Become a Creator today!Start creating today - Share your story with the world!
Start for free
00:00:00
00:00:01
/GenAI: the good, the bad, the ugly
58 Plays7 months ago
Callibrity's AI Practice Lead joins the podcast to talk about the past, present, and future of generative AI. Recorded live September 2024 at Callibrity's Annual Team Summit.
Recommended
Transcript
Introduction to the Forward Slash Podcast
00:00:40
Speaker
Welcome to the Forward Slash, where we lean into the future of IT. I am Aaron Chesney, your host, with my beautiful co-host, James Carmen, who is supporting the wet look today, and for the first time ever, a live studio audience. All right, quiet down, you busting misfits.
Guest Introduction: Mary from Caliberty
00:01:05
Speaker
All right, so today we have Mary with us.
00:01:12
Speaker
I might try it here. Let's try it. Greg Leski? Close enough. Greg Leski. Greg Leski. That's good. It's a good Polish name. I do like that. That's right. Believe it or not, Chesney is Polish.
00:01:28
Speaker
Um, it's just not one of the rare non ski non ski Polish names. Yeah. We say at the bottom of the mountain. Um,
00:01:39
Speaker
that was, that was a freebie. yeah yeah that's Wow. So a little bit
Mary's Journey in Tech
00:01:45
Speaker
about Mary. um Well, the first bullet point on her intro is currently she is the AI practice lead at Caliberty. She previously was a developer at or sorry developer advocate at IBM and also at DataStax. She's been a Java champion since 2021. Congratulations on your championship win.
00:02:09
Speaker
uh, discovered her passion and developer and customer advocacy. She's majored in math and computer science years ago with the crying a and laughing emoji. I like that.
00:02:23
Speaker
ah Graduated just in time, right around the birth of the web, World Wide Web, the WWW, I almost said WWF, in the early 90s, which would match with the WWF. What are you gonna do, brother? I don't know, but I'm gonna have to look up which came first, the WWF or the WWF.
00:02:45
Speaker
Most recently, she was service companies of various sizes such as IBM, US Cellular, Bank of America, Chicago Mercantile Exchange in topic areas that include Java, GenAI, streaming systems, open source, cloud and distributed messaging systems, I'm sure many others.
00:03:08
Speaker
an active tech community leader outside of her day job, president of the Chicago Job Users Group, CJUG for short, and Chicago Chapter Coley for AI Camp. That's what I love doing this podcast with you so much because you your preparation and your attention to detail is just You know how long it took me to memorize all that? I would have probably just read it. My rap sheet is almost not that long. Almost not that long. So welcome, Mary.
00:03:41
Speaker
Thank you. Thank you, Erin. Thank you, James. No, no, thank you. Thank you for being here with this lovely audience that we have joining us today. Thank you. Thank you very much. I think this is our whole listener base of the room, right? We invited all of our listeners to come to join us live in the audience. That's really exciting for me. It's good times. Good times. Good times.
00:04:06
Speaker
Jumping right into things.
What is Developer Advocacy?
00:04:07
Speaker
ah Why don't you tell us a little bit about dev relations and advocacy and and what that means and how you got into it. Yeah, so developer efficacy, for those of you who are not familiar, um and kind of became the thing. um like I'd say with since 2016, 2017, it's actually, well, what you find is that first of all, they are they get interpreted differently by different companies, but I'd say most of them seem to be you know kind of viewing that as a function of within marketing department, and especially for product companies.
00:04:44
Speaker
So like if you're selling products you want to be able to evangelize and talk about it and all of these things and yeah, so as advocates too is that you were hired to essentially promote what you have their your product and then you know it's sometimes too for example like companies such as like Docker that works with a whole variety of things you can actually talk about the functionality how you can support a different software running in Docker and all of these things and So in other words, it's um it's a you know ah job function in which you're going out to educate your audience and doing outreach work and spread the words and also raise the level of awareness, kind of pretty much a marketing function. um But two is that I think it's also important equally important too is to build up a community around it as well, user community.
00:05:33
Speaker
thing and And then it's not only through speaking engagements, but it can be writing blog articles or running workshops and all these kind of things essentially raising level of awareness about your product.
00:05:46
Speaker
in that's what It has been, um but then it's interesting. I have to say then I you know got invited to explore opportunities within Caliberty, which is not a product-based company, it's a consulting. so um And I remember, I actually asked James, I said, well, you know you have advocacy in a consulting firm because I never heard of it as much. But he was saying that, well, likewise, do you want to be able to raise awareness of your company through the services that you provide? so Yeah, so it's kind of same thing is promoting. Yeah, not only that, we also advocate for certain technologies and they're in the proper use of them as well as consultants as well and kind of help guide companies into, yes, if you use this suite of products, you know, you're going to see these kind of X, Y, and Z benefits and that kind of thing. And you should maybe avoid these kind of technologies for your space for the same reason. and
00:06:41
Speaker
um And I imagine you handed out lots of cool geek stickers
Benefits of Advocacy Roles
00:06:45
Speaker
too, right? Yeah, that that's true yes yeah that's that's true. Yeah, true, yeah. But yeah, essentially that's what that's what it is. I think the kind of like I do have to say the fringe benefits if you like to travel like as much as I do. I really like it. And so you speak at conferences, but of course you have to be you know you have to be responsible for applying to conferences. But then once you're known,
00:07:07
Speaker
then people will also come and invite you to speak as well, that type of thing. So it's nice. It's sort of an extrovert side for the engineers, yeah basically, because I was an engineer before, for over 20 years, and I didn't know about this. And until I tried it, then I realized, oh, this is kind of nice, because it's about time I wanted to go out and reach out, not just like working as an engineer. You're also going and share the you know the news, yeah good news, kind of in some sense, yeah.
