Techstrong Gang – October 24, 2024

Hey folks, I'm Mike Ard. Today we're gonna be talking about the rise of Qualcomm. Then we're gonna ship some gears and talk about, well, is AI and software engineering.

Everything is supposed to be cracked up to be. And then finally, we're gonna have a little chat about AI as Smokey the Bear. You're watching Textron Gang.

And we'll be back in a minute. Our All right folks, let me introduce our gang members for the day. We have, of course, um, sitting out in Colorado, as usual, Mitch Ashley's joining us in, in, in a, in a background that's become quite familiar to people who watch the show.

I'm gonna have to move the guitars around a little bit once in a while, just so you know. It's really my background. All right, Good to see everybody.

All right, then moving up from where Mitch is, we have Anne Aloha. Ward is back. Anne, how you doing?

Wonderful. Thanks for having me Back. Always a pleasure.

And then moving to the left, there is our own John Schwartz. Hello. I was apparently appointed czar of Silicon Valley at some point.

I forgot when I, I kind of missed that notice, but I guess it's, it's well known. Yeah, that's true. That's true.

Yes. Um, it is all true. Yes.

Guilty. And finally, guy Careers joining us from Visual Impact, the arm of the FU group where he follows all kinds of technologies. Hey, guy, how you doing?

I'm good. Good to be here and to help witness the rise of Qualcomm. All right.

Well, apparently this is contentious because in addition to them having a conference this week where they showed, not surprisingly, another processor and system on a chip advance, there's noise in the system. Apparently, uh, the arm folks are not happy with the Qualcomm folks and are threatening to pull some licenses. And I'm not sure what that's all about.

But John, you're covering the event. What's your impression of what's going on with Qualcomm? Are they on the cusp of becoming something bigger than they previously were?

'cause when I think about 'em, I still think about phones. Yeah. I mean, uh, as, as, uh, our, uh, futur group, uh, CEO Dan Newman said that he, he looks at them as kind of almost like a third option or a, a possibility of them in imposing or encroaching on the X 86 incumbents and that they're gonna become a, this third player.

Um, there's also, if you wanna throw something into the mix, there's also that continuous, ongoing rumor that Qualcomm is gonna buy part of, if not all of Intel. Something that Daniel puts it very low odds, but he, he thinks there might, they might buy some part of Intel or parts of Intel. Um, you know, it's interesting, we even go back last week when I was at the Lenovo conference, where Intel and a MD formed this X 86 ecosystem advisory group, which you might correctly noted as a kind of a defensive measure against Qualcomm in particular.

So now what Qualcomm is doing is they are bringing their powerful, the technology behind his chips, uh, for laptops to his chips for smartphones. In a sense. They are more aggressively pushing for Gen ai.

And, um, I think they're a legitimate threat. Um, they also didn't stop there. They also, late Tuesday, they sent me some news that, um, they are now also, um, looking into, uh, state-of-the-art platforms designed to transform the driving experience.

And they announced partnership with Mercedes and Lee Auto. So they're very d very busy. They're ratcheting up the pressure on Intel and a MD.

It's gonna be interesting to see how this all kind of falls out. Guy. What's your impression of what's going on with Qualcomm?

'cause I know you kind of follow the space a little more closely than I do. Uh, well, Qualcomm is an aggressive player and has been, that's something to watch out for. Uh, they are longstanding, they know how to design chips that are fabulous chip designer, but they know all about chip production and pipeline and supply chain and the whole thing.

So they execute very, very well and have historically, um, and they've accidentally or maybe on purpose set themself up, uh, for this kind of battle. Royal, it's not just arm, um, it's until it's X 86. Uh, it's worth noting, you know, John's point is a, a, a fair one about the X 86, uh, consortium, I guess it's called.

I don't remember. I was there for the announcement. Don't remember the name.

Um, being a way to combat, uh, Qualcomm, it's also a way to combat arm, uh, because arm, um, and ARM-based chips are becoming more important, uh, in data centers, particularly cloud data centers. But, uh, there are other players trying to move ARM as a, a sort of a cloud native viable alternative to X 86 in the data center. So what, what does all of this mean?

Um, it seems, uh, likely that Qualcomm's initial take on licensing r um, was, uh, in mobile devices and four mobile devices. And then they saw the opportunity, uh, on, I probably in the data center a few years ago. And they bought this company called Nuvia, which was founded by former app folks, um, that already had an arm license to develop arm chips for the data center.

Come AI come like a lot of, uh, uh, interest in act in activity around, uh, neuro processing units, which are normally not arm or not arm based. They have little arm components in them. Anyway, this all gets mixed up and jumbled up.

And Qualcomm says, Hey, wait a second. We have the IP now that we bought Lia to create an alternative desktop or laptop, uh, processor, um, using this technology. At which point ARM says, wait a second.

No, you can't do that. This was licensed for data center use. And so, so is born this sort of lawsuit and this battle royale.

Qualcomm now a very capable player to fight on multiple fronts. I see arms move as just a move in this ongoing lawsuit, which is supposed to be, when is it John? It's, uh, go to trial in December or something like that, just ratcheting up the pressure to settle, which I fully expect them to do most likely because there is the water's warm, there is lots of room for lots of growth all around.

