Is “Artificial Intelligence” possible?

2017 was the year of AI (for now).

We saw a bloom AI technologies and methodologies in the past year. With the latest Deep Learning techniques, we are starting to see AI programs surpassing human capabilities in many areas: Alpha Go and OpenAI being few of the examples.

Yet, with all these achievements in sight, there’s still a skepticism that remains on whether true AI is actually possible.

This however is mainly due to a misunderstanding of the term “Artificial Intelligence”.

The term “Artificial Intelligence” is an unfortunate mis-term, which has led to many misinterpretations. When the 1956’s Dartmouth conference named the new research field as Artificial Intelligence, they had good intentions for that name. But, as always, intentions aren’t preserved nor obvious.

The common misconception — taking the name literally — is that AI aims to build “intelligence” artificially. However, in reality, the term “Artificial Intelligence” was, and always, meant to be “Artificial”+“Intelligence”. Meaning that it was meant to bridge the artificial and the intelligence. The goal of AI is to observe and understand “intelligent” behavior inherently found in natural constructs — which could be human or otherwise — and attempt to build said intelligent behavior into artificial constructs. These artificial constructs could be computer programs, machines/robots, algorithms, or theoretical frameworks.

This concept is what has brought us models such as neural networks and genetic algorithms, among many others. If you look closely at these models it becomes apparent that all these models are applying (modified versions of) natural intelligence concepts on top of artificial constructs.

The ultimate goal of AI was — and is — is to build a machine with a human level or grater intelligence. (A ‘machine’ is a subjective term here, which can mean any artificial construct). We do not want to reinvent “intelligence” for that. We just need to adapt the concepts of natural intelligence on to the artificial constructs we build.

We don’t build intelligence artificially. We build machines more naturally.

What do you think?

My book on Deep Learning talks more about the history of AI. It’s available on Amazon if you would like to check it out.