Decoding AI: Making the Case for Artificial Intelligence

comments

By Shamli Prakash, Opera Solutions.

Everywhere you look people seem to be talking about one thing. Well, I mean apart from Donald Trump and his disturbingly entertaining antics. And that’s AI. There are those that can’t contain their excitement (‘Yo, AI is going to change the world!’). There are those that are uncertain (‘We really don’t yet fully understand the impact AI would have on everything’). And finally, those that have declared apocalypse is upon us (‘AI will eat up humanity. Prepare to perish!’). It has in fact become a much abused term, being used out of context and often erroneously for a lot of things.

Credit: Many Wonderful Artists (original).

But what really is AI? I am sure you have wondered too, whether or not you would admit it to a judgmental set of friends who gawk and point the moment you show a chink in your armor about this opiate of today that AI has come to be. So let’s see if we can break it down a bit.

If you stumbled over to this page, you probably have this bit of rudimentary knowledge already — AI stands for Artificial Intelligence — semantically, it is pretty easy to grasp as these two words are fairly self-explanatory in themselves. However, you do need to scratch beyond the surface to get to the heart of it. We will take it one by one.

Artificial:

The dictionary meaning of artificial is ‘caused or produced, as opposed to existing naturally’. In other words, it means something that is created, manufactured, simulated. In the context of AI, the artificial-ness comes from machines which do this creation, manufacture and simulation.

Intelligence:

There are many ways in which you could attempt to explain the meaning of intelligence. At it’s very core though, it is the ability to come to complex decisions and conclusions, while taking into account multiple parameters. Think about it — isn’t this what sets humans apart from all other species, their ability to consider facts (clear or ambiguous), weigh them, identify alternatives, and ultimately come to an inference? That decision or conclusion could be something simple like ‘I will watch Dunkirk because I think I will enjoy it.’ Or something way more complex like a move in the game of chess.

So let’s get these two things together to understand AI. In the most simplistic of ways, AI is the concept that enables machines to replicate human intelligence, i.e. the ability to make decisions and come to conclusions based on available facts. These facts could be binary or fuzzy (terms I will explain in a later post), just like the parameters available for human decision making.

This brings us to the next question about AI — the very existential question of why? Why do we need machines to make decisions that humans can? What is the benefit of AI? While this is a vast topic in itself, I will focus on answering the question in the context of the broader attempt of explaining AI in layman terms.

Let’s go back to the definition we arrived at — AI refers to machines making decisions based on available facts. In doing so, machines have some distinct advantages over humans, the following four being primary among them:

(1) Machines do not have biases. Humans do. If we are talking about pure, logic based decision making, these biases can sometimes get in the way. A machine, for instance, would not make a sub-optimal decision just because it was in a bad mood due to an argument in the morning with the spouse about whose turn it was to take out the trash!

(2) Machines can have much higher computing power. As a result, they can literally enumerate all the different permutations and combinations of relevant factors, and compute resultant values for them to come to the most optimal decision. While humans can surely understand this conceptually, their ability to actually have thousands or even millions of rows in their mind against which they put a score, doesn’t exist. This is one of the prime reasons for computers being able to beat humans consistently in logic driven games like chess. When the computer makes its move, it can consider all the possible next moves you could make, and make a decision based off of that. (Side note: Chess per se though is a game of infinite possibilities and therefore a study of chess in the context of data and AI is an extremely fascinating thing).

(3) Machines don’t make ‘human errors’. That’s a term usually used for unintended mistakes that happen simply because, well, man (or woman) ain’t a machine! Small errors that could change outcomes, e.g. you miss an item inadvertently when manually calculating the total bill amount at a restaurant. In other words, once programmed, a machine would always, until eternity (or at least until it breaks down, gets a bug, or needs an upgrade), will do things exactly as expected.

(4) Machines don’t need breaks. Or sleep. Or weekends. Or vacations. Or sick leaves. You get my drift. From a productivity point of view, this clearly is a very pertinent factor.

There are obviously cons as well to an AI driven approach, several of them being very substantial, particularly from a sociological standpoint. It ultimately boils down to the context in which AI is being used.

Before I end, a quick word on the history of AI. While the buzz around it has become intense of late, with debates raging even as self-driving cars become a reality, AI by no means is a recent phenomenon. The idea of it has been around for as far back as you can think. You can find allusions to it even in ancient Greek mythology. Aristotle (that’s 4th century BC) is credited with inventing the first system of deductive logic (which is at the crux of AI). The first digital machine to take over computing from humans was the calculator, a crude version of it was invented by Pascal way back in 1642.

Of course, the pace of change has been phenomenal in the last few decades and is only going to get stronger. I think we are now very close to where philosophy would need to enter the realm of technology for the future of AI to be determined. The question is no longer ‘can we get machines to do this or that’ (the answer is yes for most things you can think of), question now is ‘where all do we want to do it?’

 
Bio: Shamli Prakash is Senior Vice President at Opera Solutions. Management consulting. Accidental AI enthusiast. Love the stories that hide in data. Passionate about women in business.

Original. Reposted with permission.

Related:

• A New Beginning to Deep Learning
• Deep Learning is not the AI future
• A Guide to Understanding AI Toolkits