00:07:34
Speaker
So were you always kind of extroverted anyways, like speaking up in front of groups of people and and presenting? Interestingly, I have to say when I never thought it was funny. I never thought I would become an advocate. um But I think certainly, I'd i'd say I have, and and I think for most people too, it's not black and
From Introvert to Advocate
00:07:55
Speaker
white. It's not like you are completely 100% introvert or 100% extrovert. A lot of us are probably a mixture.
00:08:02
Speaker
So I certainly have that introvert side of me, but that appear as the dominant trait for me in the beginning. I was more shy and didn't want to speak. And then um I remember too, also when I first came out, you know graduated and started working, I mostly meant, I do have to say, and then happens that they are all very nice people, but also very much to themselves. So I didn't really have the influence or the role model Within me so I thought that's what it it is and plus I grew up with a very nerdy geeky brother who was very annoying Who actually was saying that no girl I remember I actually went to him. I still remember to this day I like I couldn't figure out some math stuff I think I was in I don't fifth grade or something and he's three years older so like in seventh grade equivalent and And he was like, you no, I'm not going to tell you because girls can't figure it out. I remember I was so mad. Yeah. And I was like, no. And I was like, OK. Well, but in some ways, it was weird. Then you internalize it. Maybe it really is true. Girls couldn't do it as well as guys would be. right So I was kind of under that kind of shadow. But anyway, to go back to your original question, I never thought I would. And so when I first started working, I was um also like an IT company or IT department of actually Monsanto chemical company. That was my really first job and ah in the basement and chemical plant. And so
00:09:31
Speaker
They are mostly men, but there are a few women, but they were also doing more of the interacting with customers, internal customers, and then they were all just like not talking too much and something, and I just thought that that's the way it was, and I never thought of it.
00:09:47
Speaker
um But then it was funny, too, as I get older, and then I realized, oh, something is missing, right? It's just like, I like this work, and but it's kind of something's missing. um And then, because my daughter grew up, too, and then she started going, you know, to... to ah college and all that, and so I thought, oh, let me join user group, and naturally I was working with Java, so I joined the Java users group in Chicago, and started going, and I really found out I really enjoy it, so I was going like crazy, like because too, at the time I was working in downtown Chicago, so I have to take the train in and out, and then I thought, okay, let me, and and the nice thing is that going to user group, you get free dinner, right? So you get pizzas, and so it's nice, and you talk with people, you get free dinner,
00:10:29
Speaker
So I kind of, the more I did it, I really like it. And without knowing, then IBM happened, like 2018, I remember. They were like, oh, looking for developer advocates. And then even I was trying to figure out what advocates do. And then I thought, okay, let me try it. And I started reading the job description. Oh, user group community, I said, oh, I like that. That's what I'm doing now with user group. Oh, not only, actually, before that I should say,
00:10:55
Speaker
The Java users group at that time, the organizers came to me and said, you're coming so much, why why don't you help? It's all volunteer work. So then I said, OK, sure, let me start helping. And so that naturally led to IBM looking for people and then reaching out to Chicago. That seems to be a pretty popular, well-known user group for Java.
00:11:16
Speaker
So I thought, oh, let me try it. And then it worked out. So then that's how I started doing. And then I realized, oh, I got to start actually going to speak to. That was another thing. I mean, I started talking too, but just lightning talks, all these things. And and then being a full-time professional person, I have to start talking. And then I'm like, oh my gosh, I got to start preparing all my talks. and And they asked me to talk about streaming and reactive systems. And I have to start studying and and doing all that. But I was thinking,
00:11:44
Speaker
I better do it now. you know I ah can't like make ah make a fun of myself. um and And I think also, too, that I was in a group, the Java Developer Advocacy Group at IBM at the time was four of us. And three other guys, they were constantly traveling. the german One of the guys based in Germany, he would be always somewhere. I mean, you talk about me being crazy, he was on all the time traveling. So it just kind of became the thing. And I think you know it is. It's funny, I think it brought out that extrovert side of me that I never thought I had. So yeah um so I think that's how I kind of converted to doing advocacy.
00:12:22
Speaker
yeah i did a ah kind of a It was but something that was provided by the company. I think it was over several weeks. And it was based on the Myers-Briggs personality profile things. And the way they did it, they actually like graded you in in the different um areas. And one of those is introvert, extrovert. And so you had like a percentage. And depending on how far you swung, either way she called it, she called it, you're either a raging extrovert or a raging introvert, you know, depending you know depending on how much. And I was actually paired with a raging introvert and I was a raging extrovert.
