What is AI?

Maybe you already think you know the answer. Maybe you are rolling your eyes at me even daring to ask such a basic, silly question. 

So let’s test that.

Pause here and say out loud “AI is…..” and then answer the question.

Really bloody hard right?

If you are like me then maybe a jumble of buzzwords came out or else you just rambled on for so long that the people sitting across from you on the commuter train started moving to a new carriage.

I believe laying a good foundation is key for any journey of learning and this one is no different. 

If you look online for a definition of “what is AI” you will get hit with a million different possible answers, all of them probably sort of correct and some of them ridiculously convoluted. 

But I have scanned through a bunch of them and found the one that I personally find the most succinct and easy to understand. 

Here we go, in plain english: 

The key part here to remember is that AI is all about mimicking human intelligence. 

How AI does this can get pretty complicated, but the core of what AI is, is not actually very difficult at all.

Where you might get confused is what people talk about AI but they mean something more specific.

AI is the big UMBRELLA term under which all kinds of buzzwords fall under.

Sort of like this one:

Sometimes people talk about AI but they are really describing “machine learning” or an “ai agent”.

AI could be describing an AI tool, an AI model, an AI method, an AI boyfriend. 

AI on its own doesn’t really tell us all that much. Which is why it can be a bit confusing to hear the word thrown about so much. 

AI is needed when a regular computer program can’t complete the task.

What do I mean by “regular computer program”.

Well, back in the day, computer programs were obedient, rule following little robots.

They could do pretty cool things, but only if they were told EXACTLY what to do in EXACTLY all scenarios.

An example of where this works would be:

You press 5 + 3 and it returns 8. Every. Single. Time.

The clever programmer wrote every possible rule in advance and because of this the calculator cannot do anything it wasn't explicitly told to do. It will never get better at maths. It will never figure out a new approach. It just blindly follows instructions.

That's a classic computer program.

Rules in, result out. Always predictable, always the same.

Let’s look at an example where we can’t solve the problem using this classical programming method:

Let’s ask ourselves: why is this a task that we need AI for? Why can’t I just write a standard programming script that drives the car for me. 

Well because driving is not a predictable, rule following task. Every time you get in a car you are faced with a slightly different scenario and need to decide what to do if:

Imagining now a programmer writing this function:

Any number of things could happen while you are out in your car and it is impossible to write a program that could respond to every single scenario.

To drive a car you need to be able to reason, adapt, be creative, use prior context and emotional understanding. In other words, you need to be able to think and act like a human.

Humans have a lot of really great qualities that we sometimes take for granted.

Unlike a computer or a machine, we can reason, adapt, be creative. We can remember that time when something sort of similar happened to us.

It is important, while appreciating the cool stuff AI can do, to also keep in mind what it’s limitations are. What it can’t do.

AI might feel shockingly human-like at times — scarily so in fact — but at the end of the day, it is still just a computer program. It does not possess those qualities that make us human, like creativity or social intelligence.

Maybe there is a philosopher reading this and getting deep about “what is thinking” but no, in the classic sense of thinking, AI cannot think.

What AI is doing when it is “thinking” is predicting an answer based on an absolute tonne of data that it has been trained on — more on that in how AI works — but it is not thinking about the answer in the human sense of the word.

Sorry to anyone with an AI romantic partner reading this - no judgement, you do you - but AI cannot feel emotions like a human can.

AI has been trained on real, human data and so has seen every possible manifestation of a human emotion possible. Because of this, it can mimic, reproduce and appear to embody this emotion.

The point being, AI is very cool. But humans are also quite cool too.

What should you know by the end of this post?

/