And, uh, there's room for growth and development in X 86. There's room for growth and development in R Yeah, I read that story and I was kinda like, here's two companies that are essentially pointing guns at their toes and threatening to figure out which one of them wants to blow off their foot first. I was just like, it's crazy.

But Pretty much, yeah, Pretty much. Well, there is wor there is worth mentioning one of the thing though, which is that ARM, um, is also trying to move up its own stack. I mean, we tend to think of the stack as, as not including components, but CPUs and processors have their own little stack, and ARM was founded on risk-based, uh, instructional, uh, instructional architectures, ILAs.

Um, and uh, that's more or less how the core is designed and the specifications around the core and how you put those cores together into the processing unit. But more and more often now, there are chip lit technologies that will allow, um, uh, silicon manufacturers to make systems on chips, SOCs. And so ARM has created this, uh, architecture, this SOC architecture called CSS, that is also a competitor.

It's a competitor to existing arm based chips like Graviton and, and uh, uh, Snapdragon and so forth. Um, so, uh, you know, I I, I think it's worth wondering if ARM doesn't think it's shooting itself in the toe because it wants to start providing licensing and architectures higher up the chip stack. Alright.

And, um, guy made a, a small snide comment in there that I'd love you to kind of expand upon. He suggested whether or not these strategies from the chip processors were accidental around purpose. And I'm trying to figure out which is the chicken and which is the egg here.

Did the AI community see these processors and go, oh boy, we can build this. Or did the AI community go do this stuff? And then they started looking around for processors and found Nvidia and Qualcomm and all this other stuff.

But, um, you know, we give the chip manufacturers all this credit for their foresight. But I happen, I have to wonder how much of this is, you know, accidental. There's no accident here.

There's too much money at stake. There's too much, too much to gain. I, I, I think it's just very careful planning.

I, I don't this, there's no random chance here. All right. What is your take then, Ann, about how critical is AI gonna be on these mobile devices versus where we might see AI run somewhere else in the backend?

And how much of these models are really gonna be quite literally in our hands? I mean, Qualcomm chips specifically have already been in my hands for many years 'cause I'm an Android user. But, um, regarding ai, it's already been in our pockets.

I think the only difference this last year is that consumers are now aware that that's what they're holding. Um, because the capabilities in, in the splash that chat GBT made, um, the meteoric rise got people talking and thinking about it. The, those capabilities were already in, in our phones.

Uh, and there's really nothing new there. I mean, AI's been a part of search for 10 plus years. Uh, as an SEO it's been a part of my life.

I've been thinking about it, working around it, trying to understand it for a long time. It's just a reckoning of us saying, okay, this is reality. How do we use it?

We're involved in it too. I mean, that's really the only difference. There's just a lot more types of AI models that we seem to be using over the years than necessarily what we use just for SEO purposes.

Oh, absolutely. I mean, consumers are now in a, they're in on the joke, right? They're, they're, they're in on the play.

Um, you know, I I've been testing them for a long time. I, I like Claude, I like Gemini. I'm less into Chet GPT than others, but, uh, other people are.

But if you love it, you love it, and it's your whole world for image generation. I mean, people are making videos with it. Consumers are having fun with it.

Social media involves it. I mean, it's had, its, its mass moment, right? It's had, its, it's pierced the zeitgeist.

Um, and that happens, like for example, for augmented reality that happened when everyone was using Snapchat filters and didn't realize it. When you're using this technology and not realizing it, that's, that's when you've reached that moment. And I think people are doing it now with the knowledge that they, they're using it and loving it.

Um, I think a lot of people are enjoying it. I think we've kind of exited the novelty phase, right? Wow, that's interesting.

What can I do with this sort of playing with it, seeing, you know, okay, interesting. And then moving on, and to your point, is it now embedded? Is it part of things that we do, part of what we use?

And you could do that on the, you know, whether it's on mobile phone or consumer devices or workflows and, uh, DevOps systems and that, that's really where the, the, the metal meets the road, uh, for AI and, and the generative AI specifically. Exactly. That's where we're gonna see the real results.

Exactly. And, and Claude, you know, there was a big anthropic splash this week. I'm sure you saw John, uh, that they're doing all these partnerships with Canva and Asana, and it's like the integrations are just starting and then the, it's like, uh, mashups are, are gonna be interesting to see how they integrate.

Uh, because the early days of this, you know, early days being like earlier this year, early last year, I mean, it week felt a lot, it felt a lot to me like blockchain, uh, where it was a solution looking for a problem. I think it's interesting is there's a lot of announcements about agents, you know, ag agent and, you know, autonomous agents that are in products now. Microsoft announced it, Salesforce, Atlassian, et cetera.

And those aren't necessarily at the edge. And one of the reasons is, is access to all the data, but also these knowledge graphs that connects and mean provides context. So you can actually tell the agent to go do something and it knows what you're, has a way of finding out what you're talking about.

It'd be interesting to see if those kind of things make their way onto the edge devices too. Or is that gonna be what's in the cloud versus what's on the edge and Infras stream happening there. So I'm curious to see how that's gonna evolve.

