On today’s show our guest is Bob O’Donnell. Bob is President and Founder and Chief Analyst at TECHnalysis Research. He’s widely regarded as an expert in the technology market research field and his original research and advice is used by executives and large technology firms all over the world.
Cyber Security Inside Podcast
#5 – AI’s Role in Cyber Security
Share on social:
I’d like to introduce my co-host Camille Morhardt. So hi, Camille, how are you doing today?
Camille Morhardt: Hey Tom. I’m doing great.
Tom Garrison: So what’s on your mind today?
Camille Morhardt: Well, I know this sounds like a big topic. I was going to say artificial intelligence and compute.
Tom Garrison: Wow.
Camille Morhardt: But I wanted to start with something a little bit smaller: end devices. So when I think about the evolution of AI, the smartphone, particularly, with its built-in camera kind of gave deep learning such a boost. And then when I think about when I know what I’m going to do, and I need to sit down and get something done, I still go to my PC.
So, what I’m wondering is when I think of the development of AI, it’s kind of through the smartphone as this end device. And then of course, servers on the backend for centralized learning models. And then when I think about the future, I tend to think IOT, preventive, maintenance and exciting things like that. But what about this basic workhorse that is the PC? What’s happening with respect to artificial intelligence and the PC
Tom Garrison: Yeah, there’s a lot to unpack there. In general, there’s some pretty cool things about AI. Some of them sound boring, but they’re, they’re actually pretty game-changing and one of them sort of boring sounding ones is using AI to basically guess what you’re about to do on the PC.
So if, for example, you’re working away in Microsoft Word, and you’ve been going at it for a while typing, and then you pause for a moment and you start to move the mouse up, chances are you’re, you’re either going to be clicking on Save, or you’re going to be clicking on Print or something like that. And using AI based on your, the things that you do–without even thinking about it–you can use AI to guess what you’re about to do, and then make those actions something that’s basically one click away or something that’s just right there on the screen.
And it’s kind of invisible to the user, but it gets us out of having to remember “which dropdown box do I have to click on this and then that.” And you know, like if in Excel, I don’t know how many times in Excel I have tried to find the dang “wrap text,” little check box. Those are all things where AI can watch your behavior over time and learn your behavior and then sort of present the things that you’re likely to do in a very easy to find mechanism.
Camille Morhardt: Okay. So you’re talking about, you know, basic sort of workload, help my life get better kind of a thing. So what about on the security front? Is there anything that we’re seeing there?
Tom Garrison: Yeah. So the first one was just one sort of simple example. And then on the AI side for security what’s being looked at now is around using AI to see is the machine operating in a way that it doesn’t normally operate. So knowing enough about the way you use your device, to be able to say, “huh, now I, the PC, and operating in a way I don’t normally operate” and flagging that. Doing that in a way that doesn’t induce a lot of false positives (or obviously false negatives too) but false positives are a real problem for security, because if you are sort of Chicken Little, and you’re always raising your hand saying, “Oh, there’s a problem! Oh, there’s a problem!,” then pretty soon people start ignoring you.
And so the, the promise of AI is to be able to do that and see these anomalous behaviors that you should flag.
Camille Morhardt: I would like to learn a little bit more about that and find out what other people in the industry are seeing.
Tom Garrison: Yeah. I think that’s probably a great podcast right there. What, what do you say we narrow in on that topic?
Camille Morhardt: Yeah, I like it.
Tom Garrison: All right, let’s go for it.
Okay on today’s show our guest is Bob O’Donnell. Bob is President and Founder and Chief Analyst at TECHnalysis Research. He’s widely regarded as an expert in the technology market research field and his original research and advice is used by executives and large technology firms all over the world.
So Bob, you are the perfect guest for us today. So thank you and welcome to the show.
Bob O’Donnell: Thanks for having me.
Tom Garrison: So our topic today is around Artificial Intelligence. And I wonder if you could just spend a moment and talk a little bit about your background and this topic around AI.