00:13:06
Speaker
So we we challenged each other you know to to um you know drift more towards the middle a little bit more because, yeah, you're absolutely right. There is ah there is that bit of introvert in all of us and extrovert in all of us. But no surprise, I'm extroverted.
00:13:29
Speaker
So what was the first ah talk you gave where you're you know you're kind of standing up in front of the crowd, and you're like, OK, this is something. like This is a big crowd. like this This is a whole new world for me. What was the first one like that
First Major Speaking Engagement in Ukraine
00:13:39
Speaker
for you? Yeah, good question. Now I think back, I think it was actually in Ukraine. um and yeah But i I did some small things. Oh, although there was Java 1, but that was something else. I forgot. yeah and then But anyway, for international, it was in Ukraine. and But the crowd wasn't too, too big. So it was still scary, but sort of felt more like a meetup group. And at the time we had about 60 or 70 people coming to meetup groups. So that was about the size. And again, I was told I joined the Java users group and i mean the java advocacy team at IBM and then
00:14:13
Speaker
Some folks are talking about spring. There's one guy talking about more carcass or at the time more with the enterprise microservices. And then the other guy talking about something else. And they said, oh, can you talk about streaming and reactive? So ah I started talking about reactive systems. And that was my initial thing. Yeah, yeah.
00:14:30
Speaker
But I felt like, too, at that time was a little scary because reactive streaming is a bit more abstract than the other thing. And I worked a little bit with it. So I had to come up really quickly, like try to do some samples and being able to put together a talk, yeah, like that. But yeah, it was scary, I think, in the beginning. And I was getting getting um some feedback from another IBM person. I call her IBM Mom. She's now at Red Hat, but anyway, she's funny, she's very honest. She said, Mary, you talk too fast, like that. It was really funny, but but I think it's just that over time I started, because too, is that I started talking so much that I the force i was being forced to practice all the time. And then in reality, and you know in real production.
00:15:15
Speaker
So I think it becomes more natural to me now to try to talk. And also interestingly, English is not even my first language, but it became my first language now. Yeah, so yeah. You would never know it. You do speak English. That's right. No, thank you.
00:15:30
Speaker
better new but thank you appreciate But I think I still have some accents that I certain words maybe I still can't figure out exactly but I still have those kind of stutter moments. Yeah, my daughter makes fun of me too.
00:15:48
Speaker
but yeah Kids can be great, can't they? I know, exactly, yeah! That was being facetious. Oh, okay. I used a big word Aaron, like facetious. So, streamings and stuff like that. and Doesn't seem to be I mean it's it's big in areas, but it doesn't seem to have caught on like I don't know maybe Did it fall flat like what happened with the whole streaming and reactive systems and that sort of stuff? Actually very good question and as you know right now Java 23 or even before that right virtual threats coming out and in fact, that's what people are saying and When it first came out, we were saying that virtual threats gonna wipe out reactive systems and streaming, all these things too. But of course it won't. But anyway, but going back to streaming, we were talking about Java streaming and data streaming, asynchronous processing. I think it can mean a variety of things, but primarily when we think of streaming, it's more about Kafka too. So it is definitely being used because
00:17:00
Speaker
I think even when I was evangelizing about it, I realized it's a more natural form of processing like to do streaming because it's asynchronous. you don't It's not about blocking. This is yeah like all these things. so I think it is being used. It's not completely wiped out, I don't think. And also, too, as I am talking to a lot of people, they are saying that, oh, we're using Kafka most of the time, right? yeah And they're maybe using RapidMQ. Those things kind of streaming, not quite exactly. So Kafka, I think it is still...
00:17:33
Speaker
heavily being used but just not talk so much and also I have also thought too is because it's more of a lower level infrastructural type of thing that people just don't talk about it's like you're trying to show your house you usually say oh this is the you know my living room all of these things you won't be saying that those are the pipes in my house are like that so I feel it's because of that you know the function of streaming is tend to be more of an infrastructural thing, yeah, that you must have, and it must be well-designed. But as a result, if you put in very good pipes, then no need to worry about it. But if it's only when things go wrong, then you kind of talk more about it. Maybe, that's why, that's that's what I think. Yeah, I guess I was thinking more on the reactive side, that doesn't seem to be yeah as popular as everybody thought it was going to be. Oh, yeah. It just seems like the the coding patterns are much more complex, right? it's Yeah. that's youre You have to think much more. and and
00:18:28
Speaker
Just a lot more code involved to write reactive systems. It seems like I mean everybody's tried with frameworks and everything, but I i don't know I just yeah, yeah, it's a different brain. i think for that I totally agree with you. Yeah, that's right. I don't have that brain. No, I was following the streaming thing because you know I was I always thought of like the Java streams is kind of like that. It was kind of really close to the code in how you're processing behind the scenes, but not necessarily right up in your face. It's not the thing you want to demo, right? It's not going to present well to a customer. It's like, look, and we're streaming this in, and it's coming in live. And it's like, to what? Obviously, live things aren't very impressive. We haven't had any sort of hoops and hurrahs from our audience today. They're all in there. phones and computers doing other things. Which is like most of our listeners. Any questions from the audience for Mary? No questions. Ryan, I'm sure you have one. I thank you long time listener. First time caller. Thank you caller.
