Podcast: Real world tech: Edge AI drives car-making, healthcare and retail
Artificial intelligence (AI) at the edge is changing healthcare, retail and Audi cars, as Intel’s IoT Group vice president, John Healy tells Jeremy Cowan and George Malim. It’s cutting auto safety inspection costs in half, speeding life-saving cancer diagnoses, and creating Covid-safe shopping in Latin America. Plus we learn how chipmakers globally are tackling supply problems that have halted vehicle production. The semiconductor industry is facing an “awakening”, says Healy, as it shape-shifts to meet “insatiable demand” for silicone. Finally, we hear which African country is a leader in satellite cartography, and how Amazon is playing games with its warehouse staff.
Jeremy Cowan 0:04
Hi, and welcome to the latest Trending Tech Podcast brought to you by The Evolving Enterprise, IoT Now, and VanillaPlus.com. This is Jeremy Cowan, and I want to thank you for joining the latest, sometimes serious, sometimes light-hearted look at enterprise digital transformation.
I am delighted to welcome today two guests, who are John Healy, from California-based international technology company, Intel, known among other things, for the processors that power so many of our devices. John is vice president of the IoT Group. He’s also general manager of platform management and customer engineering. So, they’re clearly keeping him a busy man. John, thank you very much for making the time to be here.
John Healy 0:58
Jeremy, thank you. Delighted to be here.
Jeremy Cowan 1:00
And it’s a welcome return to the excellent baritone of IoT Now’s managing editor, George Malim. Good to have you on again, George.
George Malim 1:08
Great to be here. Thanks, Jeremy.
Jeremy Cowan 1:10
Okay, today, we’ll be looking at some key tech news stories that deserve a bit of a deeper dive. And after that, we’ll be asking John to spill the beans on what’s really going on in artificial intelligence at the network edge. I think he can share details of some great IoT and other applications in areas as diverse as car making, healthcare, and retail. And we want to know what the benefits of edge AI are in the real world.
Then in our final section called What The Tech, we’ve got some extraordinary technology stories to bring you, the sort of thing guaranteed to make you raise an eyebrow. So, George first. What stood out for you in the news recently, George?
George Malim 1:59
Well, Jeremy, I think it’s the global micro-chip supply shortage that’s dominating everything at the moment, really, I think it’s really highlighted how essential microchips are to virtually everything we use, and how fragile the just-in-time manufacturing process is. The shortage is really being caused by the pandemic, but a kind of misread of what the impact of the pandemic would be. I think most people modeled their expectations based on the financial crisis of 2008 and the downturn that had in consumption, so people stopped making microchips because they weren’t expecting people to buy products. However, people stayed at home and started buying technology, from bigger display screens, new PCs, tablets, phones, games, consoles, milk frothers. Yes, they have microchips. (Laughter) And car manufacturers shut their factories and told suppliers to stop shipping to them because demand for cars dried up.
Suddenly, though, demand returned, when lockdowns started to lift and people needed new cars, there was pent-up demand. And there was still pent-up demand for homeworking equipment and all these other items that people have liked. I think it’s really important here that it’s not necessarily the cutting edge chips that are enabling very complex new business cases. It’s the cheap stuff, the mass market stuff that’s sold by the bucketload. And in everything you know, such as display drivers, and you have these kind of routine products that have basic chips in them, and maybe two or three generations behind the cutting edge. This is an important point though, because these non-cutting edge chips are made in foundries that typically use eight-inch silicon wafers. And the market has been moving towards 12-inch silicon wafers. And this means that there is actually constraint on eight-inch silicon wafer production. And in fact, there’s less capacity now in the market than there was in 2017.
Yeah, it’s obviously affecting everyone in technology but the car makers are really affected. Alex Partners has said they expect to see global automotive industry revenue cut by US$60.6 billion this year.
Jeremy Cowan 4:48
Wow.