Bob O’Donnell: Sure. So I have been a tech industry analyst for a little over 20 years. And prior to that, I was in the music technology business–so writing and reading and playing with musical equipment (because I’m a musician for fun, as well). But so I’ve been following tech industry trends for a long time. And as we’ve seen the evolution of computing, we’ve seen the development of more sophisticated software tools along with more sophisticated silicon and those worlds kind of really coming together in a very interesting way with Artificial Intelligence. The idea being that you could start to see the ability to do things above and beyond what basic software would allow and enable, and then unique means of solving problems and then silicon being designed to accelerate that, cause it turns out not all, everything would just be accelerated by a CPU.
But long story short is as I’ve tracked these trends in devices and core technologies and software in the cloud, AI has all of a sudden become this huge issue. And I’ve done some independent research studies on it, I did a survey of AI use in the enterprise. I’ve done research on AI and consumer applications gaming and so it’s just an area that I’ve looked at quite closely, because there’s so much interest in fascination with it.
Of course there’s so many different variations on it between machine learning and the different flavors of AI. And it gets very confusing very quickly, certainly, but at the end of the day, it’s about being able to extend some of the core, basic types of software tools that we’ve created in ways that we may not have thought of before. And it’s also a way, frankly, from another perspective, it’s a way to make sense of data in a manner with which we haven’t thought about it before.
So it’s a combination of how do you create these algorithms? how do you interpret this data? and how do you put that all together into something that goes above and beyond what we’ve traditionally done? And it’s a fascinating field, obviously, that has lots of implications all over the place.
Tom Garrison: Yeah, no, this is, this is a great, and, and I wonder through the research that you’ve done–and I understand you’ve got a white paper coming out as well–for the listeners here, what are some of the key sort of “ahas” or takeaways from your research?
Bob O’Donnell: Initially all the excitement, frankly, and all the action in AI was happening on smartphones, right? It was all about smartphones. A lot of it was we heard about computational photography, the ability to enhance image quality and do, uh, very clever processing in ways above and beyond what you could do with the traditional Photoshop filter types of things. And then we saw audio processing, as well, as some other things. But the PC was a little late to the game.
And now what we’re starting to see–and what my research is on–is about AI usage on PCs. We’re starting to see PCs be part of the equation. We’re seeing a lot of adoption of AI in various PC applications. 90% of PC developers that we surveyed are working on some sort of AI machine learning or deep learning type of effort–either by integrating into a function within their application or building entirely new applications based on that. So that’s huge, right? That’s a huge amount of focus being placed there.
And at the same time, we’ve also seen of course, a lot of effort around both companies like Intel, as well as NVIDIA and others to build algorithms and software development kits that can leverage that and to build acceleration into some of the chips that they’re creating. So, I mean, everybody is really focused on trying to bring some of that magic that we saw with smartphones a couple of years back to the PC, because there’s a lot of interesting applications, especially nowadays when we’re all using our PCs a heck of a lot more.
Camille Morhardt: Hey Bob, what is kind of one of the major use cases that people are actually doing with AI on a PC?
Bob O’Donnell: There’s a number of things. So we are seeing some of the same kinds of things we saw in smartphones. We’re seeing some of the filters, you know, for image filtering and audio filtering, especially now with video conferencing, noise reduction in the background is a huge deal, right? Because we’ve all had dogs and kids and, you know, loud noises happening in the background.
The other thing we’ve seen, actually, is workflow automation, processes totally radically different kind of thing, but using tools to leverage how data workflows are happening or process workflows. All those kinds of tools that are run on PCs are also changing.Also a lot on security and threat protection. We’re seeing more and more automated tools to look for security threats.
