Why AI is a big deal

It’s simple. Artificial Intelligence (AI) enables us to solve problems that up till now we could not solve using computers. Why? Because these problems don't have analytic solutions that rely on logic, which is what computers require. No logic, for example, can tell you why a cat is a cat and how to recognize one – you “just know”—so you can’t give a computer logical rules about how to recognize one. 

In some cases, AI even enables us to detect patterns and solve problems that we did not know existed. And sometimes, the problem and solution may be understandable, but it's cheaper to construct the solution using AI instead of analytic software that follows rules.

So basically, AI enables computers to solve problems that so far have been unsolvable or too expensive to solve.

In some cases, AI even enables us to detect patterns and solve problems that we did not even know existed. And sometimes, the problem and solution may be understandable, but is cheaper to construct the solution using AI instead of analytic software that follows rules.

So what?

Like other major changes, not much seems to be happening at first. But in ten years' time, the changes will be huge. And they will especially impact white collar work. You owe it to yourself and your organization to know what AI can do and have a plan, now.

Recipe for making an AI

All you need to make an AI solution work is to feed it data that is correlated with the outcomes you want. To “train” it how to recognize cats in a picture, feed it a lot of pictures without cats and a lot of pictures that have cats and that are marked as such. Then feed it a lot of test pictures until it’s ready.  In computer jargon, we say that legacy software is “rule-based” whereas AI is “data-based”.  

Learning from data is not unique to AI. It applies to Human Intelligence also. That is where we got the idea. Saying that software is “data-based” is equivalent to saying that humans learn through experience or by trial and error. Let me illustrate. 

All great athletes win with their brain first. A champ like Serena Williams did not become #1 tennis player by analytically applying the laws of physics to coordinate her movements or calculate the trajectory of the ball. She did it by collecting a lot of data about what works –by practicing a lot and by #TrialAndError. Of course, practice also conditions the body, but at that level and in a complicated game like tennis, conditioning is secondary.

What AI can do

AI enables Amazon to predict demand to give you what you want. It’s how Netflix can recommend the right movie and it’s how they design their series to titillate you – without understanding the psychology involved. It’s how a virtual agent responds correctly to what you say. And it’s how Google knows there is a cat in that picture. These are all examples of learning from past doings or experience represented in data. When you make a Google query then click on the matches you like, you are training the Google search AI to correlate certain web pages with the query you entered. You are generating Google's training data! That is why Google search is free and why we don’t need coupons like the one below.  Every Google user is continually training Google search to be better. It's a self-maintaining, self-learning system.

How AI democratizes software

The process of making a specific AI solution does not fundamentally need a computer science expert. So now one can produce software solutions without having an army of programmers and project and business managers, which makes financial payback necessary. AI opens the process to the others who have no such expertise.  What you need is imagination, effort, and access to a generic AI engine, and of course a lot of data. Many cloud providers like Amazon, Microsoft and others offer open AI platforms that enable anybody to build AI models about things they care about. 

AI enables more people and organizations to develop solutions that are important to them instead of relying on specialists interested in financial gain. So now people and organizations without core software competency can produce software solutions.

Familiarity with programming definitely helps, but, in theory, is not required. The reason it helps has nothing to do with AI itself. It helps because the tools to organize and manipulate the data can benefit from software expertise. In some cases it is required as we discuss below.

In practice, large AI projects still require a lot of software and business expertise. Reproducing Bing or Google search is a multi-billion-dollar operation. There will be a lot of legacy programming to scan millions of websites. A lot of programming to prevent others from manipulating the platform. A lot of programming to produce a good user interface. A lot of programming to manage the enormous physical infrastructure required to collect and hold the data. And of course, a lot of money to actually hire an army and build all of that. Conceptually, what Netflix does is also simple. But actually doing it is quite hard.  Same for Amazon. They present a completely digital interface to the users, but there is a lot of physical infrastructure behind it to enable Amazon to sell and ship products.

The intersection between the AI and digital assets on one side and real life on the other is hard to implement. 

AI is only a decision-maker, not a doer, so if you need to "do" things, you will still need regular software or some sort of automation platform

AI makes decisions and correlates things based on past data. This is a picture of a cat sleeping, and that is not. If that is all you need, great. You’re done. But if you want to translate AI decisions to actions, you will likely need to integrate the AI you built with something else, and that requires programming or some sort of automation platform, or both. AI will not end programming, at least not until we build an AI that can actually produce software. And that, is not imminent. 

Other resources

AI Primer for the non-specialist part 1

AI Primer for the non-specialist part 2: Different types of AI

Who will be the face of your company? Yours or somebody else's AI? (Consumer-based vs. Enterprise-based AI)

Consumer-based vs. Enterprise-based AI (Applying consumer-based AI in financial services)

#AIStrategist #AI #automation #RPA #strategy #AI4Leaders  #Google #Amazon #AWS #Microsoft #Facebook