George Malim 4:59
And research firm IHS Markit anticipates 672,000 fewer vehicles will be produced in the first quarter of this year due to the shortage. So, it’s obviously broadsided people, with the exception of Toyota, who, after the Fukushima disaster that was caused by the tsunami in 2011 realised that the lead time for semiconductors was way too long to cope with devastating shocks and national disasters – which I think we can all agree last year was a devastating shock, if not a natural disaster. Toyota has come up with a business continuity plan that means it holds two to six months’ worth of chips, depending on the time it takes for delivery from suppliers, but it leaves everyone else who didn’t foresee the global pandemic of this scale – so, basically everyone – stuck in a tense battle to get the chips that they need. That’s going to be ongoing while the foundries ramp up production, and people try and figure things out.
The problem is we’re seeing organisations like GM with kind of half built cars, because you can’t ship your $65,000 car for the want of a display screen chip, for example, and I think it’s just comes down to the fact that the supply chain is fragile. And we need to look at how to build redundancy into that, without it costing too much. That leads me to conclude that if you’re really stuck, you might be able to find a few chips on a certain container ship in Egypt at the moment. (Laughter.) We’ll see how that goes.
Jeremy Cowan 5:51
Or it goes nowhere, and not very fast. (Laughter.) I mean, 20 years ago, I guess we were all staring in admiration at the rollout from Japan of just-in-time as a philosophy. And now it seems, you know, that has had its day, or at least its fundamental weaknesses have been exposed. I guess you could say, “Sure, the car makers were unlucky.” I mean, who would have said that they’re fairly sensible and cautious forecasts last year, were anything other than that, sensible?
John, is there a way out of this short term problem? Because it’s obviously going to be doing long term damage? And not just to car makers?
John Healy 6:34
It’s interesting you mentioned ‘who could have foreseen it?’ I think the shock that COVID-19 brought to all of us was totally unprecedented and response to it, in terms of, you know, demand for equipment that George mentioned, and the spending patterns that we’ve seen, were something we really wouldn’t have forecast. In fact, I think a lot were forecasting the complete opposite, that there will be massive declines in demand. And so I think the return from this is a number of things. It was an increase in capacity, manufacturing capacity. You know, we recognise that, frankly, semiconductor manufacturing capacity and the numbers of jobs required and the investment in the infrastructure to support it really underpins economic infrastructure, underpins economic stability. And we’re heavily committed to that expansion, we announced some growth in our own capital investments a few weeks ago in building more capacity in the US and beginning to build out more capacity in Europe as well. Because frankly, that’s what’s required. You know, we need to create more global capacity for the supply of semiconductor. And then as a result of that, being better equipped to support these spikes in demand as they occur, because what is clear, is that everything that is connected is computing and everything that’s computing is connected. So, the demand for and the insatiable, you know, requirement for computing all forms of computing and underpinning those are the semiconductors, the silicon chips is going to continue into the future. So, I think it’s been a salient awakening for us that we need to respond to this in a much more progressive way now for the future.
Jeremy Cowan 8:10
I guess the landscape in chip production, supply and design is going to change considerably. I mean, 10 or 15 years ago, this would not have been an issue for anything other than the high-end vehicles. But that’s no longer the case. I mean, anything on four wheels and two, will now be properly connected.
So, yeah thanks, John. I think that’s fascinating. It’s a little glimpse inside what’s being felt by chipmakers.
The news I wanted to focus on is a positive tech story that I found on www.TheEE.ai. And if you haven’t already seen it, you can find it by searching under “automation”. The site is The Evolving Enterprise’s site, to give it its full title. And the story is headlined “IT leaders will lift spending on automation post COVID”. It reports some new research published by Snap Logic, and it shows that 78% of US and UK businesses plan to increase their spend on automation projects over the next 12 months. This comes after a year in which obviously 48% or so of IT decision makers – the researchers insist on calling them ITDMs. Hey, I’m sorry about the inevitable abbreviation but this is tech world. Forty-eight percent of these decision makers have already accelerated their automation initiatives, following the pandemic’s disruption and the new research for Snap Logic – who, for those who don’t know them, make integration platforms – was conducted across both the US and the UK. And it found also that 63% of IT decision makers said cost savings are the main driver behind their enterprise automation. Well, no surprise so far.