Advice for Beginners on Generative AI
00:19:40
Speaker
What's the best way for a beginner to start with generative AI?
00:19:45
Speaker
I think um to me it's you know AI is not it's not going to take away what we're doing right like say we're already doing coding let's say most of us we're doing coding you need to have data parse it and filter all of these things transform and all of these input output So you're not going to take away. But then for generative AI, I think definitely you you want to start reading up on it too. Because to me, generative AI is like a branch of AI and machine learning. And it deals with more of the generative side. And before yeah before that, it's more predictive AI that's actually not not making use of natural language processing before, not as much or at least. And then dealing more with predictive, you're predicting
00:20:31
Speaker
You know, based on some large data set, you train it, and then you also generate well you also generate results, but also it's more predictive, and also like making use of statistical model, more of that, and more of a timed type of value. But they like generative AI is dealing now, we see that you can use natural language in talking.
00:20:51
Speaker
And also to have to understand that with generative AI there we're dealing with different modes of data. There's texts and that's what it came up of to chat GPT with the text with the spoken and you ask and it comes back with answers in typing out all the strings or paragraphs of data. But then also there's images as we know and videos and code and generates code and all of these. So we have to understand the different modes ah that it can deal with text to image, image to text, video, 3D, all of these things too. So I think definitely to to kind of long
00:21:25
Speaker
One answer to your question is, yeah, start also studying to some of these theory behind and and understand AI and ML. It already existed before Gen AI and we need to understand to the models are still being trained, supervised training and supervised training, all of these things. So it's still kind of good to understand the traditional machine learning AI so we understand more of the training of the model. And then on the Gen AI side, we're also spending a lot of time to figure out how to make use of it and then develop applications, doing system design, all of these things too. that yeah So I think to kind of say to start it, to understand theoretical side plus also start working with it and using it too. right On the chat GPT we can go to different chat GPT and and use it. Now that you can also do it locally as well. So interact with it to get a feel of it too. I think that would be a good way to start.
00:22:17
Speaker
yeah in In the predictive AI, it's been around a long time, right? that's Because we've been using seicial statistical data models to to you know make informed decisions. in We've called it many things throughout the year. so is One was like knowledge base, rule systems, yeah you know some of these other things. systems Expert systems, yes. And large language models have been around a long time, too. like
00:22:48
Speaker
You know, back in the old Oracle versions, you had to use all capital letters for your column names, which was a large language. yeah And you'd model your data on it, so it was a large language model. There we got a little audience participation. a Good job. I was wondering where he was going with that. I'm like, no, they haven't been around a long time. Multi-layered neural networks have been around for a long time, but not large language models.
00:23:13
Speaker
Another thing I found like with chatgbt and stuff, like you starting out using it is ah like I just figured out the trick not long ago. like tell it like Have it teach you so you can ask, like I'm trying to understand this or whatever. What would be the way to write a prompt so that you can better answer me?
00:23:28
Speaker
And it's pretty good at that. It'll say, oh, you need to structure your prompt this way. Maybe give me some examples or this stuff. It actually does a pretty good job of like training you how to talk to it. so Kind of like a recursive function, like you kind of loop back on itself. That's kind of cool. I don't have to try that. Yeah, it's kind of like a recursive function, and it loops back on itself. Oh, and and just repeats over and over again.
00:23:49
Speaker
It's kind of like a recursive function that just kind of calls back on it. I hadn't thought of that, but it is.
00:23:58
Speaker
wow yeah how many How many layers deep before we exit? Before we stop. like we're We're looking at each other like, just keep going, man. Just keep going. keep going So you you mentioned like you know getting getting into AI. I mean, we're you know at Calibri trying to build our AI practice. And it's the one of the hard things is deciding, like how are how are we going to use this? And it's primarily generative AI. I think they figured out most of the other traditional AI flavors have been applied to business problems through the years for a long time. right So that that's kind of figured out. It's this generative AI thing that everybody's like,
00:24:33
Speaker
how do we use this to solve business problems? What are you seeing in the market as like some trends of you know where folks are starting to kind of lean in and say, this is how this can truly help us solve business problems?
Integrating Generative AI into Business Solutions
00:24:45
Speaker
I think for me, it's definitely not ready. um There's still a lot of people are are kind of getting not excited more more like a while by, okay, you can ask it and we've all done it.
00:25:00
Speaker
Go to chatgpt.com and just say, oh, I don't know about this, and then have it answer you. Those things. I think those are still wowing people, but then if you kind of think of using that and trying to put together an airline reservation system, that won't work. We know it. Yeah, so yeah that's right.
00:25:17
Speaker
So even I was trying to have been trying to figure out how are companies using it, and then I've been involved with the user group. We're talking about it, but I'm not finding a lot of actual you know situations in which companies are developing some tangible systems.
00:25:36
Speaker
But the thing is that I've seen this more about okay they're, they're a vendor trying to come up with tools that are well now their libraries lane chain for example, those things like crew AI doing agent multi agentic, it's more like on the tools side, and I think more.
