Brian Markwalter:
If an AI is trying to optimize for something, we have to be really careful about what we’re telling it to optimize for.
James Kotecki:
This is Machine Meets World, Infinia MLs ongoing conversation about artificial intelligence. I’m James Kotecki, and my guest today is Brian Markwalter, Senior Vice President, Research and Standards at the Consumer Technology Association. CTA is the organization that puts on CES, which is the giant consumer tech show every year. This year, it starts on January 11th and this year for the first time, it is all digital. So, Brian, I know there’s a lot of prep that goes into something like this, especially when we’re doing a digital show for the first time. Thank you so much for making time to be with us here.
Brian Markwalter:
Yes, James, thanks for having me. And we are getting ready for CES, January 11th through 14th, we will be there and I hope you’ll be there too, digitally. Should be a great show.
James Kotecki:
One thing that is interesting about CES and has been true for the last several years at least, is this pervasiveness of AI as a theme, which is why I wanted to talk to you. When is it going to get to the point that AI is just so pervasive in technology, so expected that it’s not even going to be something we talk about that much anymore?
Brian Markwalter:
So, we’re clearly in a period right now where it’s both. We see AI talked about because it’s AI, but we also see it embedded in many of the things that are bringing such wonderful personalization to these products. So, we have big parts of the show that are dedicated to smart speakers and voice assistants and interaction. And those whole ecosystems are not presented as AI, they’re presented for the problems they solve. So this is sort of like the IoT trend.
Brian Markwalter:
Years ago, it was a big deal to say you had IoT. You just wouldn’t do that now just — everything is connected. So I don’t know when that point will be exactly, but we have both wonderful displays of AI in action and AI embedded in the products that you have to know what’s under the hood to really understand what’s going on.
James Kotecki:
How do you define AI from where you sit?
Brian Markwalter:
We do standards for the industry. And one of the first ones we did in AI was definitions. I personally, tend to take the broader definition of a machine intelligence that mimics some of the behaviors of human intelligence. And then there are other nuances of whether they learn or not and some of those characteristics.
James Kotecki:
What can you say about the overall trust in AI? What kind of trends do you see there from consumers?
Brian Markwalter:
We do 30 or so studies a year on consumer perceptions. One of the things that we studied this year was perception of autonomous delivery and robotics and other stuff. But what showed up is people’s favorability and willingness to adopt, use — or just overall acceptance of these technology tools like autonomous delivery, which clearly are using AI. So that’s another case where people see the service being provided, probably don’t really think a lot about, oh, well that’s — it’s got machine vision, all sorts of underlying bits of AI in it.
Brian Markwalter:
So, we’ve definitely seen a little favorability increase in the technology and some of these assistance-type devices. We’re at a good point where people quit talking about AI for AI’s sake itself and focus more on, are we doing this right? Is it helping me? And are these solutions really good?
James Kotecki:
What trends do you see around AI ethics and responsible AI? It seems like those concepts have really emerged to the fore in the last few years.
Brian Markwalter:
Yeah. They certainly have, and maybe that’s the best trend by itself is that so much attention is being paid to this space. We’ve had many attempts at AI, we’ve had the AI winters that everybody talks about. And so when the tools got right and our data handling capabilities got right, we really did reach this tipping point in the last five years or so where it made huge advancements and the tools were generally open-source a lot of people used.
Brian Markwalter:
So we rushed in and we needed to learn a lot and catch up. I think the discussion around ethics and AI and the places where people got things wrong and didn’t fully understand perhaps how AI worked — that it will solve problems, but it’s also a little brittle. You have to be careful about putting constraints around it. So now the fact that people think about having multi-disciplinary teams, thinking about the ethics, thinking about adversarial AI and what can go wrong, using it more as an assistive-type thing with a human in the loop to pay attention. Those are all great trends I believe in the ethical side of AI.