But 60% said they were employing automation to increase customer satisfaction, and 59% were also using automation to improve productivity. There was one major difference between the US and UK markets. In the US, over 60% of decision makers adopted automation to grow revenues. But the UK is lagging way behind with only 38% in Britain saying revenues were the motivator for their automation projects. So, of course, automation not only saves businesses time and money, it seems it’s also enabling faster response times to meet new and changing user demands. And it isn’t just happening in IT, it’s affecting different lines of business. Again, there were differences between the UK and US respondents, with those in the UK almost four times more likely than their US cousins to focus their first projects on financial or sales processes.
Now, I don’t know about you, but I find it encouraging that nearly all the replies – I think it was 98% – said you have to take an enterprise-wide approach to automation. You can’t just rush into one-off isolated projects that won’t scale up. But, and there is always a but, it’s not all wine and roses. There are hurdles to overcome, because when asked if anything was hampering the rollout of their automation initiatives, 55% of IT decision makers cited the inevitable legacy technology, 40% said they lacked internal skills. It’s the old training issue, and another 40% were being held back by the pandemic’s shift to remote working. John, what do you make of these findings? Do you think they’re significant?
John Healy 12:18
I think they’re significant, they’re not surprising actually. We’ve worked with our partner and customer base, we see this continual move towards forms of automation. And I’ll talk a little bit about that later, in terms of even forms of AI (artificial intelligence) deployment, which really are replacements of what might have been manual processes or very human-centric, resource-centric processes, as a way of improving the business, frankly – you know, from everything from decision-making through to implementation and production. Though forms of automation, where there are mundane, repetitive in some ways, maybe even, you know, standardised processes and tasks, just makes for a more effective business, but also improves on the ability for, as you said, customer service improvements and the path or platform to differentiated services, so-called top line improvement. So, I’m not surprised by it. It’s very encouraging, because I think that as the technology continues to run forward and leap ahead in terms of what it becomes possible deployment of automation, where we can and for the benefits that can be derived from it is super important that it’s happening across all industries.
Jeremy Cowan 13:26
Yeah. George, any other thoughts on that?
George Malim 13:29
Yes, Jeremy. I mean, I think it’s aways a difficult thing when you come to a major shift, and it is getting the people’s mindset altered that is always the difficult piece. And I think part of the problem, as you pointed out, is people are dispersed because of the pandemic, so it’s much more difficult to I mean – yes, we can do Zooms and all that sort of thing – but it’s much more difficult to drive a company-wide strategy towards automation, because it does involve the whole company, as you said. So, I think that’s, that’s the issue, isn’t it? As with anything, anything new is met with suspicion and perhaps fear because automation, the immediate knee jerk reaction is, is my job safe? Or will I be replaced by something automated? So, that I think is a cultural thing that needs to be communicated properly and communication is at a premium because we’re all dispersed and trying to work in unusual ways.
Jeremy Cowan 14:26
Yeah, I think that’s interesting that you pick up on that, because from a telco background, I would say that quite often we get told that when there’s a new technology such as 5G, the best thing is go and try it somewhere small, where it doesn’t matter. Go and play with the toolkit. And this is clearly not what is being advised by those who are deploying automation. They seem to be strongly coming through with the advice, “Look, you’ve got to do it company-wide.” And I’m sensing that that means endorsement from across the board, at C-level.
George Malim 15:07
Yes, I agree. I mean, I think it must be that. It’s, you know, it’s a long term, strategic target and direction. And I think it must be from C-level, and then down to all levels, to gain acceptance and adoption, and also to operate properly. It’s new and complex, like so many of the things we talk about, and that needs time for people to digest.
Jeremy Cowan 15:34
Okay, thanks for that. It’s our interview time. And I’ve been looking forward to this. John, we want to get under the hood of what’s happening in artificial intelligence, or AI, at the edge. Can you give our listeners an idea of the range of real-life applications that you’re seeing now for edge AI?
John Healy 15:56
Yes, Jeremy, I’ll share a few really interesting examples across a number of industries, maybe starting with manufacturing. And Audi you know, the automotive powerhouse that we all know, have a very sophisticated manufacturing and implementation and assembly plant. At Neckarsulm they have a plant that’s producing about 1,000 vehicles a day. And manual inspecting quality on those vehicles is an onerous task. Now each vehicle has about 5,000 welds per car, and we’re 1,000 cars per day, that’s 5 million welds. So, it’s costly, and it’s labour-intensive. And frankly, the sampling method was really one of taking a car off the assembly line to a separate room and manually inspecting the welds with ultrasound.