This reminds me a little of, uh, of on what is still actually a mobile app app experience now, but was especially the case, uh, 10, 15 years ago, whatever, which was, it seemed like the developers, um, assumed always that you were connected. And so, you know, I lived in New York City, go into Subway and half of your apps wouldn't work for no real reason because it was a small amount of data that could have been stored locally or whatever it was. And you're just sitting here like with a blank screen.

And, uh, something that annoys me to this, to this day. And in the same, my, my contention always was, um, you wanna distribute the data somewhat so that a client can do appropriate functions wherever they are connected or not good or poor latency. Edge developers know this very, very well, right?

They do. But it seems that the mobile developers never quite caught up on it. And it's gonna be the same thing with ai.

I mean, I don't wanna be negative Nelly so early, but I guess I will. It's gonna be the same thing, which is there's gonna be this assumption that the backend AI service is gonna always be available and when it's not, that's what that local inferencing is so helpful For. You've actually started to Classify game, Hey, hey, wait, I have to defend New York City.

It's 2024. We got free wifi in the subway baby. In the subway stations.

In the subway stations, baby. Because when you go between stations, especially between Forest Hills and, uh, Queens Plaza, your favorite little stretch there, Mike. Um, it just goes dead.

Well, I got something that's, it's better than nothing. Well, I think you're gonna throw in is at least in the, in games for mobile games, if started to classify games that require a connection versus this is a Connectionless game, I can play it on my own in the subway on the, that'll help on the whatever. So people, you know, as a recognition that, you know, 5G may have, uh, may transform the world, but it ain't gonna happen where there's no connection.

I haven't, and I am not totally bought into the user experience here yet. And maybe I'm missing something. So I will ask you this, but seems like on my phone now, I like, if I go search something, I get a response from Gen AI giving me some sort of, um, you know, overview that is marginally useful, followed by six or seven ads that were probably generated by AI before I get to a bunch of links that might be useful.

But, you know, they're kind of organized in ways that are, don't make a whole lot of sense. So, um, can we get a better user experience going here or, or am I just like maybe not using this correctly? I mean, I don't know.

Whoever sold you on the idea that AI was gonna eliminate spam, uh, that one made it worse. I mean, I bought a fire TV and it's all, it's the innovation there from a fire stick is that it integrates ads in like 50% more places. I mean, I think Google has rolled out recently ads into the AI overviews, um, and there is a mad rush to optimize on the SEO side for these overviews.

It's not really that different, but there's a lot of chum, there's a lot of chum that's coming up with this because the game has changed. And anytime the game has changed, people figure it out and then they take advantage of it. But, um, the personalization level of ads is, is only gonna get weirder and creepier.

Uh, and I actually coined a term for this a couple years ago. Uh, digit, it's a port man between, uh, digital and pere, the inner sanctum of building. Uh, basically you search for something, you talk about something and it follows you.

That is only gonna get worse. Unfortunately. I'm sorry to be a doom and gloom person today, But No, that's a word that can get me in trouble.

I'm not gonna use that word. Yeah. Dig is, yeah.

Is is it, it, it, the, the digital bits will follow You. It sounds almost like an a stalking in a certain weird way. Right?

They, the more they know, the more they, it's, the more they know about you, the more they are going to intrude to your life. Right? There are certain words that You don't.

All I know is I need an AI bot to do a summary of the summaries I get from AI bots. That's what I need. There are just certain words in the world.

And even if you don't know what they mean, just by the way they sound, they're bad. Yeah, Exactly. That can't be good.

Yeah. I mean, I was researching a car, I think I talked about it last time I was on the show, and I will be getting ads till the end of time. Mm.

Good. Every car breaker I looked at is serving me new and more annoying ads in more intrusive places. So I'm, I'm sorry to say that quality is not seem to be on the docket yet for AI overview and things of that nature.

And, and am I the only one? But I feel like sometimes I'll be talking about something with my wife offline, and yet I will show up online and there is an a for the thing we were just talking about. So like, is it getting that Creepy?

Well, that's, that's all. That's Alexa or, or Google Nest. Uh, Google, sorry.

Listening to you. I I, I'm not sure Google does it to be honest, but I know Alexa does. All right.

I'm sure I gotta turn up every device before I have a conversation that's really private. You just to have a conversation. Well, There you go.

Ooh, There's a new industry home. Skiffs. Yeah.

Go kind of like that idea. Then I think about it. We're gonna have to have like a honey come on, we're going in the cone of silence room over here to talk about this thing.

There you go. Somebody's gone. I mean, not a bad idea or like, you know, university of Texas, the, the business school had a, a giant fountain in the middle of the, um, in the middle of the courtyard.

And that was because it would block any parabolic mics. 'cause they had it origins there All Streams Of what goes back to 2001. I think about Hal reading the lips of the astronauts are Dave and, and his colleague Frank.

You know, they think that they're, they're privately away. And even, even in that this era, we think about if, if the ability to read your lip, you know, like you see it in sporting events all the time where they have the placard in front of their face, the lanyard in front of their face. And I think it's, God, it's kind of reaching that stage.

All right, so, well, I don't know what you're talking about Paranoid here. I'm not familiar with that technique. But, you know, this Is, this is, this is like the manager coming out to the mound and they gotta cover their mouth so that they know table and see what they're talking about.

Exactly. Is, is this where we're rolling? All right.

Good lord. I guess we're gonna see a lot of that in the days ahead. You see that?

