Terms of Reference :: What is Artificial Intelligence

As I did with my basic primer on data science before, I want to lay down some foundations to make sure we’re all on the same sheet. Coding the same lines? We need a more appropriate phrase.

Anyway, before we start talking about how to incorporate AI into your human capital program, let’s talk about what artificial intelligence actually is. Of course, a lot of this is theory so there are bound to be lots of opinions on this matter, but I’m going to do my best to filter through those and set a baseline. And I’m always open to more education by the experts, so if there’s something you want to add, please chime in!

What is Artificial Intelligence?

ar·ti·fi·cial in·tel·li·gence / ärdəˈfiSHəl inˈteləjəns /noun

  • the theory and development of computer systems able to perform tasks that normally require human intelligence, such as visual perception, speech recognition, decision-making, and translation between languages.

We’re not talking HAL or Skynet. We’re not talking I, Robot or anything Isaac Asimov generated. Well, yet. Statistically speaking, the moment we’re currently in is the slowest moment of technological advancement we’ll see in our lifetimes. It will only get faster from here.

Artificial intelligence hearkens back to Alan Turing’s “thinking machines.” Hence the Turing Test. Does the computer behave in such a way that a human being can’t distinguish the machine from another human being by using replies to questions. The test doesn’t care about how accurate the questions are, but how much the machine sounds like a human.

The fact that this conversation started back in the 1950s and we were thinking that far ahead of the science of the time still amazes me. So much of what we were dreaming about has become real, and so much eludes us. Through natural language processing, we can make machines sound human. But there’s still a lot that eludes even the smartest programs.

Still, artificial intelligence as it is today allows machines to perform tasks we would previously have needed a human to do. Machines now famously play chess, provide automated assistance right from your phone or a pod sitting on your countertop, and are starting to drive cars. But that’s not where they’re most important to our lives, at least in people analytics.

  • Machines perform repetitive and high-volume data analysis tasks
  • Machines identify structure and find oddities and irregularities
  • Machines learn routines and recommend efficiencies
  • Machines study patterns and recommend new media or products
  • Machines detect fraud and other digital behavioral anomalies
  • Machines classify images and recognize objects
  • Machines back propagate data when an answer is wrong to learn

When used in the human capital realm, machine learning, deep learning, and artificial intelligence can help us make use of the massive amounts of human capital data we generate and ingest. It’s what helps us make the jump from long term aggregate forecasting to individual retention and exit prediction, by cutting through the huge amounts of data and automating processes so that analysts can study trends and behaviors that normally would have taken years to identify.

So what are the types of AI?

There are generally three types of AI. Two are science fiction at this point, but for anyone who says we won’t get there, remember what people thought about Star Trek’s disk-tapes, communicators, and the intelligent computer people walked into the room and just talked to {“Alexa, ask Hulu to play ‘The Ultimate Computer’. Wait, maybe not the best example.”}.

Here’s how we break it down.

Narrow AI {ANI}

Narrow AI {also known as weak AI} is what we see most commonly, which is AI focused on a narrow subset of tasks. It’s very good within that subset. But it can’t really learn outside that subset. A speech recognition engine can’t parse apart images, and an image recognition AI can’t play chess.

Self-driving cars and your Siri automated assistant are examples of several narrow AIs working in concert with each other. That gets us a bit closer to strong AI, but it’s only as good as the number of ANIs you chain together.

General AI {AGI}

Artificial General Intelligence {also called strong AI} is science fiction at this point, but hypothetically, if you blend enough ANI together and teach a machine to author new programs based on functionality it has learned…anyway, we don’t have it yet, but AGI is genera-purpose smartness that can be applied to any task a human can do, as well as learning and improving itself.

While we’ve made great strides in intelligent tech, we have yet to teach a system to think abstractly and to truly innovate. Creativity remains a human domain, and the test that separates us from creating General AI or Super AI.

Super AI {ASI}

Artificial Super Intelligence is intelligence that goes beyond human intelligence. ASI is not only able to mimic human behavior but is superior in cognition. This is the often the stuff of science fiction nightmares, as we don’t know how long it would take a self-improving AGI to become an extremely powerful ASI.

Some see it as a future helping hand to mankind, but until someone manages to teach an AI ethics and empathy, there’s no guarantee that a superintelligent machine would care whether or not it’s helping mankind. And when you’ve already developed it to be able to add to its algorithms and learn and improve? Suffice to say that we need to really think about what we’re doing before we develop an intelligent processor and give it all our data and all our functions.

This is the piece I’m both most interested in and most nervous about implementing in the human capital arena. Fortunately, there’s a lot of good debate and discussion about how to do this, how to do it properly, and whether it’s all even possible. I’m looking forward to seeing where the discussion goes!

Leave a Reply

%d bloggers like this: