In this episode of the Futurum Tech Podcast, host Shelly Kramer welcomes fellow analysts Olivier Blanchard and James Kobielus. We start off the show talking about CES and the focus this year on all things artificial intelligence, and then move on to a review of the fitness tracker market, and Intel’s CPU-based edge AI. We also check out Deeplite, a Seattle-based startup that’s doing cool things with neural architecture, important in the AutoML space, along with the exciting things to expect from Android in 2020.
Our Main Dive
As James has recently returned from an immersion in all things CES, we let him take the lead in this show on discussing the Consumer Electronics Show and his observations. His main observation this year — CES is becoming less of a gadget show and more of a connected ecosystem showcase. The AI emphasis exhibited at this year’s show is bringing more of a developer focus to CES vendor discussions, which is interesting. Also of interest is our observation that the CES AI hype, while exciting to some, was a little too rose-colored, and failed to address key risk/ethics concerns. We would like to see more meaningful discussion on the risks associated with AI, its impact on privacy (and encroachment on privacy) and the bias that is often inherently built into machine learning models. Will there be a “winter of backlash for AI?” Perhaps so. We also discussed other themes at the show and things we know we’ll see more of in the coming year, like 5G and Wifi 6, to name just a few.
Our Fast Five
We dig into this week’s interesting and noteworthy news:
- CPU-based edge AI. James covers Intel’s announcements at CES regarding the next generations of microprocessors focused on edge deployments for AI. Intel’s Xeon Scalable will be released by the middle of 2020 and will offer its third generation scalable, 60% greater performance on AI workloads, specifically the deep learning workloads like computer vision. He believes this will be a very important chip and will dominate CPU-based edge AI compared to the second generation of this product. It seems clear that Intel is gearing up for more robotics-type applications, which of course makes sense.
- State of the Fitness Tracker Market. Olivier shares stats from Pew on fitness trackers, market penetration, and demographic breakdown, especially relevant because they are the low-hanging fruit of wearables. An interesting data point: 26 percent of Hispanics use wearables, 23 percent of Black respondents, and 20 percent of whites. And 25 percent of those consumers are between 18 to 49 (fairly large range there) and 17 percent are over 50. There’s a lot of room to grow in this market, and Olivier predicts we might see 50 percent adoption of fitness wearables by 2023 or 2024. He also covered this in greater detail in this article: Global Wearable Sales Double YOY, Women Users Lead the Way.
- What Deeplite’s up to. When we turn to a conversation about Deeplite, a Montreal based startup, that is more of a software discussion about AI. As James explains, automation of the machine learning pipeline, or AI development and operations, is a big trend today. What Deeplite does, in the simplest of terms, is provide software to automate a particular function that data scientists do that has largely been difficult to automate, and that’s the automation of neural networks that are really the heart of machine learning — which is the heart of AI.
- Deeplite has formed some interesting partnerships, including one with Asia-based Andes Technology, and is on the path to commercializing neural architecture search, to be able to take a machine learning model that has been built by any data scientist and compress it down to a smaller representation to run in a constrained edge device without losing its accuracy in terms of doing the inferencing for which it was built. For more on that, check out James’ recent in-depth article: Apple’s Acquisition of Xnor.ai Aims to Deliver TinyML to Edge Devices, which covers this topic in interesting detail, with additional insights on the role Deeplite might play in the future.
- The exciting things to expect from Android in 2020 (with Samsung leading the way). An unabashed Android fan, Olivier covers the Samsung S20 leak, and other things to expect from Android in 2020
For the Tech Bites portion of our show, we cover weirdness from Teen Vogue. Business insider reported that Teen Vogue posted a well-polished and positive piece on how Facebook is “securing” the 2020 elections. It looks like sponsored content. It smells like sponsored content. It is sponsored content. At least Teen Vogue says it was sponsored content by posting a note at the top of the article. Facebook then denies it, saying the article is not sponsored content. Then it changes course and says, oh wait, we guess it was sponsored content. And then, Teen Vogue deletes the entire story. Get it together, Teen Vogue. We expect more from you.
Crystal Ball: Future-um Predictions and Guesses
In the Crystal Ball section of the podcast, we circle back to CES and our predictions on what the main technology or technology category will be for the 2021 show. Without hesitation, all three of us have an answer … you’ll have to listen to the show to find out what it is! Transcript: Shelly Kramer: Welcome to this week’s episode of the Futurum Tech Podcast. This is your host, Shelly Kramer, and I’m joined today by my colleague Olivier Blanchard, fellow Analyst here at Futurum, and by the newest member of our analyst teams, James Kobielus. Welcome. James Kobielus: Hello Shelly and Olivier. Nice to be here. Nice to be on the team. Shelly Kramer: Olivier, you’re welcome too, but you’re kind of old school- Olivier Blanchard: I know, I’m old school. That’s right. Shelly Kramer: Old hat. Yeah, you’re always here. So we are going to dive right in today, because we have made a new year’s resolution to be timely in our podcasts. In this week’s episode of the podcast, we are going to talk about the Consumer Electronics Show. James was there this last week and he has some great observations that he’s going to share. I know Olivier, you have some insights as well, so we are going to cover that as our Deep Dive for the podcast today. Before we get started, I wanted to remind you that this podcast is meant for information and for entertainment purposes. We might mention publicly traded companies. We might share opinions and insights, well, we always share opinions, but that advice is not to be taken as investment advice or anything else. This is our disclaimer, so make sure you understand that. So here we go. Without further ado, let’s talk about CES. James, you were there. Tell us maybe what your top takeaways from the event were. James Kobielus: Thank you, Shelly. Yeah, actually this was my first CES, though I’ve been an analyst for years, so I’ve been covering enterprise topics, but I really felt that this year I needed to be at CES because the hottest topic in the consumer space now is AI inside of everything and that was the dominant theme at CES. And of course it’s a gadget show, so there’s AI inside of almost every gadget, and much of this played out in the mainstream media. But I see CES is becoming less of a gadget show in its actual orientation in terms of what the announcements that are made, and more of a show focused on connected ecosystems for broad use cases like smart homes, smart cities, autonomous vehicles, streaming media and so forth. And that hunch was born out by my experience at the event. That’s number one, it’s a connected ecosystem showcase. CES could stand for that. Number two, and as I said, AI is the dominant theme, or was the dominant theme, terms of the secret sauce inside of so many gadgets and ecosystems. And then, that’s just AI, so much as I saw an increasing focus this year, having been at CES, on the developer. The developer of AI came into many product announcements, tools, geared towards automating more of the development and the training and deployment and management of the AI inside of diverse edge devices, mobile, and so forth. And thirdly, related to that was, the hype was ubiquitous around AI at the show. It was a very rose-colored hype, it was all about the promise of AI and the miracle of AI. But there was very little meaningful discussion at the event, that I saw, were on the risks from AI, and the risks loom large in the popular mind now, in terms of the privacy encroachment, in terms of bias built into machine learning models and so forth. So I felt that the hype itself was, this year, is at the point where it’s becoming unsustainable, that there might be a backlash. And I’ve heard some say that there might be another so-called AI winter of backlash. I don’t think that’s going to happen, but I do think that CES is in for sort of the harsh reality, that next year, I think there’ll be a lot more, sort of push back at the sort of the hype level around this tech. Shelly Kramer: Yeah. I know, I think that’s very interesting. And you actually wrote a blog post about that that published late last week and I thought you were spot on in terms of the over saturation of the AI buzz factor, and that that’s a very real concern, should be a very real concern of companies who are hanging their hats on everything AI. James Kobielus: Yeah, for sure. AI is both a winner in terms of the dominant theme, but it’s also a potential loser in meaning, it carries the seeds, at least the hype of its own destruction as it were, if the industry isn’t careful to moderate their claims for the so-called miracle of AI. AI is a lot of moving parts that quite often need expert human beings to make it all work smoothly. And also to build in safeguards to protect all of us from the downsides of AI, which are very real. Shelly Kramer: So Olivier, I know that you weren’t at the event, but I think that you were, well, I know that you were paying attention to it. Olivier Blanchard: Right. Yeah. CES is something I don’t need to do every year, just because the kind of technology trends and development are a little bit longer than a 12 month cycle. So I like to skip CES and just go