Simplifying (?) Artificial Intelligence

Artificial Intelligence is a vastly complicated topic. It’s not just that it has a lot of components (because it does) it’s because it isn’t really a single topic. It’s a surprisingly old field of study that has morphed and changed over the years. I worked with “AI Systems” in the 1970’s – and even then it was based on much older concepts.

Lesson timeWhenever you attempt to learn a new topic, there are two dangers to keep in mind: making the topic so complete and deep that it becomes impossible to understand it; or making the topic so simple you think you understand it, but you really don’t.

At the risk of the latter danger, let’s simplify AI a bit – at least to start our journey. We’ll start with the simplest of definitions of AI. I certainly didn’t invent this description, but here’s how I talk about AI:

Artificial Intelligence is when computers do things that humans normally do.

Now that is quite over-simplified – and “begs the question” (did you know that phrase doesn’t really mean what you think it means?).

What do we mean by “things humans normally do”, and how do they do that? Let’s use an example.

Sitting next to me right now is a cup of coffee. (It doesn’t actually matter when you read this article, that is probably still true). I see the cup  of coffee, decide that I want some, pick it up, and take a sip.

That simple action is immensely complex. Using the amazing vision system my body has, I can differentiate the cup from all other objects on my desk. The vision system, connected to my brain, recognizes that object as a cup, associates it with the concept of coffee, recalls the memory of making it recently so that I know it’s still hot (and knows what “hot” means), another subsystem “decides” I “want” the coffee, sends a signal to the subsystem that controls movement, grasping, spatial awareness and others, lifts it to my mouth (which is yet another subsystem) which drinks the coffee (another subsystem), all while my eyes and mind have moved on to other tasks.

And that’s just drinking a sip of coffee.

I think  about driving a car, or even more complicated, when I ride my motorcycle, and it becomes amazingly complicated to even describe the processes involved or even to define the subsystems involved. The field of AI attempts to do just that.

So let’s simplify that a bit. Almost all of these components involve three basic concepts, all modeled on how the brain works:

  1. Acquire
  2. Process
  3. Respond

Acquire

In this phase, the system takes in information or stimuli in some way. In my case, it was the vision system differentiating objects using light reflection. Computer vision, an implementation of AI, works exactly the same way.

Note that acquiring doesn’t have to involve external senses like sight, sound or touch. It can also be the process of getting new or different information from another subsystem. This is the phase in AI that you write to bring in data – wherever that data is from.

And stringing the output from a “Respond” phase can lead to this “Acquire” phase of your AI program.

Process

In this phase, the data is processed in some way. In my case, the memory of a cup, coffee, and heat are all involved in that processing. I used a recognition algorithm to “see” the cup with the coffee in it – that action is pretty complicated.

In the case of AI, functions such as Machine Learning and Deep Learning, and algorithms in those disciplines such as Neural Networks are used to process the data from the inputs. The Cognitive Services API for computer vision contains an example of the input from the computer’s camera to be passed on to another API for image recognition.

Respond

In this phase, the AI system either performs an action, or sends the processed data along to another AI function (or both). In my case, I “decided” I wanted some coffee. The decision subsystem acquired the data, processed it, and responded with an action to trigger my gross and fine motor subsystem to reach out and pick it up, which had an acquire phase of spatial awareness, nerve endings and so on to process where my hand was at the time, and a response of touching the cup, grasping it, lifting it, and bringing it to my mouth. And each of those is also a subsystem.

AI works in this way – a series of acquiring information, processing it, and responding according to a set of rules. AI is a chain of these smaller systems working together to mimic human behavior.

(Well, mostly. It’s more complicated than that of course)

 

I’ll flesh this out more in future articles – for now, the definition of AI, along with these three components can serve as an (incomplete, to be sure) introduction.

Advertisements

Published by: BuckWoody

Buck Woody works on the Microsoft Cloud and AI Team, and uses data and technology to solve business and science problems. With over 35 years of professional and practical experience in computer technology, he is also a popular speaker at conferences around the world; author of over 700 articles and seven books (databases, machine learning, and R) sits on various Data Science Boards at two US Universities, and specializes in advanced data analysis techniques. He is passionate about mentoring and growing the next generation of data professionals. Specialties: Data, Data Science, Databases, Communication, Teaching, Speaking, Writing, Cloud Computing, Security Clifton's Strengths: Individualization, Learner, Connectedness, Positivity, Achiever, Ideation

Categories AI, Learning Data Science, Other ArticlesTags6 Comments

6 thoughts on “Simplifying (?) Artificial Intelligence”

  1. I was exposed to AI in my early 1970’s CS classes, Minsky, Weizenbaum, et al. It is my recollection most of these theories were computational in nature, maybe pattern recognition of objects, or just really smart chess strategy.

    A decade or two later I read a fascinating book “The Origin of Consciousness….” by Julian Jaynes (Harvard researcher?) which controversial at the time.

    What impressed me (“…Breakdown of the Bicameral Mind..”) was home much of human existence can occur with no actual “consciousness” as we understand it.

    If have any clout on this excellent blog (I’m a newbie), would you consider commenting on distributed on bodily “AI”. When I reach for a coffee cup, I really don’t think much human intelligence is involved.

    Jumping ahead, I have not yet seen it, but supposedly the new “Westworld” series hints at Jaynes’ theory, also with connections to Philip K. Dick.

    Liked by 2 people

  2. Interesting thought – while consciousness is not something I think AI will reach with our current understandings, the coffee cup reach is only meant to illustrate the multiple subsystems involved in a task. Distributed AI….now there’s a topic I haven’t heard in years (we talked about it a lot in the 90’s in my circles 🙂 )

    Thanks for reading!

    Liked by 1 person

    1. human intelligence is distributed; look how much gets compiled and offloaded into the grey matter in the spine (the reflex of reaching out your arm, the reflex of gripping your fingers, the reflex of tipping the cup – before the sympathetic nervous system controls swallowing).

      Liked by 1 person

      1. Mary, at my age, I’m not so sure my nervous system remains sympathetic.

        I was thinking about Buck’s comments about the 90’s, and in fact, everything I might think I know about AI dates back to 1968 (Space Odyssey) and then in CompSci classes in the very early 70’s. However, I do remember the issue with grip, and especially an experts ability to catch a baseball.

        As I remember it, there is a specific travel time from the “command” based on a visual sighting of the ball, and a very quick reading on speed / trajectory / wind. The timing on the neural circuits is such that it is not possible for the brain to fully control the arm, and the conclusion at the time was that brain sent a command to put the arm in the computed receiving point at the time the ball would be arriving.

        Rather than the fake robots you see on television or the street performers who perform linear movements until reaching a noticeable stopping point, the ball player’s arm performs a very quick ballistic movement. Only possible because control is local and signals travel much shorter distances.

        Years later, reading Seymour Cray’s book, he says most people think designing a super-computer requires a super-intellect, but the real engineering challenges are in creating extremely short travel paths within the processor, and next is heat dissipation. And these are electrical signals, much faster than neurons.

        Like

  3. Yeah, a robot coffee drinker would not need to be conscious, in fact from my early days working on a Ford production line (summer job) I would have preferred to have been unconscious at the time, yet fully functional. There is some truth to the phrase “bored to death”. Others solved this problem in other ways, but drugs and heavy machinery do not mix well.

    Like

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.