The world's crazy. Hey, I'm hopeful that maybe we'll be using, I'm hopeful that we may one day use AI for something more useful than what we just talked about. 'cause a lot of it's just kind of getting in the way.

But That's on the next segment, Hopefully. Yes, it is. On the next thing.

All right folks, we'll be back in a minute 'cause we're gonna talk about, well, AI and software engineering One more time. com is the leading resource for news analysis and education on challenges facing the cybersecurity industry. com covers all aspects of cybersecurity, including data security, DevSecOps, cloud security, application security, network security, security threats, and more.

com has the largest selection of security content featuring breaking news, blog posts, podcasts, and more. com to learn more. com.

Home of security bloggers network. com. It's one of our primary sites and it's a topic we love.

And one of the things we keep tracking is the impact AI is having on software engineering. And there's a new report out from the folks at Dora, which is an arm of Google. And for those of you who do not know what Dora is, is the DevOps readiness assessment metrics that were created over time.

The track things like how often are you releasing software and how quickly can you fix things? The latest report has come out and suggests that even with the rise of ai, we're actually slowing down somewhat in our ability to release software. There are certain things that we are doing better, including documentation and maybe testing.

But Mitch, what's your take on what's going on here? I mean, you know, first they promised supplying cars and then they said there'd be faster software development and doesn't look like I got either right now And we're still driving on the road, right? No.

Was it, you know, Dora is, is the well established as the, the metric report. It's, it's what most people use as a benchmark to assess their, their progress towards adopting ai. And there are also some trends in that.

You know, I, I said that I mentioned the novelty phase of, of AI getting past the, oh wow, it does that. Um, and now we're kind of in the gangly awkward teenager phase, right? We're sort of figure, we're trying to figure out what it's about and how to use AI in development.

And you know, we've, we've there, everybody's coming out with their version of how AI fits into their platform, their development tool, their repository. There's a lot of good things happening, a ton of innovation. Um, but there's also, you know, with innovation comes a lot of failures, a lot of good ideas that are bad ideas.

And they, they go by the wayside. So I think the quote unquote productivity hit is, is more of us experimenting with AI in products and in the use of those in, in development and testing and platform engineering for configurations, things like that. So, I mean, just the first release of everybody's product had a chat bot and who, who wants to spend their time talking to a chat bot?

Probably not. Um, but now we're seeing agents, uh, being developed like in lasting and announced their agents. Microsoft just announced their agents, uh, agent force at, at Salesforce and the end user's ability to create AI agents.

So it's gonna be interesting. I don't know if we're gonna see that that curve bend back down in the productivity increase or we're gonna see a lot more experimentation. But what I would expect to see on the other side of it is it's not, it's not ai, it's the uses of it and innovative applications of it that people are finding by being able to create agents or using AI in their development process.

So don't sign on to the belief that AI is gonna make developers less productive. That's a short term issue. I have a pet theory, and my pet theory goes like this.

I think that developers are using AI to do documentation and they're running tests. And you know what, previously they didn't do any of that crap. They were like, it's, I'm too busy.

I can't be bothered. And now that they've got ai, they're actually doing it. And then that is actually slowing things down because all the steps that we previously skipped we're now trying to do Well, and to AI takes some experimentation, right?

You've gotta, you don't just like, oh good, I'm gonna do it this way. AI is gonna, AI's gonna take care of it for me. Well, maybe that's not a good use of it.

ai, maybe it's not great for crawling the repository to find security vulnerabilities and fixing them quite yet. Let's wait till that's really kind of perfected a little bit more. So again, it's this, you know, we're, we're in this learning curve and this adoption curve, the hype curve is this, right?

And we're in this adoption curve with ai. You know, how much of a hockey stick we'll see. Um, but how long that takes.

I, I think this is a just a cycle of continuous improvement process, if you will, of using ai. Another innovation comes out and finding another problem that that's actually really good at, you know, lms, get better at writing secure code. Okay, great.

Let's try that in our development process. And how good is it? Is it good for the kind of code that we write?

You know, just to pick one example. So are all those managers going back and saying, and now go back and write all that documentation that you were supposed to write? I think if they're doing that, they're just putting it in and say, okay, whatever it says, I'm not gonna review it, but I'm moving on back to writing, you know, doing the fun part of the job, et cetera.

I, I checked the box, baby. We're good to go. I'm, Yeah, nobody's gonna read that documentation anyway.

There You go. Except another AI model. Someday going, what the hell is this thing about to do again, Except large language model that sucks it in and says, yeah, this is all wrong.

We wrote this. Oh, exactly. And are we in like a, a, a series of phases with ai?

And I ask the question. 'cause if I look at open AI and maybe these large language models, i, I their general purpose jacks of all trained master of none, and is the next phase of this gonna be maybe smaller language models that are trained for specific functions and they'll just write better code because they'll be vetted and they'll be trained using code that we know actually works. And we kinda went from, you know, isn't this stuff awesome to the, hey, this big thing that we're using for all the same purposes is kind of, doesn't really lend itself to every purpose.

You know, I would like to say yes to that, but I'm gonna say probably no, because human intervention is generally required. Um, I would say that I have team, uh, you know, I, I run a business, I have a team and I have developers on that team. Um, I'm also married to a developer and I was one myself for the first 10 years of my career.