You know, a lot of what AI does at a simplistic level is it Looks for patterns, right? You teach it a bunch of patterns–a lot of these AI algorithms–and then from that, it can determine other patterns. That’s a classic, deep learning application. It was initially, you know, it was show 50 pictures of, uh, of, of dogs and then show some more pictures that they haven’t been trained on and decide if it’s a dog or not. Well, take that a million times further, here’s a signature or here’s an application that’s functioning in an unusual way on a PC,
could that potentially be a security threat? And so you’ll see a lot of AI based tools around security and threat protection also being used.
Camille Morhardt: Who’s owning those models, then? If we’re doing AI on the PC and looking for threat protection, in particular, I guess maybe, you know, is that the IT department who’s owning the takeaways from that? or are there managed service providers that are collecting that?
Bob O’Donnell: I think we’re seeing all of the above. Obviously in a lot of corporate environments, and even in our extended corporate view of the world with a work from home, IT shops will install, obviously, a number of security tools–there’s the traditional MacAfee, Symantec types of things. There’s obviously what Microsoft has done with Defender. But there’s more advanced other technologies we’ve seen from Cylance and some of these other companies–some of whom have been purchased by some of the big PC vendors.
But there’s a number of tools being deployed, sometimes by corporate IT, sometimes by individuals because, you know, the boundaries between personal and work of course have completely been obliterated during the pandemic. And so you have people working on personal PCs and they’re installing those kinds of tools there. But you also, in fact, have service providers, uh, who are involved with this at a corporate level. You’ve got people who provide a managed security type services that are watching what goes in and out, past the firewall. Again, things are very different now because whereas everything used to be behind the firewall, now, literally everything is outside the firewall and that’s changed the dynamic of what the things you have to look for, the types of threats.
So there’s all kinds of services being offered from a variety of vendors. You’re seeing it as well in network equipment, from the large networking companies. So folks who are in charge of the network at many organizations as a part of IT they might be monitoring. Um, so it’s being approached and attacked on many different levels with AI being applied to almost all of these different security applications.
Tom Garrison: So do you see Bob then that the AI is basically just being integrated into many of the sort of existing products that are out there? And it just makes their products better?
Bob O’Donnell: It is. It’s a good question, Tom. And yeah, I mean the bottom line is a lot of what’s happening is not necessarily that the entire– I mentioned that some people are trying to do entirely new apps with AI. But the vast majority of what’s happening is they’re taking a function or two, and they’re integrating AI into that. Or they’re building a couple of special new features and capabilities leveraging AI models or deep learning or what have you. So that’s typically the way that we’re seeing, developers on the PC, as well as other platforms do that, right? We saw the same thing on smartphones. There were always photo apps and camera apps on smartphones, but they just got a little bit smarter through the integration of some of these technologies.
And frankly, in the case of smartphones, Qualcomm had a bunch of software development kits and APIs and things like that, along with Android and the two worked together to create a suite of tools that developers could use. Now, we’re seeing the same thing with Intel doing that with OpenVINO on the PC side, as well as Microsoft. So there’s a lot of efforts. And then of course there’s, you know, and then special instructions being integrated into the latest generation of CPU’s again from Intel as well as from AMD. So lots of different parties working together to bring AI more to the mainstream.
Tom Garrison: We’re certainly doing a lot of work in the hardware side, making sure that our platforms are, uh, highly performant doing AI type workloads. I wonder, from your perspective, is there anything that really has caught your imagination? Cause I’m envisioning now our listeners are listening to this podcast saying how is AI gonna impact my business?
Bob O’Donnell: Well, I think it’s going to happen across a number of areas. Sort of a big picture one is around analytics. You know, we’ve talked about analytics and big data in the corporate world for, I dunno, 10, 15 years. It seems like forever. And the reality is that a lot of the initial analytics efforts, frankly, were not very successful. They were trying to dive into big chunks of data and try and discover patterns and, and they really weren’t particularly successful in doing so.