Well, we’ve worked with them and using Intel’s industrial edge insights software solution, Audi and Intel created an edge solution for AI that allows us to help them to do defect detection using machine learning algorithms that are trained for accuracy, comparing predictions with weld quality, comparing the behaviour of the welding machines and allowing them to do 100% inspection across the vehicles, which has improved the cost for them by anything from 30 to 50%, in terms of the of the cost of providing that inspection, and quality control, but also expanded it to 100%. Which leads to a whole range of future benefits of preventive maintenance, proactive maintenance to understand the behaviour of the machines long before issues arise. So very, very significant.
Another interesting case, though, is in the healthcare sector, specifically in medical imaging. In 2019, GE Healthcare used our Vision technology to develop an AI algorithm that helps medical staff in the triage of potentially life-threatening cases and allows them to do it more quickly. In what they’ve described and defined as their critical care suite, it’s a set of AI algorithms that are built to detect critical findings on a chest X-ray. Now, chest X-rays represent about 50% of all medical imaging today. So, the ability to detect critical findings quickly, and to prioritise which X-rays should be top of the stack for the ER doctors to review is significantly important for them to cut both wait times, but also to improve upon workload for busy healthcare professionals. So, it’s not only better in terms of the process of diagnosis, it’s better for getting to results more quickly for the patients that are waiting.
Jeremy Cowan 18:23
And presumably better for outcomes as well, because you’re triaging the right ones to see first.
John Healy 18:29
Exactly, exactly. Speed of diagnosis, and then speed of treatment. And then lastly, is in the world of retail where we’ve a couple of examples actually, recent examples. And you know, in in the world of retail, a lot of the times the video data is key and processing that data right at the edge, which is why edge-based AI is so important. It allows the company to understand in analysing the video feed from their retail environment where their customers are spending their time, and tracking them across the store allows for placement of goods and product. But during the Covid-19 pandemic last year, we worked with Claro360, based out of Latin America, where they operate 100,000 camera feeds in retail environments across Latin America. And we took this same product, same tool, same application that was video processing for retail environments, but it was retooled. And in this case, to allow them to manage social distancing guidelines in a retail environment, processing that data close to the edge and enabling them to make real-time decisions and provide for guidance and management and social distancing within those retail environments. Super exciting, but it enabled them to support the application in the environment that they were already deployed and had their camera feeds deployed. And we’ve worked with them very recently actually, when we launched our third generation of the Intel Xeon scalable processor and saw significant improvements in performance and application right at the edge; critically important because that’s where the video feeds are captured.
George Malim 19:57
That’s really interesting, those use cases. But I’m interested to drill down a bit more into the benefits that Edge AI brings to both your customers as you set out and then users. Could you share a bit more insight into the benefits that customers and end users get from Edge AI?
John Healy 20:18
Yeah, absolutely George. You know, it really is an improvement in workflow efficiency as a primary. That’s when we see that sort of spans across each of the different industries as Edge AI is deployed. And I think of the healthcare example, I spoke to earlier. You know, if a doctor is treating a cancer patient, they oftentimes need multiple comprehensive, detailed body images, particularly through an MRI, to understand if a tumour is spreading, where the tumour is located. And each of these scans produces hundreds of images. So, AI at the edge, assisting the clinician to identify the most critical images is a significant assistance to their efficiency and workflow. They’re analysing those images and identifying which ones need attention. And we found that, you know, with accelerations within the product, particularly in our scalable processors, we’ve provided for support for that AI inferencing, right at the edge so they can gain insight more quickly. And then of course, speed the path of diagnosis. So, it’s an efficiency in their implementation and process as I mentioned with the GE Healthcare earlier. They’re using then algorithms built around and accelerated by our Open Vino computer vision tool suite to enable them to really get greater insight into the X-ray images, as I mentioned, you know, deep inferencing, to enable them to understand those X-rays much more quickly than they could before to get, you know, speed of analysis from from three seconds to under one second. That kind of increase in efficiency to ensure that they can take action on it more quickly. And I mentioned the example with Audi, and we see this in other manufacturers as well, where the ability for them to do inspection work both at 100% of time, but also with up to 50% reduction in labour costs significantly improves on their business flow and business process. And it’s really this improvement in existing processes, because of the application of Edge AI that is deriving at least the initial primary benefit across all of these markets.
