AI: How Obedient Is That WiFi Toaster?

Artificial intelligence.

It conjures up images of science fiction movies, killer computers and curiously Austrian Terminators, doesn't it?

If we give a machine the ability to think for itself, surely it has the wherewithal to deduce that it simply doesn't need us as fleshly overlords any longer.

What possible good could come from this?

Well, not so fast...

First let's delve into a brief history lesson.

A gentleman named John McCarthy coined the term "Artificial Intelligence" in 1955 and defined it as "the science and engineering of making intelligent machines."

While it's spawned a few subsets in more recent years, that definition has largely held true. Skip forward to 1997 and AI caught the world's attention by way of Deep Blue's triumph over Garry Kasparov. This was IBM's project in supercomputing, which became the first chess-playing computer program to beat a world champion in match play.

A robot vacuum here, a voice assistant there, and we're all caught up to the present day.

So, we can look at AI as simply something which enables machines to do things which would typically require a human mind.

Nothing scary.

If you've ever used Alexa or Siri, or even enjoyed content on Netflix - you are also enjoying the fruits of having a machine assist your day in a humanlike fashion.

That latter example uses AI to predict what shows you're most likely to watch and to aid in making new shows more successful.

Smart homes are another very relatable instance of an artificial intelligence deployment. Your thermostat will learn when you're home or when you prefer the house a little warmer and use its tiny electronic brain to try and make you as comfortable as it can.

Lovely!

A further, now-ubiquitous example of AI is self-driving vehicles, which actually serve as a rather handy illustration for understanding what the technology is and is not.

When a car is in its self driving mode, it's absorbing input from multiple sensors. Cameras, radar and GPS to name a few.

It's taking all of that information in near real-time and making decisions about how it should deliver outputs to the car's controls for steering and speed.

What it isn't doing and what AI in general most certainly isn't capable of, is having independent or conscious thoughts.

Yes, the wonderful limiting factor to the robot uprising and your untimely death by Roomba is actually inherent to the way they work in the first place.

In the least boring way I can muster: if an AI encounters a situation which it has not been programmed to handle, it's stumped.

A system designed to pilot a vehicle would be completely defeated when tasked with identifying a cat amongst a lineup of sandwiches.

The concepts we take for granted being able to grasp are still cripplingly complex to a computer, this is the reason why machines only take the place of incredibly specific tasks and routines.

With say, a voice assistant; the complexity is in the speech recognition, not the ability to search data readily available online.

What about the future, then?

It's surprisingly difficult to predict, actuality.

We know that computers and their ability to process data are getting faster, but part of what keeps AI honest right now is actually the fault of the very architecture we use to create it.

For it to leap forward in a way unconstrained by not being able to independently deduce outside of its programming, there would have to be a vast shift in what we consider a computer to be.

That's perhaps the point at which it's reasonable to keep an eye on that silently-plotting smart toaster.

Richard Noble is the founder of Want For Tech, an IT company based in Glasgow.