TechRepublic’s Karen Roby spoke with Noel Calhoun, CTO of Interos, an artificial intelligence supply chain solution, about AI in the supply chain. The following is an edited transcript of their conversation.
Karen Roby: Noel, we’re going to talk a little bit about AI today in our supply chain. You spent many years in the public and the private sectors, working with the CIA. When we talk about our supply chain, I mean, never before has the light been put on it as much as it is right now. I mean, I think the vulnerabilities are really showing through.
Noel Calhoun: It absolutely is true. I mean, basically every day some event is happening that is showing companies how fragile their supply chains are. It’s so complicated and so convoluted sometimes as to where all of your material, where all of your software, where all of your services come from, that a lot of people have a tendency to put it in a box and pretend that it’s working fine until the day it’s not. Then they’re made really brutally aware of how fragile it is. I think that’s what we’ve seen over the last year, year and a half, especially with COVID, but with trade wars, with China, with software attacks, like SolarWinds, with ships going crazy in the Suez Canal. I mean, it’s just a never-ending string of things that really disrupt the flow of material and services and goods. That’s been something that has brought the issue of supply chain really to the fore, I think, in the last six months especially.
Karen Roby: Talk a little bit about how is machine learning helping with the supply chain? And down the road, how much more can it be involved to really make a difference in terms of securing our supply chain?
SEE: Healthcare is adopting AI much faster since the pandemic began (TechRepublic)
Noel Calhoun: There is a lot of talk about AI and machine learning. I think they’re siblings in a way. AI tends to get a lot of buzz because people didn’t think of Terminator, and Sarah Connor and protecting the future, and all these things. I know certain celebrities have helped propagate that to some degree. But really I think from a practical perspective, what we’re really talking about is the ability to develop machine learning software that can interpret the facts, the things that are going on around the world in a way that a human would. Then being able to figure out what’s anomalous, what’s weird. What has happened that maybe you haven’t seen happen before, and isn’t a trend that you don’t really want to go that particular way. Identifying that and alerting you to that, or making it really clear when something happens that you can really quickly investigate something without having to spend weeks digging through a bunch of information.
When I was at CIA, everyone, all analysts at CIA are considered all-source analysts. Which means that you’re not focusing on one particular type of information, human-reported information, signal-intercepts from NSA, or you’re not recording. You’re not thinking about any one particular type, you’re thinking about all the types of information that you could possibly bring to bear on a problem. That’s the way I think machine learning and the current trend is to do an all-source analysis approach toward the supply chain. Is it a news story? Is it a weather event? Is it the position of a ship in the Suez Canal? Is it telemetry on satellites? What information do you need to give you the insights on your supply chain?
Any one of those streams would be too much for a human to process constantly day in and day out. You would have to hire tens, if not hundreds, or thousands of analysts all day to be looking at the information, to figure out, what’s going on in my supply chain and how does this affect me? Everyone realizes that’s not doable. I think machine learning really is the solution there where it’s not perfect, you’re going to have errors. You have human errors too, so it’s not like humans are perfect. But you apply machine learning to basically process all the information and give you a superpower to really observe everything that’s going on without having to invest all that human research and effort into it.
SEE: AI in the OR: One company is closing the gaps in surgery using technology (TechRepublic)
Karen Roby: I know when you talk about the acceptance, do people really get that, that this is where we are, and this is the technology we need to look to? Because there are so many layers here. I mean, and like you said, you just can’t wrap your brain around how many humans it would take to analyze all of this on a daily basis. Is the level of acceptance there?
Noel Calhoun: It’s interesting, I think, I would almost say the level of acceptance is there, but in a very uninformed way, I think people are looking for a silver bullet, and saying, can AI solve my problem? Can machine learning solve this problem? The answer, for those who have worked in this space for a very long period of time is, it is one tool in the toolbox, and it’s a very important one, and one that probably has not been used as much in the past. But just as equally important is how you apply that and the data that you apply to that problem. It’s being able to pull in all the right data at one time and in real time analyze it. That in and of itself is a challenge.
This is a combination of machine learning, which everyone gravitates to and says, “OK, if I could just apply some AI to this, I would solve my problem.” But what they ended up finding out is that’s the shiny surface of it. Underneath is months and months and months, if not years of grunge work going through data and combining it together, putting it into a place where you can analyze it and apply the machine learning algorithms to it. That’s a little bit of an education for most people. You’re not three or four months away from a magic solution, if you can just get some AI, bring an AI team in and apply it to your data. There’s a lot more to it. That’s where, from my perspective is, always, at the CIA, at Kensho my previous company. Today it’s always been that data, that data groundwork, that data plumbing that ends up taking up a big portion of your time, and that’s true for, I think for most machine learning engineers.
SEE: Company uses cognitive neuroscience to help train police officers (TechRepublic)
Karen Roby: 2020 and now into 2021 have been teaching us a lesson when it comes to our supply chain. What has it really taught us this last year?
Noel Calhoun: I think it’s taught us you can’t take anything for granted. I think probably the worst thing to happen to the supply chain was probably Amazon because you basically got used to the fact that you could request something from Amazon and it would show up magically the next day or the day after. What the last year and a half has shown us is that even Amazon is dependent on these things. I went looking for a piece of woodworking equipment and it doesn’t ship until 2022. I mean, that’s how messed up some of the supply chains are right now. It’s just amazing machinery. The demand on manufacturing, the demand on machinery, idling car factories for weeks because there are no microchips. You just can’t take for granted that if it’s needed and you have money to pay for it, it’s just going to show up. It may not.
Karen Roby: Yeah. I think my teenagers need to learn that lesson because that’s all they’ve ever known, right? They click on their phone, “I want something.” In three days, two days, some of them the next day, it’s on the front porch, right?
Noel Calhoun: Exactly.