Episode 61 – Artificial Intelligence: Supercharging Project Management

NICK WALKER:  Welcome to Manage This, the podcast by project managers for project managers.  Every other week we get together to talk about the things that matter to you as a professional project manager.  And it doesn’t really matter whether you’re a PM veteran or someone simply exploring what the field is all about.  We’re here to offer some ideas, some perspective, and draw on the experiences of others who have been down that road and have realized success. I’m your host, Nick Walker, and with me are two who are still on that road, Andy Crowe and Bill Yates. ANDY CROWE:  Thanks, Nick.  We’ve had so much interest in the topic of artificial intelligence within project management, and we’ve got somebody here who knows a lot about AI who’s going to be processing that with us. NICK WALKER:  Our guest here in the studio is Chris Benson.  He’s an artificial intelligence machine learning strategist, a solution architect, and a keynote speaker who specializes in deep learning.  That’s the computation technology that is driving the artificial intelligence revolution. Chris is the co-host of the Practical AI podcast, produced by Changelog Media, designed to make artificial intelligence practical, productive, and accessible to everyone.  He’s the founder and organizer of the Atlanta Deep Learning Meetup, one of the largest AI communities in the world, with nearly 2,000 members.  Chris, it’s great to have you here on our podcast. CHRIS BENSON:  Thank you very much.  Happy to be here. NICK WALKER:  Could we start off by just defining for our listeners what artificial intelligence is? CHRIS BENSON:  So artificial intelligence means a lot of different things to a lot of different people.  In my view it’s really a marketing word more than it is anything else because over the years the definition of artificial intelligence has changed and evolved.  So what you might have thought of in the 1980s is vastly different from what it is in 2018.  So before I define it, I want to point out I was in a group of artificial intelligence experts that Adobe was hosting about six weeks ago.  And in doing that, they asked us all that same question; and all 10 of us gave 10 different answers. ANDY CROWE:  Well, and the joke is, if you ask two economists for a definition, you get three answers. CHRIS BENSON:  Absolutely. ANDY CROWE:  Same idea, huh. CHRIS BENSON:  Yup.  So it was very much that.  So I wanted to note that.  Take what I say with a grain of salt. ANDY CROWE:  What do you think it is, yeah. CHRIS BENSON:  So what I think it is, is a narrow definition.  I would consider that in 2018 artificial intelligence is synonymous with deep learning, which is the application of deep neural networks. ANDY CROWE:  Interesting.  Well, learning is certainly a part of AI that I think that’s almost a universal component that goes across most definitions.  Most definitions talk about the ability to imitate intelligence and things like that, imitate human intellect.  But that ability to learn and grow as a neural network is an interesting part of it.  So how do machines learn? CHRIS BENSON:  So there’s different techniques.  And those all broadly fall under the definition of machine learning.  The thing that separates deep learning, which is how I’m defining AI, from the rest is that it can take an enormous number of inputs – we call them “features” in data science – and process them in a highly nonlinear manner and give inferences, which are essentially probabilistic predictions on what the answer might be. For instance, to make it real:  If you have machine vision, and you are putting a cat in front of the camera, and it will come back and identify that it thinks it’s a cat.  It might come back 97 percent.  But the difference is these technologies aren’t going to come back with 100 percent.  They’re probabilistic technologies.  But they can make these identifications using a model that is many orders of magnitude more complicated, and therefore in some ways more capable, than previous models of machine learning. ANDY CROWE:  I have a funny comment to that end.  About two weeks ago I looked across the street, and I saw something, and it’s funny what your brain does when it doesn’t have something that fits a pattern or that makes sense.  And I saw a cat coming across a street.  It was a couple hours before dark.  And I looked at it, and this cat was enormous, and it was walking funny, but my brain’s telling me, well, it’s a cat.  And I called my wife over to see.  It was a giant raccoon coming across the street. CHRIS BENSON:  Oh, okay. ANDY CROWE:  And, yeah, now that’s a daily ritual.  That raccoon crosses the street.  But it’s interesting that you say that, that it’s not 100 percent certain, because I was pretty certain, and I was wrong. CHRIS BENSON:  Yes.  