The Impact of AI on Cloud-Native Engineering | Cloud Native Now 2023

Hey, good morning everybody. Really excited to be here to open the day with our keynote panel of experts, uh, really awesome people from the, uh, ecosystem. Uh, some of the smartest folks that I've had the pleasure and honor of, uh, really working with, uh, in the cloud native ecosystem.

Uh, today we're gonna talk about the very hyped up, uh, impact of AI on cloud native engineering. Um, but before we get into all of that, I wanna just introduce briefly the folks that you will, we will be talking to today, because they are awesome and you should know who they are. Um, so Iran, Bebe, c p o, and co-founder at Firefly.

Good morning. Good morning, Sharon. So happy to be here.

Uh, I, yes, we're excited to have you a vp, strategic insights and Analytics at Slim ai. Hello. Hey, Uh, Ron Arani, c e o and co-founder at Fox.

So everybody here. Welcome folks. Really exciting to have you all here.

Um, so yeah, we've all heard the phrase, it takes 10 years to build an overnight success, and nobody really knows who said that. Some say Jeff Bezos, some say Tom Clancy. But the fact is that plenty of research and development goes into building technology for it to become a rapid and wild success.

And we're lucky to actually be in an era now where we're witnessing a breakthrough of decades of data engineering and modeling, machine learning, artificial intelligence. It's culminating in this breadth of knowledge and capabilities we probably haven't seen since when Google made the internet accessible to all of us. AI and specifically generative AI will certainly impact many domains.

It's already impacting them. And this panel of experts today, we're going to with them together, we're gonna unpack how we foresee this, uh, really, uh, influencing the cloud native domain engineering as a whole also. Um, so there's a lot in store for, uh, for us.

It's gonna be a really good one. Um, but yeah, without further ado, I love when folks actually introduce themselves and, uh, emphasize the things that they're passionate about and interested in. So, um, let's get started.

Uh, so go ahead and introduce yourselves each, each one of you. I'm gonna start off with Aisha. Tell us a little bit about yourself, your journey to date, and what excites you and excites you about ai.

Sure. I love the energy, so thank you so much for that intro. ai.

We are an end-to-end software supply chain security platform. And, uh, last 10, 15 years of my career has been focused on the intersection of AI and data analytics and cybersecurity. Um, I'm very familiar with technologies, with the hockey stick growth curve.

In fact, my, um, thesis at m i t was focused on emerging technologies and technological forecasting. Um, I'm also very familiar with, uh, ai, not just through the quantitative lens. I am, um, part of the AI ethics advisory board, uh, at Northeastern University, um, and the last couple weeks, months.

Um, although I'm very familiar, uh, with the technology, I feel like we've been on a high speed rail. Uh, every time we are trying to take a snapshot of the scenery, um, the, the time. By the time you look at what we captured, the environment has changed drastically.

Uh, so it's a very exciting time to be around, to be a human. And, um, I'm, I look forward to, uh, discussing, uh, decomposing, uh, what the feature has in it for us. Amazing.

Ron, go ahead and tell us a little bit about yourself and your journey. Awesome. Hey everybody.

I'm Ron. Uh, background, started off as a software engineer in Israeli intelligence units, and then, uh, went into kind of the whole developer experience space as kind of the direct passion and area that I've been tackling for, for a long, long time. Uh, there, prior to what I'm doing today, I was, uh, leading the developer products and experiences space at Facebook and, and now working on flocks and also, uh, board member at the Nixa West Foundation.

Um, small just intro to Nix, just so people are aware of it. It's a open source build system. It's a lot of different things, configuration management mechanism for deploying software, but it's really based on the huge and powerful ecosystem that includes nix packages, uh, which is the largest and most up to today's, uh, repository of, uh, packages in the world today.

Um, and today working on flock, which is working to bring all of that into the work ecosystem and into the world. Um, touch points of ai pretty much even before, obviously it was called ai. So machine learning.

And before that, just like making recommendations based on stuff that groups of engineers thought that we needed to make recommendations on. So across the developer experience journey, kind of seeing how we can implement these things that empower and enhance our ability to have more productive developers. And excited to talk more about that.

Amazing. And Iran, take it away. Hi guys.

My name is Iran Vivi. I'm the co-founder of Firefly. I've been, uh, in the industry for the past 20 years, mainly, uh, in system engineering and infrastructure.

Uh, in the last 10 years, I'm doing, uh, DevOps, uh, uh, technical roles, and then, uh, moving to more leadership, uh, positions. And two years ago, uh, I founded, uh, Firefly together with my co-founder, uh, two help teams, uh, to get better control over the cloud. And to be fully honest, I, uh, was not involved in AI before what happened, uh, six months back when, uh, open air released G pt.

But instantly, uh, I got that feeling that, uh, something is very big, is coming, and as one that is managing the product for Firefly, I, I saw immediate, uh, kind of, uh, conjunction between the technology and the stuff that we are offering, uh, in Firefly. So, um, we embedded a lot of AI flows recently. Amazing.

