The definition of AI

I have a love hate relation to buzzwords. On one hand they are fluffy and not well defined, on the other they drive technological advances (more on this in another future post). However as a Technologist it is important for me to understand technological buzzwords and hence a lot of my time is used on reading and discussing buzzwords. Artificial Intelligence is such a buzzword, and my favorite for the time being. When learning something new I always try to approach it by first defining it, which again makes it easier to put it in a basket and comparing it to similar terms. With AI this is not a simple task.

There are almost as many definitions for AI as there are people writing about it. There are numerous reasons for this, however one of the biggest ones is that the definition also changes over time (which in itself is a definition). Artificial Intelligence is often referred to as what we presume require human like intelligence, and this again changes over time. For instance it was assumed that Chess required intelligence until Deep Blue beat the world champion Garry Kasparov in 1997, at which it was understood that this wasn’t the case. The same can be said for image captioning, language understanding, driving cars etc., we all thought that these needed intelligence until it was proved not to, at which time it no longer is defined as AI.

But to fully understand this it is important to break AI into two, namely Artificial and Intelligence.

• Artificial is easily defined as something being human-made rather than occurring naturally (https://en.wikipedia.org/wiki/Artificiality)
• Intelligence is however a completely different story, it has been defined in a multitude of ways and there is no simple way of defining it. I won't even try to do so however while reading through dozens of definition one of the commonalities I see is that it is often referred to as adaptive behavior outside of normal context (eg. The possibility to learn new things)

This is however where it gets interesting. As no one actually are able to defined AI it has been broken down into subareas (which again has been broken down into subeareas). The main split is between Weak and Strong AI:

• Weak or narrow AI is a process created to perform a small task and cannot learn new tasks without the help of a human. This means that all current "AI" processes for instance Chatbots, Robotic Processing Automation even self-driving cars can be put into this definition.
• Strong AI or Artificial General Intelligence (AGI) on the other hand would on the other hand be a machine/software that is conscious and has a mind of its own.

This also means that all current buzzwords flowing around in the AI sphere such as Machine Learning, Deep Learning, Neural Network etc. are methods that would fall into both definitions, however their scale of course differs. I won't get into those now, they are the topic for another really interesting post.

As for AGI it yet again has many definitions to it, however one of my favorites is Technological Singularity (https://en.wikipedia.org/wiki/Technological_singularity). Here we are borderline sci-fi, but this means the moment that a computer can evolve itself, at which time it's evolvement will explode. This is often depicted in apocalyptic movies and books such as the launch of Skynet in Terminator or "The birth of AI" in the Matrix.

What are your thoughts on the topic? What do you think is AI today, and what isn't? Love to hear.