00:25:51
Speaker
But at the same time, too, it's interestingly, I was just recently in Norway in Oslo for the conference, and one guy came and we had talked because I did a workshop there, and then he was telling me he works for a company that does that process large amounts of data, but they generate different graphs. and all that stuff for you know like you you can be taking a lot of data and basically have the bot, right the AI itself going LLM and figuring out all the data and then sort out everything and then generate all of these beautiful graphs or kind of diagrams to kind of show things. That seems actually be working quite well. So if you think of it, it's more about, okay, I'm looking large amounts of text to image, I need to do those things.
00:26:37
Speaker
I think that would be a more safer bet to use generative AI approach because you can be still, you let's say, huge tons of numbers you get from IoT devices you gather. Then you want to maybe see some things with those data, maybe like how many trucks you know in my fleet is like doing some things, where do they go, that type of stuff. We can quickly figure things out for you and express it in diagram format. I think those things probably will see it a bit more.
00:27:06
Speaker
as a real use cases, but like not not quite yet, right, for doing ah an airline reservation or other the kind of things where Caliberty's doing maybe for financial institution trying to, oh, doing something, you you need some real data to predict some something, let's say, that you need something more dealing with accuracy of of your information. I think that's not ready yet, but doing more of the image yeah So we can go on the record as as saying that developers are not going to be replaced by AI any time soon. just yet not just Not just yet. I think the hallucinations is probably like probably the biggest thing yeah the industry is worried about right now right that with the AI and hallucination. It's like, I don't know, I just thought of this analogy. I'm an analogy person. So it's like like talking to Ron Burgundy from Anchorman. When he says something to you,
00:27:54
Speaker
has the utmost confidence, like when he explained what san San Diego meant. He was saying it with the utmost confidence, and he believed every word he was saying, but it was completely wrong. Sometimes that's how it feels when you're dealing with this AI. I'm very important. Yes. I have many leather-bound books. Yes. In my apartment smells of rich mahogany. Well done.
00:28:20
Speaker
teing you up for that yeah thank you i appreciate that i only said that about like ten or twenty times yeah until i got a write it is now committed to memory you got it now i think so question from the audience So for companies who want to jump on the AI bandwagon and they feel like it can help them out, but they have no clue how to get started, what do you tell them?
Utilizing AI to Solve Business Problems
00:28:47
Speaker
Call Caliberty. Yeah, that's right. call I like that. Call Caliberty. Yes, we have yeah the best solution, um I think. right um i well First of all, we need to figure out the kind of problem we want to be able to solve.
00:29:04
Speaker
um Actually, one thing too is that James gave me idea because he's from our financial officer talking about, okay, we need to process all these invoice every month. um And then from the invoice, right now we're manually looking through and sort through and look look for specific types of data inside each invoice. So those kind of things. So first of all, I'd say, you know, for a company to figure out what areas We need more you know help in terms of reducing the redundancy of human beings having it to intervene and and sort out, get data, maybe those kind of projects. And once you identify, then you can kind of look at all the tooling to see how it can you know we can kind of leverage on Gen AI to do it.
00:29:48
Speaker
So, let's say one thing by with GenAI to me, it's like it's not it's it's like we're shifting to the kind of programming that we need to do. Let's say traditional programming. We need to process data. We have to exactly know what types of data comes in. We need to process, transform, doing all these things. But now we can maybe use GenAI to say, okay, I don't care what data it is. Here it is a file, this invoice.
00:30:14
Speaker
Go figure it out for me, like from the data, figure out um I need to look for certain types of invoice for this category of spending. I'm just using an example. Go figure it out for me. Those kind of things is what we can leverage it on rather than us having had to explicitly program some things to do it. I see that as a possibility of of working with it.
00:30:36
Speaker
Now, another kind of solution, as I'm also learning too, as I'm talking about this, is that I'm reminded too, let's say of some big companies, right big big type of companies like Microsoft, they have Copilot, and within Copilot, they have all these document processing studio doing this and that, all of these things, and they are purposely built for so processing certain types of document, for example. Then you can also use that kind of like higher level, I guess, product,
00:31:04
Speaker
um kind of cookie cutter type of product less interesting for the programmers but maybe more useful for business and users you can use those things and leverage on those and then maybe then in some companies environment I'm already a Microsoft user then I can leverage on this particular feature of the whole platform set I can use for processing documents, something like that. And so I think, you know, just I hope I answer your question, but I think that's also another way of like solving end user type of problem using kind of this kind of big company platform product.
00:31:40
Speaker
I think another thing, too, about using the tools that are out there is, like you were saying, you know, learn to learn to use the prompts. And I think, you know, one of the one of the things that from the Agile and Beyond Conference that we were at a couple months ago AI was a big topic there, and one of the things was it's a very powerful tool, but you know you need to know how to use it. And I think, you know, a lot of ah lot of discovery needs to be involved in how can I use this, you know, properly to enhance what my my processes or my my generation of certain things. You want to break it for ship it or skip it?
00:32:27
Speaker
Ship, ship, ship, ship, ship, everybody, we gotta tell them if you ship or skip. So the first thing, oh let's go with something in the Java world. Let's talk about Spring Boot. Spring Boot, I mean, it's just in general, right? We're talking about, yeah, ship it. It's kind of enough usages everywhere, ship it, yeah.