I mean, there’s an analogy to be made there is that, saying this very loosely, neural network technologies are essentially modeled after the brain, a mammal’s brain.  Not just a human brain, but any mammal’s brain.  The cerebral cortex, specifically.  And so, with that said, you can take sometimes tens of thousands of inputs into that.  And, yes, we make mistakes.  And just as we make those mistakes, today’s neural networks make those kinds of mistakes all the time before the training gets right. BILL YATES:  I think it’s worth noting that your wife pointed that out to you, that mistake. ANDY CROWE:  She did not.  I self-corrected. BILL YATES:  You self-corrected. CHRIS BENSON:  He got there before she could get to it. ANDY CROWE:  As we talk this through, project managers are looking at this idea of AI.  And a lot of people believe that it may have an earlier impact on project management than some of the other domains.  Which is interesting to me.  I’m not sure I agree with that.  What do you think? CHRIS BENSON:  So I don’t know if it’s having an earlier impact because talking with people all the time about this, I see it having an impact everywhere, in just about every industry on the planet.  And matter of fact, I haven’t been able to come up with an industry that I don’t think will be impacted in the years ahead.  Some maybe sooner than others, but you’re already seeing it across medicine and transportation, financial, you know, security, you name it.  It’s already starting to have a place.  Machine vision’s everywhere.  Natural language processing is everywhere.  These technologies are becoming pervasive.  We’re all using it every day, every time you’re doing Google searches or using your email or whatever.  So it’s already affecting our lives in a profound way. BILL YATES:  That’s true.  Even in my home, you know, I think of my friends Siri and Alexa.  iTunes is getting smarter; whatever streaming service has these recommendations and suggestions.  It’s as if they can see inside of me. CHRIS BENSON:  They literally know more about you than you consciously do yourself in many ways because everything that you do is data for them, and it is constantly crunching that data behind the scenes. ANDY CROWE:  Well, so now that gives us an interesting transition because project managers are also supposed to predict, to some degree.  That’s an important part of our job.  It’s not all of our job by any stretch, but it’s an important part, is to look at things going on and to spot some signal in the noise, if you will, or some trend that maybe the team doesn’t even consciously know yet.  Maybe the customer hasn’t picked up on this.  Maybe the developers don’t know.  But the PM sees it.  That seems like a pretty natural fit for AI. CHRIS BENSON:  It’s a very natural fit, especially so – and there’s a question that I’d like to even pop in before that, and that is, what will AI do well for us in general?  And that is today very specific problems that are highly complex.  So if you have many, many different inputs that come into a problem, but you’re narrowing the scope of what it’s trying to accomplish to something that’s very specific, then in many cases we’re seeing AI technologies that are improving upon even human experts.  And I would say that that is likely one of those. ANDY CROWE:  So that’s interesting to me that AI is sort of tuned toward very specific and very complex problems.  The human brain is amazing at general things.  Not everybody can make change, you know, for a $5 bill.  And so it’s kind of funny that the human brain can do a lot of broad things, but not everyone is really good at super complex things. CHRIS BENSON:  Yeah, you’re making a great point there.  And that’s that we should not think of today’s neural networks as analogous to an entire brain.  So you could think of it as a very small collection of neurons in your brain that has been trained through your own activity to do a very specific task or identify something.  That’s what today’s neural network would be.  So if you were going to get to a level of complexity in dealing with daily life, where you’re doing that kind of generalization, that would be like having lots and lots and lots of deep neural networks that are all put together to sort of simulate what your brain is doing. ANDY CROWE:  Well, and psychologists tell us those neural networks sometimes compete, as well. BILL YATES:  Right. CHRIS BENSON:  They do. ANDY CROWE:  And that’s interesting from my standpoint, that then you have some kind of function that prioritizes those things and knows which ones to listen to and which ones to tune out.  That’s fascinating. CHRIS BENSON:  Yeah, our own brain creates all sorts of noisy signals because each little piece of our brain is being trained for specific things, and they don’t always go together well.  And so just like that, that’s actually in robotics right now we’re seeing that,