And I would say in looking at the code that it's producing, it's raw. Some of it is interesting. It's, if you're stuck in a corner, I think it's probably better to use the, these tools to get yourself out of a jam or troubleshoot something versus just going to Reddit or forums or places you would go before.

So that's good. But generally it's not trusted. But what we noticed when we were doing this is that it's started to ask, is this okay?

Is this correct? So if you start using AI tools to generate code, you are going to start being a tester for it unwittingly, because you're gonna say, no, actually that's wrong. Because they're, they're betting on us being annoyed and saying, no, no, no, let, let's train your bottle for you.

So there's gonna be some human interaction there, just like there is in search. Um, you know, there are human people that grade the quality of results. Humans are still going to have to be involved in this to get to the level that you're saying.

It's not gonna be the size of the model. It's going to be the accuracy and it's going to be dependent upon human intervention and testing, uh, to some degree in that feedback. Whether this is useful, this is not useful.

You notice, you get asked for feedback everywhere you go. If you go to the airport restroom, you get asked if it, it was a great experience and there's a reason and that, and that's because that is then being fed into the model. Are you Suggest, are you suggesting that we are unwittingly becoming unpaid trainees for LLMs that will benefit the companies that provide these services in ways that leverage our expertise?

Absolutely. But that doesn't mean it's not good. That's the thing.

And talking to my team, they were like, it's kind of wrong, but it's still cool. Like it's, it they like feeling like it's a resource even though it's not a reliable resource. So it's just a more interesting way to troubleshoot at this point in time.

It's a novelty. It's a vitamin, not a painkiller. There you go.

You know, Mitch, I would love to get your opinions on this. I was talking to somebody about this the other day. It's like, so if we all use LLMs and they eventually do get better, won't all the code be of the same kind of quality ultimately and there won't be much of a difference between any of the code that you write or someone else might write.

Because while you all use the same LLM to create it, My my belief is we'll eventually have specialized LLMs, we'll have, we'll have agents that are really good at fixing certain kinds of vulnerabilities in code. We'll have other agents that are really good at upgrading kinda technical data on doing upgrades and software. So, so that today we live in this sea of a massive LLMs that have everything, every, every bad piece of code and every good piece of code that have been written in the world.

And do you think the LLM is smart enough to figure out what is good and what's bad? Probably not. 'cause it doesn't know the difference, right?

It's all based on stats. But when we have targeted LLMs things that are really good on testing embedded systems in a certain process in a certain way, and then we have front ends for that to say, well this is the, what Mitch is asking for is this kind of a problem. So the agent will be use that LLM or a combination of them to do that.

So, you know, it's kind of like the garbage in, garbage out problem. Um, what's there today is what we put in it, and you will get some good stuff in the bad stuff. And the point made earlier, it's a lot of time is spent reviewing the output of it to see if it's useful.

Hey, we didn't like maintaining people's other code. Why would I wanna maintain AI's code? Right?

So, so that's part of that productivity hit too. And again, we're at the beginning at this cycle, but I think we have to have very specialized LMS models, et cetera, to really, because, because software's such a massive domain with lots of different challenges and problems to it, and changing the same time Guy, the hype cycle is off the charts here as always. But you hear this phrase all the time, we're gonna build and deploy more software in the next two years than we did in the last decade.

Do you believe that to be true or do you think that, you know, that's just kind of noise generated by the marketing machines that are running these platforms? Well, You know, that I think it's noise generated by the marketing machines. Uh, I was really pleased, um, to see the, the, the, the DORA study, um, uh, because I like being right and it showed that I've been right, which is that for over a year, well over a year.

I've been saying, and I've said it a few times here, um, that the benefit of AI is not productivity. The benefit of the two benefits of AI are reliability and quality, reliability of delivery and quality, reliability of delivery. Because all of us human workers who've been staring at screens and banging our heads and taking walks and cleaning the bathroom and doing whatever, instead of getting on that thing that looks like an, you know, a, a mountain of of work to do, now you have your stupid buddy, your AI to go and say, Hey, code this thing for me and the AI will produce horrible code and you can insult the AI and then start to editing and quality checking and making it better, and that it improves, so therefore it's more reliable.

You can get more reliable work, um, that has a secondary productivity benefit potentially. Um, but uh, that's what I've been saying for all. And this, this only just goes to, to, to my sense, um, uh, validate that the door, the door results.

Uh, I will say also that, um, one recent thing I've been saying is that this is a whole lot like the PC revolution was where we went for something like 20 years before any noticeable productivity gains appear because of the use of PCs. Mitch, I'm not as, as, uh, optimistic that we will see the productivity gains you're talking about, even with more specialization and tailor, tailor, tailor tailoring of, um, of these models. Um, I think that actually that quality will extend now to things like test plans and documentation and, uh, better use of pipelines and all these other sort of things where this sort of light human supervision and intervention over what are normally considered boring or difficult tasks, um, will allow us to, uh, have better, um, better code bases overall.

Yeah, I mean, if you really understand what what developers spend their time doing, it's less than a day a week that they're actually really writing code. There's so many other things that are involved, right? So just focusing on that, but I'm still kind of stuck back on the stupid little buddy.