00:32:57
Speaker
You're going to be the contrarian on this one? Nope. No, I liked it. It wasn't a real controversial thing to throw out there. Was this just letting her warm up or what? She's a Java freaking champion. I know. Well, I wanted to hear from the champion. the champion She's got the ship, the champion ship. Did you get a belt with that? Did give you like a big wrestling champion? Even like a rodeo buckle or something?
00:33:23
Speaker
got my drama notebook All right, so that was that was a warm up question. Yeah, that was the warm up. Oops, now it's... Okay. Okay. What about aspects? Like aspect oriented programming in Java? What do we think from these days? Oh, sorry. Yeah, retain your cell phone there. you can I guess skip it. Skip it? Yeah. Why?
00:33:47
Speaker
I don't use it. No, I'm just kidding. But it's, yeah, I remember, yeah, it was being talked about, right, earlier, maybe 10 years ago, something, but I don't know, maybe this, I think because I also haven't been using it, ah or kind of no opportunities to use it, I don't see it as something necessary, I suppose, yeah.
00:34:11
Speaker
You probably can tell me more about that. I don't know if it really is used as much anymore. Maybe the frameworks themselves use it, but I know I haven't. I think they use aspects for things like the at transactional annotation processing and that sort of stuff. They do weave them in with aspects and you just don't know about it anymore.
00:34:32
Speaker
behind the scenes and plumbing maybe, I don't know. But yeah, i don't I haven't used like myself aspects directly in many, many years for sure. You are so lucky. Am I? Yes. OK. So i've I've lived aspects nightmares. There are specific use cases where it's it's really useful. um Annotations is a good example. And I think you're right. I think that's probably we've woven in with aspects.
00:35:00
Speaker
um you know, method logging is another one. So if you wanted entry and exit logs to your methods, aspects work really well there. um But in general, if something is breaking in your application and you don't know why, check your aspects. Because that's usually where it's gonna break down and you're gonna have, you know, spend a week or two trying to figure out why.
00:35:24
Speaker
So for me, it's i I guess it's more of on the fence. If you have one of those special use cases where you need something integrated across the board and it's you know it's exactly the same code everywhere, yeah, go ahead, use it. But those use cases are so rare that for me, it's mostly a skip it. There's there's probably a tool that's already doing it somewhere.
00:35:54
Speaker
That's what I was going to say, probably debugging may be hard. I mean, if things work fine, by hidden is fine. Yeah, that's right. That wasn't possible. I think a lot of the the use cases where we were where we used to reach for aspects, I think they're just falling back on dynamic proxies most of the time for that stuff anyway, right wrapping methods with those dynamic proxies at runtime. and it And it does a good enough job where you don't need to go crazy with load time weaving and all that stuff.
00:36:20
Speaker
I used to say this was somebody's doctoral thesis, you know, code submission yeah for computer science and and not meant for reality. All right, let's let's switch to an AI one. ah Building your own LLMs. So that's large, or yeah, large language models.
00:36:46
Speaker
Was that the uppercase thing? Or is it the real ones? No, no, the real ones. Not the database ones. We're going to update to 2024.
00:36:58
Speaker
getpbi Skip it. Skip building your own. Why is that? First, it's very expensive. We know training models are not easy. It uses a lot of GPUs. Yeah, so you don't probably don't want that. and And also now you can actually use like RAG, for example, to augment right what what's missing in a general purpose LLM set, which are all out there now, chat GPT. Anthropic has their own and all these Mistral, LAMA3, for example. So you can use RAG to supplement that too.
00:37:33
Speaker
Otherwise, it's yeah as such, right lms are like all the generative AI, they are very expensive, and from a sustainability point of view, it's expensive. You need to train it. Nobody even thought of the fact that your data center has to work harder to be able to produce these models, and inferencing, too. Likewise, too, it's not cheap, too.
00:37:54
Speaker
So yeah, I mean, as much as like, yeah, chat GPT is fancy and everything, I don't think at this point, you know, there are a lot of, I mean, there there have been efforts to try to reduce the carbon footprint over these things, but it's still gonna take some time. so So not training your LLM for sure, not do that. So it sounds fancy, you train your own and adjust all the parameters, but it's not very practical.
00:38:20
Speaker
Yeah, I wouldn't think so either. The only thing I could think of is if you're if you're doing like a gen AI that is like something that somebody hasn't done yet. But even at that, I would think you'd still want to leverage someone else's LLM for that.
00:38:41
Speaker
i agree um ah skip skip it on run and Yeah, I don't think I'd ever want to like in unless you're working at a like a place that you're doing research and and you have a lot of money to burn No, I don't think I'd be training doing full You know training of my base my foundational model, but you know like doing training on top of it and fine-tuning those models, maybe that can still get a little expensive, but that that would I would go for that. But you're right, like with retrieval augmented generation these days, you can provide enough context and give some one-shot or two-shot example of here's what I'm thinking and it can figure things out pretty darn well. So I'd rather go that route to start. All right, you want to throw one more out there?
00:39:23
Speaker
ah
00:39:28
Speaker
Maybe we get one from the audience? Yeah, let's get one. The audience got any ideas for topics that we could ship or skip it? I don't think we're in the audience. Audience? Hello? Hello? We're thinking? Oh, cucumber. That's a good idea. Yeah, cucumber. What do you think about that? Ship it or skip it?
