Sam Charrington of TWiML&AI: Thinking AI is Magic is a Dangerous Proposition

If I had to pick one overarching theme from the 15 or so CRM industry events I’ve attended so far this year it would probably be AI. More specifically it would center around the crossroads of where automation and AI intersect.  And during last month’s PegaWorld user conference, I had a chance to speak with the founder and host of the very popular This Week in Machine Learning and Artificial Intelligence podcast Sam Charrington. In just over three years, Sam has amassed over five million show downloads.

Sam shares his thoughts on where we are today with technologies like machine learning. He discusses AI and natural language processing. And he talks about how companies are just scratching the surface in putting these tools to work to improve all aspects of their businesses. But they have to be implemented with the proper perspective.  And he also addresses another theme that I heard addressed at multiple conference this year; how AI will eventually lead to AE, artificial empathy.

Below is an edited transcript of our conversation.  To see the full interview watch the video, or click on the embedded SoundCloud player.

Sam Charrington of TWiML&AI: AI is Not Magic

Small Business Trends: You hear about AI, you hear about machine learning, you hear about deep learning. What is the biggest misconception about this area that just drives you crazy every time you hear it?

AI is Not Magic

Sam Charrington: That’s a good question. I think generally, the thing that drives me crazy is there are all sorts of statements that give you the impression that the person who’s making the statement thinks that AI is magic. It belies that they don’t really understand the process behind making it all work. And that certainly drives me crazy, I think. In part, because it’s a dangerous proposition.

I think one of the things that was talked about in Rob’s [Pega VP of Decision Management and Rob Walker] keynote was some of the dangers of AI bias and things like that. And if you don’t understand that these models that we’re putting into use are created by data, and that they pick up on biases that are inherent in that data, then you probably don’t know that that’s something that you should be thinking about and trying to manage.

Small Business Trends: Right. You talk a lot of people in this space, and a lot of companies that are implementing AI and machine learning. Do you feel like they’re actually ready to take full advantage of this? What are the things that they possibly aren’t prepared for as they start going down the road with this?

Machine Learning Usage is a Spectrum

Sam Charrington: Yeah, it’s definitely a spectrum. As it always is, right? There are some companies that are more mature than others. One of the things that I’m seeing a lot of, and have been speaking a lot about recently, is that among large enterprises, a lot of them are in this really interesting place where they spent the last two or three years, maybe five depending on how early they were, doing machine learning and AI, proof of concepts, building out … they’re building out their initial data science organizations and experimenting with building models, and often the use cases that they’ll start out with are things like churn prediction or recommendation systems or things like that. It depends a lot on the industry.

But they’ve got a handful of these proof of concepts up and running, and they’ve been selling within the enterprise, and they’re all at this place where all of a sudden everyone’s bought in. They’ve had some interesting early results, executives are reading about it on airplanes, what’s your AI strategy, and they’re seeing these interesting results and they’re like, “Okay, let’s get AI in everything.”

Most Businesses Aren’t Ready

And most of the enterprises that I’ve talked to aren’t really ready to scale that up, and they don’t know how. And understanding how to scale that up requires thinking differently about your processes, but also your tooling and platforms, than just adding on another snowflake project.

So this topic of AI platforms is one that we’ve been covering a lot on the podcast, one that I’ve been writing a lot about and publishing some eBooks on. In fact, we just announced that we’re hosting a conference in the fall on the topic.

Small Business Trends: Nice. Where’s that going to be?

Sam Charrington: It’s going to be in SF, San Francisco.

Small Business Trends: Oh cool. When are the dates?

Sam Charrington: October 1st and 2nd in San Francisco. It’s called TWIMLcon: AI Platforms.

Small Business Trends: Very cool.

Sam Charrington: Yep. Yep.

Small Business Trends: So you’ve been following this, at least through your site, for a little over three years now. What have been some of the biggest changes in the space? Not necessarily the perceptions that companies have of it, but how has it changed in your eye, as you’ve been watching this?

What’s the Latest in AI Trends?

Sam Charrington: There are so many interesting things happening. Three years ago, it was just an idea. You talked to companies about this idea and they’d be intrigued, but not really understand what you’re talking about. Nowadays, there’s usually someone or some … usually more than one, usually a group of folks that are really working on this stuff now.

So I think in that way, in that sense, it’s become real from an enterprise perspective. I think from a technology perspective, it’s also moving incredibly quickly. So three years, we’re starting to have some interesting results in areas like the application of deep learning to computer vision. And now, I was here in Vegas just a few months ago for CES, and every other booth at CES had a camera, was showing bounding boxes around detected objects in a video. The technology to do that kind of thing is really becoming commoditized.

And a lot of what made that interesting is this technique called transfer learning. Basically the ability to train a model and apply it to other types of models or other datasets. Recently there’ve been a bunch of developments in applying that same idea to natural language processing. So on all fronts, the space is moving really quickly.

