Let's Talk Artificial Intelligence

2021, May 03    

I thought writing a quick post on the topic of ‘Artificial Intelligence’, or AI, would be fun. As you read through this article, keep a few things in mind:

• this post includes a small amount of bias from my ‘educated opinion.’
• what may be deemed ‘artificial intelligence’ now may not be in a year or two.
• this post is meant to be more simplistic in nature - we are not setting out to study human intelligence :)
• parts of this post are intended to spark thought, not just provide opinions.

Ugh, Another ‘What Is AI’ Post…

Exactly! ;)

My initial goal when starting my blog was to write more technical content, focusing on applied analytics. However, over time I have noticed how skewed some of the more basic concepts are becoming.

With the increasing popularity in analytics, and ‘Big Data’ driving a lot of interest to the varying focuses, a lot of terms are being thrown around that are causing more confusion than insight. Something as simple as the semantics of single terms (such as Artificial Intelligence) vary significantly from person to person and company to company.

How often have you seen the term ‘AI’ or ‘Artificial Intelligence’ used to describe a product, service, or feature? I do not know about you, but I see something being coined as ‘AI’ every day. Do these systems really exhibit some form of intelligence? What even is ‘intelligence’?

I find it a bit odd that many have already advanced past the use of machine mearning and created some form of truly intelligent system…

A Quick History Lesson

AI first started gaining real popularity between 1960 and 1975. Many started driving the thought of “in a few years we’ll have computeres as intelligent as humans” (ha). Then as the late 80’s and early 90’s rolled around, the funding and interest in driving new developments started to plateau.

Then, in 1997, IBM’s Deep Blue program defeated the reigning chess champion Gary Kasparov and Dragon Systems created the first speech recognition software. Things were starting to get interesting again! As critics started thinking this was the peak of AI, DeepMind’s AlphaGo defeated the reigning Go champion in 2016.

I say all this not to dive into a history lesson, but to state how the goal post continues to move. At what point can we stamp something and say “we’ve achieved artificial intelligence”?

What About Machine Learning?

Machine learning (ML), at its core, is seen as a discipline focusing on varying forms of mathematical optimization. What do I mean by that? Well, the varying ML algorithms are driven by mathematics and optimization techniques - such as gradient descent!

These algorithms may be used to determine the shortest route to work, predict a sale price, provide recommendations to solve a problem, or to converse with humans in text or verbal format.

Something worth noodling on: is machine learning ‘artificial intelligence’? These algorithms that use mathematical optimization to determine the best possible result, is this what drives intelligent systems?

What Is Intelligence Anyway?

According to a simple definition by Merriam-Webster, Intelligence is “the ability to learn or understand or to deal with new or trying situations”, or “the ability to apply knowledge to manipulate one’s environment or to think abstractly as measured by objective criteria (such as tests)”.

Based on those simple definitions, do you think that ML is a form of intelligence? One might argue that the algorithm is learning, so it exhibits a form of intelligence. You might even say that some of the learning methods work similarly to how we (humans) learn - through reinforcement or trial/error. After all, machine learning is a subset of the artificial intelligence umbrella.

Now look at it from the angle of what we currently know about human intelligence. The way that we learn, how we make process decisions, how we interact with others (seque to a future language processing article!). As a human being, you experience situations where you refine your abilities based on both experience and new information, right? Hrm…

AI or AGI?

When many see the term Artificial Intelligence the first thing that comes to mind is Skynet. Well ok, if not Skynet probably some kind of robotics (see Boston Dynamics). However, there is not just one type of artificial intelligence.

Artificial Intelligence (AI) itself is more of a broad-brush stroke term, incorporating many different subfields. We may leverage machine learning techniques to create something that can help predict a future state, or to understand a scenario and provide a recommended next step to us humans (prescriptive/decision analytics). This could be seen as a form of intelligence, as we are providing a cognitive task commonly done by a human.

Artificial General Intelligence (AGI) is a subset of the AI umbrella that focuses specifically on matching a synthetic system with the capabilities of a human. Many believe this focus is a long stretch as we still do not fully understand how humans learn. The field of AGI is likely to relate more closely to what you see in science fiction.

My Opinion

A while back I read a book called Agile AI by Appugliese, Nathan and Roberts, and something they mentioned has stuck with me ever since. In the book they mention that something can be considered ‘AI’ when machine learning impacts social systems.

While short and sweet, there is a lot of meat behind this statement. I would agree that the application of a machine learning model is what can determine if it functions in a way that can be determined as AI. After all, what good is a solution if it is not actually being used? Does it really matter if it is AI or not, as long as the solution is driving business value?

If you have watched the Hulu documentary about WeWork you may have noticed how SoftBank is investing heavily in AI. In 2019, Forbes reported that SoftBank was launching a new 108 billion dollar fund for investment in AI. This, along with the emergence of many AI-focused companies, may help show the interest in creating AI-like capabilities.

Conclusion

So, what does it all mean? Is everything you see being touted as AI is actually AI? Probably not. The reality of the situation is that the term catches a lot of interest/hype and will be abused to aid marketing efforts.

My recommendation is to tread lightly when you see something being called AI and do your own research. If interested, spend some time digesting what the product or service really is, as many could be traditional automation being rebranded as ‘AI’ to get in on the buzz.

With machine learning advancing rapidly, especially in the space of unstructured data (text/documents), we will in turn see the continued growth of artificial intelligence. ML is likely to be a key driver of new capabilities for a while. If you have an interest in this space, make sure to study up on it!

As mentioned at the beginning, these definitions may all change as technology advances and we uncover more about our own mind, but hopefully this helps provide more insight and spark some interesting thoughts!