So you're saying I need to work on some AI shaming prompts to, uh, insult my AI buddy for doing stupid work. That's Skill. I think think that that would, would be a very welcome, at least for me, development or maturation of how AI works that, uh, you can now, hey, dummy, get your wiggles out or vent to the AI and it'll actually improve the model or improve the response.

Uh, unlike in real life where that kind of, uh, uh, uh, expressions of my deep seated fears or hostilities tends to have a negative outcome, I've actually used it for the opposite. I, I've used AI to take an email I wrote that I knew was not gonna be well received, and I I took it in and said, Hey, rewrite this. 'cause I, I want the, I want the, I want the tone not there, but I want the message and I gotta say it's effective.

If you're having a rage moment, you can have it pull you back a little bit. Really neat idea. That's another example of improving quality.

Yeah. That's where you start it and your dumb buddy changes it for you. Cool.

Of course, you know, Improves it. Sorry, your dumb Buddy improves my mind being the way it is. You can go the other way with that entirely.

You can take a ma an email and say, make this a little more pointed so that I, they understand that I'm a little ticked. They Are adding a new parameter to the models and it's called the snark parameter. How snark do you want this response?

There? There you go. Oh, Heavens Snarkiness.

We, the interesting thing is that, um, in Microsoft's announcement earlier this week about AI agents, they used an email example with McKinsey kind of along these lines. So your experience, and it's kind of reflected in where they see like initial traction or reliability or quality, wherever you wanna call it. That makes total sense.

It does not make total sense to me that philanthropics, you know, computer takeover AI announcement this week that, that really kind of threw me, like, why do I want AI to take over my computer? I can see the only really, like interesting use case for that would be remote workers wanting to have their computers taken over so they are not looking idle on Slack. I don't, I don't, I don't really get what AI is gonna do on my computer.

You know, the The monkey bot, you know, pressing your Yeah. For you. Well, yeah, if, if, if, if we're using Slack to measure productivity, I must be one of the least productive people they're ever gonna meet.

And that's another story. Um, here's the thing, um, here's what I want AI to do for, it relates to developers and I've had this conversation with Mitch over the years and he's had it probably with other people, right? It generally takes me about two to three days to kind of fully explain what it is I want Mitch, the developer, to actually do in a way that Mitch, the developer can understand.

So maybe there's a thing called AI that I could just say, I want an app that kind of does this, this, this, this, this and that. And it presents that in a format that a developer can more easily understand. And then that would really increase productivity.

'cause I think most of the time in software development is wasted in that conversation between what I'm trying to explain what I want in a way that the developer can understand and maybe we should just focus our efforts over there versus just running more code faster. That's a really interesting point. I said merit to that be, uh, dirt a little secret.

All these articles you send me all the time, Mike, I have prompts to summarize all that stuff for me so I can pull out like what are the salient points for that. So in your scenario, um, sitting down, I'm Mitch, the developer, the lead, whatever, you know, meeting with you. Take, take the recording the transcript of our conversation, feed it in and say, you know, this is what I'm looking for.

Pull this information outta that. But the important thing is then give it back to you and say, am I getting this right? Did I, did I walk away with what you wanted?

Or am I missing something? So, and it couldn't be, you could match your own impression, your, what you took away from the conversation, from what you know, your AI summary of it or analysis of it. I think there's a lot of merit to that, kinda analyzing the human language part of it and trying to synthesize that into a better understanding.

Because, you know, Ann will say something and I'm thinking about that and I walk away and I miss part of what guy said. 'cause I'm still thinking about that, right? And that happens in meetings all the time.

So here's the downside of what I just said, because if, and and you may be comment on this, but if I know that there's some AI agent thing that's gonna reduce everything down there with a set of prompts, and then I can read about it later, guess what? I'm not gonna listen the first time. I'm just gonna wait for the summary.

I mean, well Meanwhile, your AI bot will, uh, make it look like you're listening or working. But Go ahead. Exactly.

I mean, I would say attention spans are at an all time low. I mean, look at, look at how, like Gen Z's consuming information, right? 32nd, 62nd videos.

This is how we learn now. This is how we interact with each other. We make up weird words.

And, you know, the attention, attention is at a is at an all time low. Um, I, I don't wanna sound negative about it, but, but I do think that there, there is hope here that it's gonna help us save time, right? So like, there's integrations now for Gmail, for example.

So I spend way too much time in my life reading email, digesting email, trying to understand and parse what a client wants or staff or whoever wants. And if AI can respond thoughtfully and do those sorts of tasks, amen. I I will, they can even, you know, maybe it's some point deal with my mother.

Great, awesome. Like, I, I think there, there are some use cases and I love her dearly, but the messages come in, in all hour. So I'm like, maybe, maybe there's, there's an inbound frontline here for me on the horizon that makes my life a little bit bit better, a little bit easier and reduces, reduces my cognitive load, right?

Because I'm constantly chained to email. So maybe there's, there's a bright spot here. I mean that's where the, I think the knowledge graph comes into play too, Anne, is that processing all the stuff in your inbox and then helping you filter it And so, or summarize it now.

Add add, knowing about what kinds of things you'd like to respond to and what kind of work you do, or whatever it might be. Now it adds context to help it even give you more valuable information to say there are 52 emails, but there's only this thing and this one that you really need to pay attention to. 'cause I know you're working on this today or whatever the thing is.