And I realize I didn't actually introduce myself, so I'll do that really quickly now. Sharon Zisman, uh, I lead the cloud native and open source telaviv, uh, community as well as the DevOps community in Telaviv. Uh, but I also have, uh, my own company, R T F M, please, you should always read the manual.

Um, so, and I've been in this space for some time, and for me it's, uh, really exciting to see everything that's happening. So I'm so happy to have all of these awesome folks with me. Um, so yeah, let's get started.

I mean, I feel like the first place I wanna start, also for so many of the things that you said, Aisha is with you. Um, just a really quick question. Like obviously you're at a company that has AI in its name, and so I feel like you guys have been a little bit ahead of the curve, and since the revolution that's coming, I'd love to hear your kind of take on everything that you've seen kind of to date, um, and all the things that you, where you think it's all going.

ai, we've been thinking a lot about the, the junction, the meeting point of artificial intelligence and, and security and its implications for, uh, container tech. Um, but in, in general with AI especially, um, as Iran mentioned, the last, uh, few months, the rate of division of, uh, AI has been breathtaking. Sheron.

Uh, in fact, I was, um, just two weeks ago I was at M I t, um, at, uh, I AI conference, uh, which was celebrating the 16th birthday, birthday of C sales. C sales stands for computer Science and Artificial Intelligence lab at M I t, and it's the 16th birthday. Um, and multiple touring award winners found fathers of ai, Marvin Minsky's, uh, Rodney Brooks of the, the universe.

Um, it was one of the, um, most ambitious and, and definitely, um, most, uh, content rich AI conferences that I have ever attended. Uh, this is a lab that, uh, in its mission trying to, um, uh, create a world, uh, of computing that enhances all human experience. And I have listened to, uh, talks from 60 labs, from data-driven health tech to, uh, how to speak to proteins, uh, AI for genome, uh, editing medicine, uh, more e equitable, uh, uh, research, uh, robotic learning, uh, exploring the potential of gen AI for, um, education to k to 12 education, et cetera.

It's just like, it's simply breathtaking. I I'm really excited on the flip side, uh, the set time where, um, it is really, um, difficult to wrap our heads around how, uh, much information that we need to process and remember and work that on a given day, right? We are all trying to lead very, um, modern, complex lives, lives.

And it is very hard to make sense of all the information our brains have not si changed, significantly changed, biologically changed in the last, uh, thousands of years, maybe 20 thousands, uh, thousand years. And, uh, it is a time where we need to be very productive and do creative work, uh, with this information abundance and, and information overload. So what my take of all of this is that it is a time that's very exciting, high stakes, and, uh, there's a lot to be gained.

I'm an optimist, uh, a very cautious optimist. I'm not going to dismiss the moral implications, uh, of, of this, but I'm very excited, uh, as a person, uh, as a parent, um, as a, a, a, a friend, uh, you know, a human being. I think this could be a time where, um, AI can definitely change the dynamic for everything that we care about.

But we need to be very intentional, intentional, very thoughtful about our next steps, because as I said, uh, it's going to be, uh, a, a time where it takes our very high and, um, we cannot go into this without planning it, uh, properly. Sure. It's interesting cause you mentioned m mit, and first of all, I said decades in the, in the opening, but I wasn't even aware that it was like six decades of like, you know, uh, research that really has been powering this, uh, revolution.

Um, but a but m i t was known to be the first that that built the moral machine and, and, and gathered a lot of data about ethical decisions in, in AI and all of that. So yeah, they've been doing some tremendous, tremendous work. Uh, and there's the benefits, of course, like you said, but there's also flip sides that we need to be aware of and kind of also planned for.

Those sorts of contingency have contingency plans around that. Um, so thank you for that, Isha. That was really, um, interesting.

Uh, Iran, you come from deep in the trenches of DevOps, um, and I feel like AI for, I have at least because maybe this is, uh, my own, uh, echo chamber, but I've seen, um, AI being applied very liberally across the DevOps space right now, cloud native space. Uh, obviously the, the conversation that we're having is around cloud native engineering, and actually Firefly was one of the first to launch, uh, you know, really interesting tools in that space. Open source tools like AAC and, uh, policy yak, or I don't even remember what it's called, or policy ai, something like that.

Um, but tell me a little bit about like kind of your experience, like from a early stage DevOps engineer till today, a seasoned kind of DevOps veteran, some of the things that you wish you would've had then maybe, um, that AI is powering now, and what you think is coming next kind of in the DevOps and cloud native space. So for anyone, uh, that familiar with DevOps is, uh, basically a combination of a, a lot of, uh, kind of skill sets that, uh, will allow the development team to work faster, uh, into promoting stuff to production. So it's a lot about, um, docking tapes and integration between systems and also, uh, a little bit about firefighting.