every other year and that way I feel like I’m not getting too much repetition. But I’m actually really fascinated by the whole discussion about AI, because AI is such a catchall term. It’s a huge topic. And when people think about AI, they might think of IBM Watson, for instance. Which is technically not really AI. If you talk to people who are in the AI space, right? They’ll cringe a little bit, right? But that’s a different discussion. James Kobielus: We’ll push back later on that. Olivier Blanchard: Okay. All right. But so there’s this kind of this sense of the broader AI with how is AI going to kind of take things away from us and just kind of eventually take us over and control us. Right? It’s the Skynet AI fear. But then at the same time you have these kind of tiny, little AI components or features that are found in phones for instance. So there’s the AI controls the performance of your phone camera. There’s AI just in everything. AI allows your phones to wake up just from a passive mode, passive listening. And there’s a lot in between, as well. And so, I understand that it’s really popular for companies to preach AI and say, “Yeah, we’re an AI company now.” And they kind of are, because AI is built into everything and the cloud is moving to the edge so everything’s moving on device. Everything is moving on chips, so a term that some of our listeners might be hearing a lot from us in the next year is SoC or system on chip, because a lot is moving there. I think CES, and by the way, I love the connected ecosystem show rebrand. That’s actually clever. But I think that CES is always the kind of show that’s just so big, that tries to be everything to everybody from connected toilets to autonomous vehicles and XR and phones, that it’s easy to kind of get overwhelmed by just the sheer size and lack focus of the show. But at the same time I think it’s kind of great because you get this sea of possibilities and applications and use cases. But then, out of that you can kind of distill the two or three winners in the next two or three years that are going to lead technology in a different direction and branch off a little bit from where we thought we were two years ago. And definitely AI is, AI five, while ubiquitous connectivity, it’s not necessarily 5G because there’s also WIFI 6 and other things. But those are two of the big winners, or potential winners, I saw from a distance about this year’s CES. Shelly Kramer: Well, yeah. It’s always interesting. That’s one of the shows that I like to take a regular break from, as well. It’s just too much hype and not enough that keeps it always interested. Now we’re going to jump into the section of our show that we call the Fast Five. And Jim, I’m going to toss it to you, and talk about what is Intel’s doing? James Kobielus: Yeah. Intel made several announcements last week at CES regarding the next generations of fair microprocessors focused on edge deployments for artificial intelligence. Really there, one of their core announcements was the so-called Xeon, well, it’s not so-called, it’s the product brand name. It’s the Intel Xeon scalable, which will be released by the middle of this year. Which will offer, the third generation of the scalable, will offer up to 60% greater performance on AI workload, specifically deep learning workloads like computer vision. 60% greater performance than generation two, for both training workloads and inferencing workloads for AI at the edge on a CPU. And the reason why I’m calling this out, why this is important, because, first of all, most of the inferencing that happens at the edge is on CPUs, already much of it is Intel. Number two, the dominant workloads are training and inferencing everywhere in the AI universe. And usually, only inferencing happens at the edge, on the edge devices, but training on the edge is becoming more important for next generation apps like robotics and robotics things like drones and autonomous vehicles. It looks like Intel with scalable third generation, by boosting the training performance at the edge seems to be gearing it up for more robotics type applications, which I think is the future. Really, the future of CES, in terms of the entire gadget ecosystem, will depend on the heap and help of robotics. So I think the scalable generation three will be a very important chip and will be widely adopted when it’s in market this year. Shelly Kramer: Olivier, you’re going to share some interesting stats from Pew about fitness trackers. Olivier Blanchard: I am. Since our theme is CES, I thought I would stick to the gadgets a little bit or as much as possible this week. And by the way, inference is something that we’re going to be talking a lot about this year, I think. So mine is about a bit of research from Pew that focused on wearables. And specifically by the numbers, I can give you what I found, which I thought was interesting, because wearables are, they’re a relatively young category of technology, but at the same time I feel like it’s maturing a little bit. But according to Pew, 21% of Americans currently report using a smartwatch or a fitness tracker of some kind already. So that’s not bad, one in five. I thought it would be a little bit higher, so there’s a lot of room to grow. That breaks down into 25% of women, but only 18% of men. So I was surprised to see that women actually use them more than men. The break down is 26% of Hispanic respondents use them, to 23% of black respondents, versus 20% of whites. So as it turns out, the white demographic is less into the wearables and the fitness trackers than their neighbors. So that was interesting as well. 25% of those are 18 to 49. 