00:39:55
Speaker
Oh, hmm. Like more testing, right? Testing to produce a BDD thing. Yeah. That's right. Yeah. Yeah. Cucumber. That was another one too. I worked on not cucumber, but it's easy. Jade.
00:40:11
Speaker
da some shame behavior ja behave that's why that's what i use yeah you remember that yeah there was something with the name dave in there i thought at one point too um anyway's right yeah but okay cucum but I have to say I haven't worked with cucumber, but I understand it is becoming quite popular. But otherwise, too, before that, I work JBehave, those things. yeah kids Yeah, I don't know what to say. I haven't been working so much with testing. I think I'm more neutral. Yeah, like you can use it if it's, you know, kind of in your environment you have already set up. But if it's not, then it may not be necessary is what I'm thinking maybe we can leverage on AI assisted tools to
00:40:53
Speaker
Do all the testing and then maybe right so then you don't need to actually use this tool. I'm just and think I've heard of that use case before of having AI analyze your code and suggest and then goes Yeah, that's right. Yeah, I like and I like that. Yeah, exactly. Yeah, I've got a soft spot for testing. Yeah i when i'm coding hits right between yeah When I'm coding, I do have one of the was that like copilot or one of the things turned on in my ID and it's it is really nice as I if I if I write the like the production code and then as I'm writing tests.
00:41:28
Speaker
it will generate like, oh, this is the test you need to write. and And I find myself, myself spending a lot less time writing my test cases now with, uh, with a little bit of AI assistance because it because it kind of sees the production code and in it. You've got an edge case here and you've got, so it kind of knows. who kind too Some of us right here just before we code. Yeah. I'm going to know what I'm testing. If I don't write that first, yeah write a better prompt.
00:41:55
Speaker
So for for me, Cucumber is one of those things where it depends on where in the stack you're developing. If you're if your're structured horizontally to where you only have backend code and you are on the hook to demo something, I think that Cucumber gives you a good way of verifying that your story, the way that project owners are expecting it,
00:42:23
Speaker
is passing to the specs that they've laid out. And it puts it in plain English instead of having to show postman or code, um you know, behind-the-scenes stuff. You can have something that is in plain English. You can see, here's the here's the values that we pass through it, and here are our results, and you know and and this is the story the way you expected to see it. And I think it presents for a little bit better of a demo of those backend systems that don't have a UI in front of them. If you have a UI, I don't think so. I mean, you can visually test very easily in the UI, and it's probably less effective ah to have a heavy testing system like Cucumber in place. Because you do end up writing a lot more code when you're using Cucumber on top of your unit testing.
00:43:17
Speaker
I don't know. I'm kind of a skip it guy on BDD. I haven't really, at least from the the selling point, was, oh, business people can write the test because it's in English, right? And of course, we can't do it. We can skip that. That's not going to happen. Yeah, skip that part. that' Now what I will say, and I think we we talk about this because we cover BDD in our our training module, but like the one thing that it can be really good at is when you set up your test cases where where it's written in such a way that it involves a table with the examples below it.
00:43:46
Speaker
Setting up the structure and the kind of you know how to do it and then underneath it with the table because business people love spreadsheets They can generate cases and cases and cases that you would put in there and you get all those basically tests for free So it's just cranking through that all that that list of of examples That I think is beneficial, and I do think business folks, I'm using air quotations, business folks would absolutely embrace that. But but as far as like, oh yeah, all ah the the business folks are just going to sit down and start writing out this gherkin files, they're they're not going to do that. That's never in the history of ever been done that I've seen no in my experience. Well, I did do it when I was a business owner.
00:44:24
Speaker
but Yeah, but different. But I came from development. yeah So i was I was different. I was i was that rare zebra right with white stripes. I actually remember that one TDD thing I used to use, probably very old, is called fitness. yeah Have you heard of it? Yeah, OK. Also, I remember. We had one test suite that was written in fitness that was um amazing. And it was for citation generation.
00:44:52
Speaker
So like in English when you put all your citations in for the different resources that you use, it had all the formats and all the different types of citations and it would test every every permutation through this generator and it was was a beautiful thing.
00:45:10
Speaker
I've got one other thought for skip it or ship it. was Very related to what you guys are talking about. What do you think of generative AI for creating test data, especially like nicely anonymized test data to give you a decent sized data set to test with?
AI for Test Data Creation and Anonymization
00:45:23
Speaker
Great question. And I was going to say, yes. In fact, um one of the things I'm actually asking some of our bench folks, if they can have time, is is basically making use of this technique called named entity recognition. And it's essentially a technique within like natural language processing that you could, like and again, we know it's the name implied. It's named entity recognizing your data and being able to apply that technique and mask sensitive data in your, you know, for for testing purpose, especially at banks, you know, you want to use real data, but you want to make sure all the secure, you know, all the data, privacy data, not expose all of these things. So, yeah. And I think it is actually a but very good, would be a very good usage too for Gen AI and NLP, this type of, yeah, usages too. Yeah. And yeah.
00:46:18
Speaker
But I think it's also I'm trying to figure out, too, if there are actually tools out there that are making use of this technique. Yeah. But but here in Caliberty, I'm already asking some people to help us look into it. So yeah.