James Kotecki:
When you start having these ethical conversations, sometimes it doesn’t take too long to get to very philosophical places around what are the ethics we’re supposed to be imbuing these things with in the first place, or what are our moral principles in the first place, let alone what we’re trying to get AI to do. Have you seen the emergence of more, call it philosophical thinking among the tech world? Is that a real trend or is that just something that I’m making up?
Brian Markwalter:
I don’t think you’re making it up because if you start down this path, you’d have to end up there. And there’s a lot of work going on trying to understand where does AI fit what we’re trying to do here? So if you’re just plugging it into a business process, okay, you still need to be really careful that you don’t drag biases from the past into your current solution and just amplify them with AI. But AI doesn’t help us with the fact that we all have different cultural backgrounds, different sensibilities around privacy, different judgment. So, if an AI is trying to optimize for something, we have to be really careful about what we’re telling it to optimize for.
Brian Markwalter:
It sort of goes back to Alan Turing, the Turing Test, famous early researcher. That paper where he proposed the Turing Test sort of begins saying, “Well, we really don’t understand human intelligence, so how are we going to decide when a computer has humanlike intelligence?” And then he just kind of posits this Turing Test, this language test, as a way to think about it. But we’re not going to get past that. We are humans and we have these machines that are working with us. So our judgment matters.
James Kotecki:
What’s an AI trend that has most surprised you in the last few years?
Brian Markwalter:
One, I think we still have a lot of promise to unfold is in healthcare. Clearly a lot is being done and a lot’s been written about particularly like using AI for medical imaging and everything. But there’s just so much to be done with, in particular, like wearables and AI and personalized apps, all in the interest of helping people stay out of the chronic conditions and out of the healthcare cycle. Somewhere between a fifth and a sixth of our economy goes to healthcare. There’s just so much benefit to be derived there.
James Kotecki:
What’s something that business leaders need to understand about AI as we head into 2021?
Brian Markwalter:
Yeah, so I think we are developing a lot of more understanding about the business implications. I’ve seen some studies, I wouldn’t say they’re contradictory, but what I do see is there’s quite a few companies who spend on AI that don’t feel like they get their ROI. But there’s this analogy that goes with it that suggests, well, some companies are sort of getting the lion’s share of the benefit. And I’ve seen some economists talk about this J adoption curve. Like, you do your investment, it feels like your ROI is going down and then once you really get it pulled into your systems and operational.
Brian Markwalter:
So my advice would be, it’s not just an unplug one thing and plug in another. The research suggests that the companies who are really good at monetizing it are also good at data and they have a holistic strategy for making sure that their workforce is trained, that people know what to do, that it’s treated holistically. So I think that would be the advice is, look at the big picture and have a roadmap and plan for it in your whole organization.
James Kotecki:
That J shape is a pretty scary curve, right? Because you start by going down and you think to yourself, “Well, this is expected. It’s just the first part of the J curve.” But you don’t know whether you’re going down because you’re on that J or because you’re just on a downward slope.
Brian Markwalter:
Yes, well, on the other hand, the successes are piling up. It’s clear it’s here. I mean, just too many companies are accomplishing too much with AI. Retail industry and banking are the biggest spenders on AI and they’re using it on the retail side. So, the sales and marketing cycle is the biggest beneficiary for revenue, for just straight-up payback on AI investment and that has to do with pricing, prediction of purchases, putting the right products in front of the right customers. All that is very tangible, very short-term ROI.
James Kotecki:
Brian Markwalter, SVP at CTA, the Consumer Technology Association, which puts on CES, which is coming this very January. Thank you so much for being with us on Machine Meets World.
Brian Markwalter:
And thank you, James. Very happy to be here.
James Kotecki:
And thank you so much for watching and/or listening. You can like, subscribe, share the show, you know, give the algorithms what they want. You can also email the show, it’s mmw@infiniaml.com. I am James Kotecki and that is what happens when Machine Meets World.