Explaining AI to Non-Technical Individuals

Question

How does one approach proposing AI to management? This is something I have struggled with for a long time. I want to implement AI toward a specific problem in my place of work. My supervisors are generally willing to listen; but they want to know how the algorithm(s) is going to work. They are not programmers. My tendency is to write out the math and step through it. However, most of them don't want to do that because they have a limited amount of time to sit there and listen. On top of that, some of these algorithms can get somewhat complex.

Lets take a simple neural network for example; how would you explain the way it works without diving into the math?

Answer 1

There are a lot of ways to describe "Artificial Intelligence".

• An artifact that "makes a decision"

This form of automation/computing/AI goes back to neolithic times.

Early AI was purely heuristic. (Also known as "good old fashioned AI" aka "Symbolic Intelligence" aka classical expert systems.)

The current generation of strong (narrow) AI is statistical, which encompasses both neural networks and evolutionary/genetic algorithms.

Artificial intelligence is a machine that makes a decision. Modern statistical methods allow these machines to learn and improve their decisions.

Current best AI is "narrowly superintelligent" in that it can exceed humans at most definable tasks, but machines still lack the intuitivity of biological brains, and this strong intelligence is narrow—restricted to single problems or classes of problems.

Answer 2

In this context, I would focus on the what and not the how.

• What part of the business problem will it solve?
• How does that fit into the bigger solution (AI model is probably making a prediction - is that it? Is there an application or report built around it?)
• How do you expect it to perform compared to alternative solutions?
• What do you need in terms of resources: data, computation, time?

As far as how it works - I would just describe it as a “probabilistic model” and leave it at that. If they want to go deeper, they’ll ask. You may not even know the exact model/algorithm/approach yet, as often experimentation and iteration are necessary.