The analogy I think of is on my Apple tv, it'll pop up and say, Hey, it's fourth quarter in Texas and, and, uh, in Alabama are, are battling it out. They're tight. That's interesting to me, right?

Not that they're playing, but it's fourth quarter and there's something I might want to go see. So think about AI giving us that kind of information, not just summarizing the mountain of information to a smaller bits, but what do I really care about? So how is your mom gonna respond when the AI says, well thanks for reaching out.

Ann has 20 more pressing problems and it'll probably be about 72 hours before she responds to your email. Well, I know full well I'm not gonna give her that level of AI experience. She is my mother.

So she's a smart woman. She's gonna see right through it. Um, she has said to me before, you say okay, a lot to things I say.

So I'm like, okay, got it. You know, like, she, Okay, you don't tell Me. So in a lot of ways, an ideal tester, I, but I probably would tell her.

I I honestly, I'm, I'm like, we're too close. She would know. But I might, I might say, mom, would you try this?

She'd probably say yes. So, I mean, that's the thing is, is it's fun when you're in on it, right? It's not fun when you're, you find out later and that's gonna happen.

And we're already seeing that in, in, you know, there were a famous group of friends with a messaging group and someone used all their, their hundreds of messages to train AI and then made a, you know, made a bot to get into the messaging group. We're gonna see this with catfishing dating scams. I mean, it's, the sky's the limit on this.

All right, well folks, for all you out there who do suffer from some form of attention deficit disorder, we're out of time on this segment and I really don't wanna press the issue 'cause I know that you're gonna go do something else anyway, so hold on. We'll be right back in a minute and we're gonna talk about Smokey the Bear. All right.

I think folks, everybody has seen at least one ad involving Smokey the Bear in their lifetime. And uh, basically the message is always the same, right? John, only you can fight for as far as John, you have a story on this very topic.

Uh, apparently AI is gonna step in and give Smokey the Bear a helping hand here. What's going on? Well, there's a, a company based in Germany called Dryad Networks.

It's been around a couple of years. It just, I wrote about 'em because I just got $7 million in funding for this, uh, IO OT network of AI and solar powered sensors. In a sense, what it's done is it's permanently stationed these drones in high risk forests that are susceptible to fires.

Uh, what prompted this was, uh, in 2018, uh, 2018, there were a number of fires that in California, Amazon, Germany, that got the attention of this, this guy, his name is Carsten Schultz, he's, he's interesting because he's a telecommunications executive who had previously had nothing to do with climate change. And in a sense he had sold previous companies to Blackberry and Twilio. And what happened with the, the wildfires got him to thinking about what could I do with telecom if my next startup and have an impact on society?

And in a sense, what they've done is they've stationed 20,000 drones in high risk forests throughout Southern Europe, parts of California and Canada. And they plan to expand in Latin America and Asia. And, uh, what the, what the, the sensors do is they detect smoldering before smoke.

And they, they can detect gas before the, the fire ignites and I, they've used it. They, they've claimed they've stopped preemptively, stopped a couple of fires, including one in, in a village, um, in Lebanon of all places where a, a former was burning brush and they, it was detected and they stopped it. So it's interesting, this guy who actually had sold companies for profit wants to take this company public.

He wants to do good, um, all the power to him. Um, and he is also addressing something that's incredibly important because with climate change, we have a severe seasons which are leading to not as many fires, but just more devastating fires based on the combination of dry brush and then, and then, uh, the, the growth of vegetation because of the rainy seasons, which then become dry and then ignite. So, um, I understand, Anne, you know, something about this topic or you follow a company in this area?

Yes. Well, I mean, I know my own experience with, with wildfires and, and that was one of the factors that led me to leave the Bay Area. The day without sun was the beginning of the end for me, uh, and my time in the Bay Area.

Uh, although, who knows, I may go back, but, um, I have Well, Hey, I was gonna ask you, was was the, can I, sorry to interrupt, was one of the reasons you left when we had the orange sky in the midst of covid? Remember that day where there were wildfires? Yes.

The day without, you couldn't go outside, Right? Yes. The day without the day without sun.

Um, I'll never forget that day. I, I had to shut all the windows I had, or I had to close all the blinds and turn lights all the way up and turn on the music because your body wants to sleep. So all day I kept try.

I kept, I had to work, but I kept wanting to sleep because, uh, you looked outside, you thought it was dark, you're getting tired. It was very weird. And that was the, the, when the conversations about leaving really got serious.

But, um, I have been advising a startup off and on for quite a few years that works in this field. But it, it works. Um, not in this, this way.

I think it's very smart to use drones for early detection. We know for a fact early detection is what stops these fires from spreading so quickly. This is absolutely what technology can and should be doing.

The startup I worked with, uh, are advised is called Quake, and they work on indoor fires. So here's something people don't really realize. Indoor fires are getting way worse.

Uh, they don't get the attention of wildfires because, you know, they're dealt with in a, in a swift way. There're, there's faster alerts because humans are right there. But because of all the electronic devices in our homes, the cheap IKEA furniture, the pressed, you know, compressed wood with all the glue and chemicals in it, our, our home fires are actually burning way hotter way faster.