17% of those are over 50, so you should definitely buy a fitness tracker or smartwatch for your parents if you haven’t already. That will make the tech companies happy. And 31% of those who make $75,000 per year or more, use fitness trackers, versus 20% from people who make $30,000 to $75,000. There’s a lot of room to grow as well. The difference between income levels is not particularly significant, but definitely it’s still considered a luxury item, generally, for at least a third of users. And Apple, which isn’t always my favorite company to talk about because I’m still a little bit disappointed in their performance in the last few years, still does win in this particular category. And the wearable market, by the way, has doubled since last year. So those numbers are actually growing very quickly. We could see 50% adoption by 2023, 2024, which I thought was also very interesting. Shelly Kramer: Yeah, I agree. I think it’s very interesting and I’m surprised that some of those adoption numbers are as low as they are, but in my world you don’t go anywhere without people checking their … I was at a volleyball tournament all day yesterday, and one of my fellow parents was saying that her Apple Watch was tracking her as working out because her heartbeat was so elevated because of the tournament. So I mean, I do think that they’re everywhere. Continuing to move along, Jim, you’re going to talk a little bit about DeepLight and what they’re doing. James Kobielus: So my first comment at Intel was semiconductor, it’s down the hardware level. The DeepLight is more of a software level discussion AI. Automation of the machine learning pipeline, or AI development and operations, is a huge trend in the business world. And there’s been lots of startup activity providing these data science workbenches that enable that degree of automation. What DeepLight does in the broader AutoML universe is provide software to automate a particular function that data scientists do that heretofore has been hard to automate, which is the optimization of the performance of the neural networks, the automated neural networks that are at the heart of machine learning, which is the heart of AI. This approach called neural architecture search is not only DeepLight, there’s been a lot of R and D going on in, throughout the world for several years in this regard. But DeepLight has taken a leading position in terms of commercializing neural architecture search, to be able to take a machine learning model that’s been built by any data scientist really, and be able to compress it down to a smaller representation to run in a constrained edge device without losing its accuracy in terms of doing the inferencing for which it was built. Whether it be for recognizing visual objects or whatnot. And so, DeepLight, I’m very impressed with what they’ve got in terms of their approach. They have partnerships, and they published recently a press release about a partnership in Taiwan to deploy the technology into more, develop once, run anywhere, machine learning deployments out at the edge. So I think DeepLight was one of the, well, really in many ways, was my takeaway, from what they’re doing is actually more mature than I expected in terms of ready for broad commercialization. And I think there’ll be others jumping in … When I say others, other vendors of data science and AI workbenches will be jumping into this space with neural architecture search capabilities to automate modeling this year. Which some people will say, “Will that put data scientists out of a job?” I say no, there’s plenty of things that data scientists need to do to in the context of AutoML environments. But this will be hugely important in being able to basically to fit or to retrofit almost any machine learning model to almost any edge device going forward. And that’s absolutely essential in this world. Shelly Kramer: Yeah, I know. That’s really interesting. Thank you for sharing that. So we are going to, again, we are limited by time today, so we’re going to be happy with four Fast Fives today. And we’re going to jump real quickly to our Tech Bites section. And I’m smiling because our colleague and fellow analyst, Fred McClimans, actually shared this remotely during our conversation. But I think it’s super interesting. Business Insider reported that Teen Vogue posted a well-polished and positive piece on how Facebook is securing the 2020 elections. It looks like sponsored content. It is sponsored content. At least, Teen Vogue says it was sponsored content by posting a note on the top of the article. Facebook denies it, says it’s not sponsored content, and then it says it was sponsored content. And then, Teen Vogue deletes the entire