Small Business Trends: So what’s been the most interesting, maybe unique, but also the most successful use case you’ve seen a company actually implement and see some positive results?

Examples of AI at Work

Sam Charrington: There are a ton. I have a slide in one of my presentations that, kind of tongue-in-cheek, says that if you want to find an application for AI, just throw a dart at a dartboard. There are so many. I talked about some of the sales and marketing use cases, like churn prediction or recommendations, identifying next best offer. That’s what Pega’s all about, talks about frequently. But there are applications in logistics, understanding and optimizing your supply chain. Tons of applications there. Certainly for companies that are in industrial sectors, I wrote an ebook on industrial AI a couple of years ago, there are lots of applications in robotics and IoT. The whole idea of digital twin is becoming an interesting one.

Small Business Trends: What’s that?

Businesses Keep Collecting Data

Sam Charrington: Basically, for companies that have significant industrial assets, whether these are generators or oil rigs or aircraft engines, a GE aircraft engine has several thousand sensors in it, just a single engine, and so they’re collecting a ton of data, and a company like GE is trying to shift their business model from one of selling this physical asset to selling engine as a service, if you will. So part of what allows them to do that is the ability to anticipate when the engine is gonna fail. Predictive maintenance is a big use case.

And one of the techniques that they long evangelized for doing that is what they call digital twin. It’s basically kind of IoT plus AI plus simulation. So you pull all of the sensor data off of your engine, you use it to create a set of models about the engine that you can use in a simulation environment to predict how different factors are going to impact the performance of the engine.

Small Business Trends: So in other words, as more and more time goes by, there’s more and more use cases coming up.

And the Hits Just Keep on Coming

Sam Charrington: For someone like me that just loves to learn and understand and study, it’s like being a kid in a candy store, there’s so many ways that folks are applying this stuff.

Small Business Trends: So if we were to look out two years, or I don’t know, maybe five years, whatever the timeframe is, what do you see happening in the future, in the near future, when it comes to this area in terms of uses, use cases, adoption and success in companies?

Sam Charrington: Mm-hmm (affirmative). Yeah, I think that in five years, we’ll be beyond this stutter-step point that I mentioned where enterprises have figured out how to do the one-off machine learning project, and they’ll have the platforms and processes in place to repeatedly put models into production.

One of the observations that I’ve been making is that it’s a mistake to think of machine learning and AI as like another technology. And I think it’s more another wave or frontier in business. So this other sequence that I have in the presentation I gave yesterday talks about how in the eighties and nineties, it was all about being process driven. It was like, let’s get our business processes out of managers and line folks head and document them. And then we moved into this era in the 2000s that was all about being data-driven. Hey, we’ve got these processes, can we automate them, can we instrument them and pull some data and use that data to help us make decisions. But it’s usually like, let’s create the TPS report every quarter, have somebody …

Small Business Trends: I love that movie by the way…

Latest AI Trends Include Machine Learning Models

Sam Charrington: …So now we’re shifting into this era of being model-driven, and model-driven, in this case, refers to creating machine learning models that pull patterns and insights out of this data and puts it in a real production so that we’ve dramatically decreased the lag between an insight and decisions being made, because the machines are making the decisions. Sometimes with our help, people’s help, but more and more of these decisions will be seated to online systems that are making decisions at the point that they’re required.

So I think this a is a new idea now, but in five years, this will be, I think, apparent to a lot of companies, and they’ll be very heavily invested in going down this path.

Small Business Trends: And does that possibly then free up humans to leverage their empathy and allow machines to help them with the decision making?

AI Steps from Sci-Fi to Reality

Sam Charrington: Yeah. I think that’s the idea. But going back to Rob’s presentation, there are interesting opportunities to allow the machines to exhibit some kind of empathy. And now I don’t think of that, and I don’t think he does either, like some, what we call AGI, artificial general intelligence, like movie, sci-fi AI, but more being explicit about these trade-offs and baking that explicit knowledge or consideration of these trade-offs into the systems that we build around these models.

Small Business Trends: And also, he mentioned, and I kind of agree in some instances, some people lack empathy just in general. So sometime maybe the machines will be more empathetic. I don’t know.

Can Machines Be More Empathetic than Employees?

Sam Charrington: Yeah. Well, you talked to him, I talked to him about this yesterday, and he’s very quick to say that they’re out evangelizing to their customers to be more empathetic, or at least to think about it as part of their customer experience transformation. But if the company is led by someone who is not empathetic, it’s probably not going to work.

Small Business Trends: Not gonna be a really good ending there. But anyway, let’s hope for the best.

Sam Charrington: Yeah.

This article, "Sam Charrington of TWiML&AI: Thinking AI is Magic is a Dangerous Proposition" was first published on Small Business Trends



RSS Business Feeds

0 nhận xét:

Post a Comment