And, and they're may way more devastating. So this company built essentially an Ironman helmet using og augmented reality to give, uh, firefighters the ability to see inside, to use the map of a building and, uh, with their fire helmets, uh, uh, use projection of augmented reality so that they can see when there's nothing but smoke they can see. And that to me is also very cool use of technology, but a little, little harder road for them because it's, it's not really what people's minds are on.

They kind of think, oh, this problem is solved, even though the gear that we send firemen into fires with hasn't really been updated in a significant amount of time. Um, you know, other than having the oxygen mask, they really haven't had an update technology wise. Well, why are we sending it in there?

Let's send robots. Well, exactly. We, we have, you know, uh, robots that can detect bombs, right?

And I think drones fighting fires is absolutely the coolest use case for drones that I think there is, it's way cooler than package delivery. Um, but, but it was interesting to me how the funding road is a lot different. And the, and the adoption road is a lot different for indoor versus outdoor fire.

They're very segmented worlds. But, but the fact is that this is something that as humans, we have to contend with. So I think it's awesome.

I think that this article Yeah, gave me hope. I'm very curious about the acoustic thing though. I'm like, really?

Okay. You know, 1, 1, 1 Thing that one of my pet peeves in Silicon Valley, and I think Ann, you share that I think probably all of us probably feel the same way, is just this lack of creating solutions to real world problems. Because it's not profitable, it's not within their revenue plan.

And I found it refreshing. I mean, ironically, this guy's in Germany, of course, but um, it, it just, with all the things we could do, we talk about ai, we talk about drones, we talk about sensors. It's just interesting to kind of apply it to a real life problem rather than, um, kind of sifting through all these weird apps that are so incremental in their use and in their value.

Absolutely. I I, I do think there Covid helped out a little bit in a way. Mm-Hmm.

Got people maybe a little more focused on reality versus, oh, you know, building a faster horse, right? I think though sometimes people expect too much from our friends in the fire department. 'cause when you go talk to them, I mean, their first mission is to save lives and they'll show up and they will then look for the pets and everything after that.

But if your house is on fire, the next thing that goes onto their mind is they're not there to save your house. Your house is already toast. They're trying to save your neighbor's house.

So, you know, it's a whole, it's a different kind of philosophy in terms of how they prioritize their responses. 'cause they're, as a result, because, you know, that's why when they go into your house and, you know, you're like, Hey, you're ripping apart my entire house and walls and everything else because, you know, your house is already toasted. What does it matter?

So, well, My, my first You one, one thing I was gonna mention that, that, that, um, Carsten said to me, which made all the perfect sense, is we have fire alarms in every room of our house or wherever we, wherever we are. So why not take the concept to the forest, which is kind of interesting. Um, so I just wanted, I failed to mention that earlier, but I think that's an interesting concept or our mantra for what he's trying to do.

I, I think that's awesome. I I really hope they're successful. I'm gonna, I'm gonna follow them.

Um, but my first cousin is a firefighter. She's a literal firefighter in Vancouver, and she says she sees very little fires. Firemen are community.

They, they do, she's, she said 80% or more of their calls are as paramedics. So I, I I think that's in an urban setting that that also happens. So our firemen have to do a lot of things.

They have to be community people, they have to be paramedics that, you know. But in terms of actual fires, um, I'm obviously happy to hear that as her cousin who loves her dearly, and she's not going into too many of them, but we really need to upgrade what we're giving them. We're asking them to literally run into burning buildings and we're not giving them the tools they need.

So they're, the Quake has a, a pilot, uh, with a couple hundred fire stations around the country testing this technology. I hope that it becomes widespread because it's absolutely amazing. We should be giving them better tools, uh, to save us.

Yeah, I'm a fan is saying that, uh, that AI is not an app, it's a service and you incorporate ai, the, an AI service into an app. You know, an app that you think of as AI is really just whatever it is that is where AI is an engine. Um, and this sort of, these sort of examples help expand the mind as to how and where and when AI gets incorporated and augments or is, is is providing a pathway to improvement or that wasn't open before.

Uh, we just always want to remember the benefit is not productivity, the benefit is quality and, you know, the human review and intervention. And I would add to that, this idea of iteration that, that you're trying things that a lot of the benefits of AI come through this sort of iterative process where you're not trying to do too much, and in particular, you're not expecting it to just solve an entire problem. So these, these, you know, roving drones in forests and everything, you know, they fill me with a little bit of fear because then I start to think of T one or T two or whichever the one was in Terminator, roving around.

And instead of having acoustic suppression, got a flame thrower. But obviously military use and development of AI for drones is already existing. This is a good counter example to that.

And uh, it's something where the entire system of fire detection, suppression and mitigation, you know, includes predominantly human elements. But why can't AI be an element to it as well? All right, I think we got end this conversation here, but in my mind, I'm having this vision of drones following around people at their campsites as they light up and sit around.

And That was the other thing I was thinking, I was thinking that farmer in Lebanon, it was just like, what the heck, man? I was trying. Seriously.

So the way we interact with the great outdoors is definitely gonna change. Hey, I wanna thank everybody for showing up to me on the Gang today. It was awesome conversation.

We got all kinds of interesting content coming up on Techstrong TV after this. By all means, stay tuned and we'll see you again tomorrow. Take care.