The beauty of AI is you’re unleashing algorithms onto these huge datasets and they are finding more success. So I think anything that involves traditional analytics types of applications, where you’re searching for patterns in data–and that can happen across any industry and we’re seeing that all kinds of places. We’re also happening, see it happening in IOT tape type applications. If it’s in manufacturing, you know, predictive analytics where you can not only be, you know, searching for data, but you can see patterns start to emerge of sensor data that might make you say, ”Oh, I think that piece of equipment is going to fail. We’ve got to deal with that.”
We’re starting to see that as well on PCs, right? I mean, it was back from the old days of smart hard drives, right, where you have these sort of basic tools built into the hard drive, they would try and be able to warn you, “Hey, I think we’re in trouble here.” Now we’ve got the same kinds of things happening on other components, right?–whether it be memory or other elements of a PC. So we’re seeing those, that predictive analytics happening.
The other big area, frankly, than I think most people are starting to see is in basic office productivity. So now, for example, if you use your, either Office 365 or G Suite, or is now Google calls it Google Workspaces, you’ve got these tools, the editing applications that give you content recommendations, right? They’ll say, “Hey, not only is it a spell checker, it’s a grammar checker. Now it’s even a content type of checker. Here’s some suggested content for you.” One of the things I love in PowerPoint is a feature called Designer and Designer is an AI powered function that will create layouts for you. If you don’t have your own in-house art department who designs all your slides, you’ve got to create your own. And even if you have a preset template that a lot of companies have, you still want to jazz it up and create some varieties and do some cool things with images. And the beauty of Designer is it can take some images and come up with some suggested layouts that look awesome and require very little effort on your part.
We’re seeing things like the ability in video conferencing applications to track someone if someone’s walking around, uh, or they’re swaying, the camera can track the person and keep them centered in the frame. Uh, so all kinds of subtle– and that we’ve also seen things like, you know, A little creepy, but you know, they raise your eyes up so it makes you look like you’re actually looking at the person instead of looking down. Cause you know, a lot of times your camera’s above your screen, so you really looking up, but sometimes you’re looking down at the people you’re talking to. And so it’s a little weird. So it literally just tweaks the position of where your eyeballs are looking to make it feel like someone’s actually looking at you as they’re talking to them in a Zoom call.
So like I said, all kinds of different real world applications that I think pretty much everyone has started to see and there’s creeping their way into the mainstream.
Camille Morhardt: Okay. So you’ve used the word “creepy” and “creeping” a couple of times. So I’m going to run with that just a little bit. What are we worried at all about privacy when we’ve got all of this kinds of tracking and voice, and now content suggestions? I won’t even go there?
Bob O’Donnell: Yes. Look, people are a little worried about it, right? Analytics, one of the, one of the analytics that people are doing is personal analytics, as in it’s tracking everything I do and then making suggestions on what I want, right? We’ve seen this with advertising. We see this with all kinds of things and so yes, there is obviously some concern with that.
The beauty of what’s happening is we are now getting the intelligence and the compute power to do what’s called Inferencing, locally. So, you know, the idea you’ve got training and inference when it comes to AI training is when you take a whole bunch of data and you create these algorithms by essentially training it what to look for, what to think of that’s classic machine and deep learning types of algorithms. Then you apply those and you do inference by taking input and comparing it essentially to the algorithm and figuring it out.
Now in the past, you used to have to do that inference in the cloud, meaning everything you did had to be sent to the cloud, to someone else’s data center and the data was processed there. By doing it locally–even though that sounds like sort of an arbitrary distinction–it’s huge because it means all of a sudden, all of that inference work looking at my own data or your own data who’s ever owned data happens on the local device. So all of a sudden that means my data isn’t necessarily being shared out to the entire world and that makes a big difference to people, as well. They want the benefits of smart suggestions and content suggestions, all this kind of stuff. But, you know, they don’t necessarily want their entire life out there, for the world to analyze. That’s what I’m referring to there. But it’s an excellent question and something that we do have to be aware of whenever it comes to AI.