Jeremy Cowan 22:18
John, there are clearly quite a number of players in this space at the moment and trying to stand out amongst them is always going to be one of the challenges. What are Intel’s and your partner ecosystems’, particular skills in edge AI? What’s your USP here?
John Healy 22:36
It really starts with the technology and what that technology enables. That’s the, you know, begins with the investments we make deep in the silicon itself to improve on the performance of the types of algorithms, and workloads, and processing needs of these kinds of applications. So, in the processor, you know, acceleration support, like our deep learning boost technology was really designed to enable AI applications to work incredibly well. And then the investments in software, our Open Vino toolkit really accelerates AI inferencing, this need to, to analyse the video stream in a computer vision environment, and derive insight from that really quickly. But it doesn’t stop there.
The technology is great, but you need a way of ensuring that the market can consume it, and that’s where our partners and our relationship with our partners is so key. You know, we invested in a dev cloud for the edge, a Developer Cloud to provide infrastructure for developers to both discover and test different combinations of hardware and software as they’re determining what would be best for their solution, what would be best for their application, in a manner that’s very low burden for them to do because it resides in the infrastructure that we’ve built as a back end as a dev cloud for them. And then we created a software, an edge software hub, as a one-stop resource for edge computing software. I mentioned earlier, our industrial edge insights solution that’s accessible to our edge software hub. So, our developer partners can access that code, can customise it can validate their implementations and deploy them more quickly, and with greater confidence for the applications that they’re trying to serve in the market. So, major combinations of the innovation in the technology and the products themselves, and then how they’re consumed by our partners. And we’re very excited and encouraged by the burgeoning numbers of fellow travellers and partners we have in the market, that are then taking those capabilities and building applications that are specific to the needs of the individual vertical markets, and enabling them to come to market more quickly and more efficiently.
Jeremy Cowan 24:40
Another time I’m going to have to get you back to tell us how you choose partners, because there are an awful lot of partners available. It will be fascinating to hear that, but I’m afraid we’re gonna have to move on because that could take us a whole ’nother podcast. Thank you, John.
We’re into the home straight, the section of the pod called What The Tech. And that’s where we share something tech-based that either amused or amazed us. Go on George, what’s struck you?
George Malim 25:08
Well, I’ve been there preparing an article, a report on the usage of LoRaWAN in private networks, and it’s taken me to some really interesting places. Perhaps the most unexpected was the Tunisian Space Agency. And once I’d got over the fact that such a thing existed, it turns out that Tunisia – although it’s only a small country with a population of 12 million – is extremely advanced. It has a National Commission for Outer Space Affairs, has been doing a lot of IoT (Internet of Things) and telecoms-based activities, predominantly in lower space for many years. The space commission was founded in 1984 but the country also has a Centre for Cartography and Remote Sensing, which was launched in 1988. And it’s great, they’ve launched their first satellite, which was designed and built in Tunisia, although launched in Russia, like many satellites. And it’s in orbit now, utilising LoRaWAN to connect IoT objects on earth.
And the great thing about this is it’s going to really help agricultural yields, things like that. But I think the thing that I’ve taken out of this, aside from just being interested that this level of innovation exists, is that it’s not just California, China and the Nordics that the innovation comes from. I’m delighted to have discovered the existence of the NCOSA and the TSA, and to find further proof that IoT is truly global, and everyone’s engaged in it. So, it’s cheering. I think that’s just been one of the most interesting things about my job over the last year is to, although being stuck in the office, to see what’s happening all around the world. So, I’ve really enjoyed that.
Jeremy Cowan 26:49
Well, congratulations, Tunisia. And I have to say I didn’t see that one coming down the pike. But that’s fascinating and indicative of the temptation sometimes to take a sort of rather First World view or a Western-centric view of these things. It’s a wake-up call for journalists like us to see that there is a lot going on sometimes in places you just didn’t expect.