00:46:31
Speaker
so you ship that prove Yeah, this one is ship. Yeah, sorry. yeah you I like that idea. It's a huge problem, test data management. yeah yeah i you know i was I actually worked on something that did statistical analysis on chip test results. And we had to try and model normal distributions and then anomalous and data distributions so that we could test whether our software was working because it was supposed to detect these things.
00:47:02
Speaker
And man, yeah, if we would have been able to generate those things with an AI, go like give me a normal distribution with one positive outlier. you know That would have been amazing. That would have made my job so much easier. I might not have even needed you. and Maybe.
00:47:20
Speaker
I don't know, they were they were idiot-proofing their software. and I got the biggest idiot, so I was in charge of QA. so so
00:47:31
Speaker
cra yeah i guess so What do you think about the new OpenAI 01 model? like would i I haven't really had a chance to mess with it much. I haven't played around, but I think it's basically is better in terms of the reasoning part. it is That's what its strength is. It's is still in like preview right now. yeah I took a quick look. I haven't used it yet. But I think also, too, there's another brand within this AI thing. It's called AGI.
00:48:00
Speaker
um artificial general intelligence. That is, I wonder too, I've been like wondering if it's that O1 Mini is kind of moving towards that, the the general intelligence. you're You're actually really training your bot to have actually general and intelligence, yeah as opposed to more artificial, I think, in that sense.
00:48:21
Speaker
Yeah. I haven't used it, but I'd like to try. Meaning it's like more well-rounded, like it went to college and learned like humanities and stuff like that? Yes. It's General intelligence, yeah. So it can win Jeopardy. Yeah. It's something like that. Oh, OK. That's why. I thought that already happened. I already did that. Wasn't that new? Oh, Ken Jennings beat it. Nice. OK. So you think it's going to be an interesting step towards that? OK. All right. It seems like it.
00:48:49
Speaker
Well good, now I don't have to get general intelligence because I can use AI for it.
00:48:58
Speaker
Alright. Alright, let's go to our lightning round.
00:49:23
Speaker
All right. So the way our lightning round works, we're going to ask you a series of questions unrelated to technology in most cases, but short answers. We're not looking for, you know, any explanation. There's no right or wrong answer for these. These are mostly based on your personal opinion and feelings. So.
00:49:42
Speaker
What age do you want to retire? Well, i i you know I'd say I'd like to retire sooner than later. However, even if I retire, I still like to do things, I think. Yeah, but just maybe like less restricted with the company, you need to follow all these rules kind of thing. But really, I'm passionate about doing some things. So i'm I'm going to continue, right? Like, say, I can be a consultant working on some things that I like to work on type. But otherwise, too, I'd like to retire sooner.
00:50:12
Speaker
All right. These are like really the search deep within yourself to answer these type of questions that we we give you. All right. Sweet, salty or sour. So tough. Oh, my gosh. Sweet. Sweet. All right. We got a wrong. Let's go with is if you could travel back in time, what period would you go to?
00:50:40
Speaker
Maybe the 80s maybe? I've been there. It's a good time. yeah i think good time good deal I like the 80s. I miss the 80s.
00:50:57
Speaker
okay If you could eliminate one thing from your daily routine, what would it be and why? I don't know, really don't know. I'd say my commute, you know. Oh, yeah. Well, and I hardly commute now because it's like I'm. So I've already done. Yeah, yeah that that's right. well But it's true if I have to daily going into work. Yeah, commute would be would be good. Yeah.
00:51:22
Speaker
But sometimes I think of, you know, I don't have enough time in a day. Sometimes I wonder, Oh, if I actually you don't need to sleep then like more time, that's kind of a weird thing to think of. Although when I'm sleeping, then I enjoy sleeping too. So yeah, but if I'm going, keep going, like i want to just keep going and not have to to be tired and, and, um, yeah, it's just too many things to do. If you had to sing karaoke right now, what song would you choose?
00:51:50
Speaker
Oh my gosh. it you You don't have to. Let's be clear. you yeah I don't have to sing. But by you can choose what song. If you if you had to sing karao karaoke, what song would that be? And we won't ask you to do it.
00:52:04
Speaker
Oh my gosh, it's just sometimes things that slip my mind. The other day I was just singing some songs. I think ah going back to like, you know, the 70s, 80s, like America, that band, I actually like those songs too. Yeah, yeah, yeah. I like Ventura Highway and things like, I'm like old. So, but anyway, I like that kind of rustic but kind of. Yeah. very kind of junior LA or New York. There's an easy one for you. New York. Good choice. i think we were wrapped up we've got enough content here yeah and we have plenty of know the audience like yeah i like yeah you guys like ri as they say So I'd like to thank our studio audience.
00:52:54
Speaker
Our full production team that was here. ah they're family lugger There's one left. are Our lovely guest, Mary. Thank you. My beautiful and now dry co-host, James Carmen.
00:53:09
Speaker
It is good to be together in the same room to do the podcast. It's been fun. Good times, good times. All right, so this has been the forward slash where we lean into the future of IT. Subscribe for your latest podcast episodes from us and keep on listening.