story. And I have to say I’m a huge Teen Vogue fan. I mean, I’m a grown ass woman and I love Teen Vogue’s content. I love it as it relates to content that appeals to my 14 year old teenagers and young women in general, and they are, their coverage has typically been pretty amazing. It actually surprises me to no end that they would publish a piece of sponsored content like this, from Facebook of all places. But anyway, that was sort of an interesting … I think you shared something on that as well on Facebook, Olivier, not necessarily about the Teen Vogue piece. Olivier Blanchard: Right. Well, yeah. The background on this is that Facebook has decided to be fine with misleading political advertising, and so it’s kind of created this backlash. And obviously Facebook is trying to write articles and have people write articles and steer the conversation towards, no, this is actually a good thing and it’s freedom of speech. And there’s an argument to be made for that. But the problem there, obviously, is that Teen Vogue’s editors or whoever published the content should have disclosed that it was sponsored content, and failed to. And there was a little bit of a snafu and miscommunication somewhere, and it just turned into this thing. It was just really poorly handled. And I don’t think it was malicious. I don’t think Teen Vogue or Facebook were trying to do anything unethical here. But the disclosure checklist was probably not- Shelly Kramer: Followed. Olivier Blanchard: … followed to the letter, yeah. James Kobielus: You know, the thing is about freedom of the press and speech, that’s important, of course, but in any ecosystem of free dissemination of news and information, there needs to be some degree of trust in the providence of the information. It can be traced back to its source. That there’s some authoritative record or ability to roll back a record of who’s responsible for the content that originated, and then from the various edits and so forth along the way. Clearly, here you had no one organization that was able to provide a thorough data lineage or providence to the content that was put out. That’s really disturbing when you have these major organizations, publishing groups and social media, they don’t know what they’re doing. They’re confused themselves. That’s like, that’s disturbing. Olivier Blanchard: Yeah. Train your interns better at Teen Vogue. Shelly Kramer: Well, yes. And I think equally is important, the words “Facebook” and “trust” for most of us do not go together. Will not go together. Will not ever go together. I mean, it’s just kind of, yeah, well, not for me. Olivier Blanchard: Maybe someday, yeah. Not anytime soon. Shelly Kramer: Yeah. So this leads us to the wrap up of our show. We’re going to touch on a crystal ball question. Olivier, you have any thoughts on what would, what you want to ask? Olivier Blanchard: Oh, wow. Crystal Ball. Well, let me consult my crystal ball. Shelly Kramer: And we usually tie it back to our main dive, which is the consumer, the connected electronics, the connected ecosystem. Right? Show … Olivier Blanchard: Let’s go with this. Let’s go with this. What will we be talking about a year from now, going in to CES? What technology or technology category will be the leading topic at CES next year? James Kobielus: It’ll be 5G. Shelly Kramer: Agree. I mean, that’s, yeah. Well, that’s kind of … That was an easy crystal ball [crosstalk] question. Olivier Blanchard: Initially, what do you know … ? I mean, it’s, yeah. 5G is part of it, but I still think it’s going to be Distributed AI next year. James Kobielus: Yeah. Well, AI is easier to distribute in great … it is more scalable over, will be over 5G- Olivier Blanchard: Absolutely. James Kobielus: … because I did a post about a month ago or so, ago, I think it was on information week, laying out the ways in which the AI revolution will be accelerated through 5G. They’re really symbiotic in many ways. Olivier Blanchard: Yeah. It’s kind of like the mobile world’s congress. It’s kind of the 5G show, right? And then you have the gadgets at CES that use 5G, just kind of sit on top of it as a layer and then give us these features. I think we’re all right. So, that’s a good way to close the show. Shelly Kramer: That’s a great way to close the show. All right. With that, we will thank you for listening, and thank you, James, for joining us for your initial run on the Futurum Tech Podcast. And Olivier, always a pleasure, and we’ll talk to you all next week. There will be plenty of more tech topics and tech conversations right here on the Futurum Tech Podcast, FTP. Hit that subscribe button. Join us, become part of our community. We would love to hear from you. Check us out. futurumresearch.com or Futurum Tech Podcast. Disclaimer: The Futurum Tech Podcast is for information and entertainment purposes only. Over the course of this podcast, we may talk about companies that are publicly traded and we may even reference that fact and their equity share price, but please do not take anything that we say as a recommendation about what you should do with your investment dollars. We are not investment advisors and we do not ask that you treat us as such.