Camille Morhardt: So just to clarify, you’re saying, for example, if we’re going to work on removing background noise in my audio on a video call, you can make a suggested edit to the algorithm and then send that back to the model, as opposed to sending, say, my raw audio file, which would include the specifics of my conversation?
Bob O’Donnell: That’s exactly right. And so, first of all, they can do the analysis of that audio file, locally. But what they can also do is they can maybe come across a variation that occurred in your particular situation or someone else’s particular situation, upload that data, in turn, refine the algorithm, and then that algorithm in turn gets re-downloaded onto your system. So it’s a constantly iterating type of process. That’s the ideal. We’re not always, we’re not quite there yet in all cases, but that’s the concept is that you can get the benefits of AI, you can even get the benefits of an upgraded algorithm, without having to share too much of your own personal data.
Tom Garrison: We’re starting a new segment. So you’re the very first one of a brand new segment that we’re doing in our podcast now. And it’s basically what have you learned lately that you want to share with the podcast? Something cool, interesting. Could be something related to technology or it could be something in entertainment or something else you found intriguing and, I think, maybe our listeners might learn something from it as a result.
Bob O’Donnell: Well, I have two things and they’re radically different, but I’m going to throw them out there anyway. So recently one of my personal musical heroes passed away and that was Eddie Van Halen. I discovered Van Halen–I’m showing my age here–but at a young age and he has always been an amazing rockstar and just such an icon to me. An interesting factoid that came out after his death that I never knew is that he was part Indonesian. He was actually part Asian. And he actually suffered a great deal of bigotry for being Asian. I never ever knew that. So that was an interesting little factoid, about Eddie van Halen,
The other thing, and it’s again, totally unrelated, one of the things I’ve been doing with a little extra time during the pandemic is I– I’m a car guy and I have a few car Lego sets and I’ve discovered that there are lighting sets. You can put lights into your Legos. And so you can turn on the lights on your legos. It’s super cool. It’s a totally nerdy geeky thing that not everybody’s going to appreciate, but if you’re into stuff like that, there are lighting sets.
Tom Garrison: I, you know, I, I didn’t know either of those two, but, uh, the Lego one that is a, that is intriguing. (laughs) Camille, any, uh, items you want to add?
Camille Morhardt: Okay, well, what I learned this last week, probably anybody who spends time by the ocean already knows, but, uh, I learned that the best time to boogie board is not exactly at low tide, which I had previously thought, but it’s right after low tide when all the water is pushing you on shore, as opposed to dragging you out with that rip.
Bob O’Donnell: That would be an important thing to learn! (laughs)
Camille Morhardt: (laughs) Trial and error.
Bob O’Donnell: What about you Tom?
Tom Garrison: I am going to go into the world of entertainment. I’m always a big fan of these shows that I can just binge watch. And my son turned me on to a new show called “The Boys.” And let me just first tell everybody out there, do not watch this show with kids around. It is completely, completely inappropriate for kids. But it’s a world where there are superheroes, but they’re self-interested superheroes. They’re not like the Superman or Batman that we grew up with that are all about the public good. These people are in it for themselves. And, anyway, it’s, uh, it’s a fascinating to me. It was a fascinating kind of re-think about the whole superhero genre thing.
I think it’s very well done. There’s two full seasons. Now you can get on it. But anyway, Bob, thank you again for taking the time stopping by, sharing what you know about AI. It was really interesting. And I appreciate your time.
Bob O’Donnell: Well, thanks, Tom. And thanks Camille, thank you so much for having me. I really enjoyed the conversation.
Tom Garrison: All right. And for all of our listeners, we look forward to sharing with you the next podcast, which will come out in two more weeks and we’ll see you then
Subscribe and stay tuned for the next episode of Cyber Security Inside. Follow at @tommgarrison on Twitter to continue the conversation. Thank you for listening. .