to calculate the immediate trajectory or planning the next navigation decision, and so on. There are dozens of “little” Narrow AI’s (some are actually very complex) at play in self-driving cars.

The same can be said about other popular systems such as the almost ubiquitous voice assistants (e.g. Alexa, Siri) that can “speak” with us, or even about Google Search, which almost guesses what we will look for next. All these are not one big AI system, but a collection of many narrower automated intelligences that work together as an orchestra.

So, how will AI impact Healthcare? Realistically!

In the next few years, we should see many Narrow AI systems being deployed to all fields of application. For many reasons, Healthcare will be instrumental. So, the question is: how will AI actually impact Healthcare in the next 5-10 years? What can we realistically expect in terms of AI? Or, on the other hand, how can we spot what is still science fiction (for now)? Those are all very good questions, which I will leave for a future post, here at Tonic App’s blog!

If you are a medical doctor, I invite you to download Tonic App for Android here or for iPhone here.