There’s just time for me to tell you about Amazon. And I’m not talking geography here, I’m talking the company. It’s getting its workers to work faster, Amazon would probably prefer it I said, working happier, you can make your own mind up on that. They’re doing it by gamifying their warehouse work. According to a report I saw on TheInformation.com Amazon warehouse workers in the US are now able to spend their shifts earning rewards by competing in games that are designed to help them complete repetitive tasks quickly, in order to fulfil Amazon customer orders faster. The company points out that workers don’t have to do this. It’s only if they want to; fair comment. Amazon started testing this apparently back in 2017. And doesn’t that seem a long time ago. But it’s proved really successful or successful enough for them to start rolling out what it calls FC Games in fulfilment centres across 20 states in the US. And the company now admits its gamifying work to boost productivity but says it also wants to relieve the burnout – their words – that come from grabbing items from shelves to prepare orders for you and me. So the games are shown on screens in Amazon’s stowing and picking stations. Employees who opt to play one of the six available games can earn digital currency. They can use that to buy and care for virtual penguins and all sorts of other pets. If you’re not interested in virtual penguins, there are other games with names like Dragon Duel, Mission Master and, inevitably, Picks in Space. (You see what they did there.) The information reports that workers see them as kind of symptoms of Amazon’s obsession with productivity. And the teams also say they see them to be fair as a welcome distraction from their work. What did you make of that, George?
George Malim 29:33
Well, I’d rather be a robot, I think. (Laughter.) I’d rather be replaced by a robot. For sure.
Jeremy Cowan 29:39
You may be.
George Malim 29:40
Well, I mean, I look at my 12 year-old and I look how much time he spends playing games. Again. You know, it’s a losing battle. Every parent knows that. The screentime situation, but I mean, I kind of get it because warehouse work is incredibly dull and repetitive. So, I suppose trying to liven it up is on one hand nice for the workers but on the other hand, good for productivity. But it does sort of seem, is this really a job? And if it is a job, how long can it go on for? Until automation comes in?
Jeremy Cowan 30:12
It must be indicative of the need for automation. Because you do wonder how long that is the solution? John, what was your take on this?
John Healy 30:20
I think similar to George, I can guess it. I mean, there’s certainly a generation, a younger generation than me are very game-oriented and competitive. And I think maybe that’s part of what they’re tapping into, is the competitive spirit. It’s not that dissimilar to some of the fantasy football and things that go on, you know, in normal work environments, where folks are participating as a way of just, you know, engaging in a team sport and using it as an opportunity to maybe break some of the monotony of the day. But you know, operating in a safe way, in a safe environment, I guess it’s providing some value for them.
Jeremy Cowan 30:53
I’m sure it is.
Well, that’s it for today. If you’ve seen something amusing, or amazing in the news that you want us to share next time, do let us know. And we’ll give you a shout out. You can reach me on Twitter @jcIoTNow. But, as I say, time’s up. Let me finish therefore, by saying a big thank you for sharing your expertise with us, first to John Healy of Intel. John, thank you.
John Healy 31:19
Thanks, Jeremy. It’s a pleasure.
Jeremy Cowan 31:20
And, John, how can people contact you if they want to know more of what you’re up to?
John Healy 31:26
Best is www.Intel.com/IoT .
Jeremy Cowan 31:29
Brilliant. And thank you to George Malim.
George Malim 31:32
A pleasure to do this. And I really enjoyed John’s comments, particularly the use cases. I thought those were really interesting and demonstrate the reality of all the technologies we’ve been talking about.
Jeremy Cowan 31:42
They really do. How can people reach you, George?
George Malim 31:45
I’m reachable on Twitter @GeorgeMalim, all one word.
Jeremy Cowan 31:51
Brilliant. I’m sure people will be in touch. Ladies and gentlemen, I can’t let you go without thanking you too, for joining us around the world. Don’t forget to subscribe to the pod wherever you found us. And if you can spare a moment, please give us a 5-star rating and say something nice about the pod. It makes a huge difference to our ranking.
So, keep safe and keep checking TheEE.ai, IoT-Now.com andVanillaPlus.com for tech news and features. And join us again soon for another Trending Tech Podcast looking at enterprise digital transformation. Bye for now!