Uh, if you are more close to that operat. Yeah, that's exactly what, um, if you are more close to the operational and like the SRE kind of roles and, um, what we had in mind when we see that AI capability, uh, introduced six months back, uh, we thought about the use cases to help DevOps doing better jobs. And one of the main use cases was about generating code and Firefly.

One of the main, uh, capability of the product is to generate I a c infrastructure infrastructure as code. So it was totally makes sense for us to take that capability of those models that can generate codes, uh, like G P T models and Codex and other, that, uh, doing a very good job with generating code and giving a flow to generate iacs. And IAC is not just about generating infrastructure for your cloud, it can be also generating, uh, your pipelines.

So it's also major capability for a DevOps to generate the pipelines. It can be like GitHub, Jenkins, everything that is described as code. And you also, man, uh, mentioned the policy as code.

So we introduce the policies code by ai, so now everybody can, with the natural language to type the desired policy, and the AI engine will give that policy as code, uh, in response. So, um, for us, it's very fascinating to take that advantage of a machine that is generating code and embed it into the DevOps day-to-day flows. Wow.

I don't love, well, you don't love the yaml. Go ahead Ron. No, um, uh, I also think I want to add on top of what Iran is saying, I just fully agree.

I think that the back kind of tying that into where I also see it kind of coming in and starting to make some magic happen, and obviously I think Iran has like a lot of that experience. It's probably a lot also on the automation side, right? Um, Iran, please, please throw something at me if I'm way off here, but for me, a lot of the DevOps processes when I was working on it, you know, 10 years ago was like, we needed to get things to be more automatic than manual.

Um, and I think what's exciting about AI is if there's so many people doing so many of these, you know, converting so many of these standard manual processes into the automated parts of them, ideally, if we were able to tap into that knowledge base and then start reproducing that fully autonomously, I think that's a lot of the magic that can start coming into DevOps, right? Um, I might touch on this later, but there's Well, platform ai. Exactly.

I have A, so yeah, you need to go into that website. Will, uh, will AI be taking my job, right? So everything, eventually all of us are gonna be working for the machines.

Um, that's, that's interesting. It's true. I think that eventually when we get to that, that place of automation and, and I mean the, the sky's the limit.

It's really interesting. What's, what's happening. It's fascinating.

And maybe we could have spent less time being on call and there could be just, uh, you know, AI on call that just, uh, yeah. Machines that are, uh, a machine that knows how to roll out the duct tape. It's like, it's really easy, Right?

But if you take into the, you know, what's happening right now, you see those vendors, uh, they're doing, um, the monitoring like New Relic and Datadog and, uh, PagerDuty, they are embedding AI flows, uh, to tackle that exact, uh, kind of challenge of how can I reduce that friction of somebody getting on call? Maybe I can have an AI to investigate it first and determine whether I need to wake up someone or solve it by myself. So we totally going to this direction.

I was, I was just gonna, I was just gonna add on that, cuz I'm starting to riff. So Cheryl, please do stop us. No, go ahead.

But like the, the empowering decision making, right? Where you a lot of times have like the DevOps engineer or the DevOps team, and we've had two kind of trends in the ecosystem where the concept of DevOps started to be in order to optimize engineering and make sure that production kind of gets out the door better and more informed. But then there started to be, because we took DevOps out of the standard engineering process, there started to be a chasm.

Uh, and then we did a lot of work to try and bring DevOps back into engineering with all of these shift left concepts. What would be very cool, right, is if like, we have the DevOps engineer and the ai, is that counterparty helping make those decisions? Uh, should we do this or that?

Or what did the engineer even mean? There's no documentation to it. Like there's, I think there's a lot of cool areas there.

Yeah, absolutely. Um, actually, uh, one of the things I would love for you also to continue touching on, I know Ron, you were just, uh, speaking, but um, a little bit on the world of open source. Like one of the things that suddenly came to mind is like, you know, one of the things that we've experienced many times, you know, as open source maintainers and, and people that have worked in the open source, uh, ecosystem is, is like when you have wide scale and you know, very wide adoption, like you noted brewer nicks and that, it's like one of the largest projects in the world is kind of like the, the onslaught, the like, kind of the flood of like issues and, and poll requests and things that you need to manage D C A I coming into play in that kind of a world, like to help kind of make a method to the madness, maybe kind of close prs that aren't relevant or escalate issues that repeat themselves or I don't know, things like that.

Or maybe even adding more to the problem first before remediating or relieving the pain. Yeah, I mean, I would say, I would say, um, so just going back, I think we tackling the open source question, I'm, I'm extremely biased. Um, I've been doing open source for a long time.

I think, honestly, while we say AI is the future, I think that that's partially correct. I think open source is the future. I think that's how we innovate as a humanity, kind of like single resource.

All of us trying to get behind these things that can be groundbreaking for, uh, for what we're trying to do and what we're trying to develop and kind of set ourselves all on a higher threshold. Um, obvious plug for Nick's, it's like, this is how Nick's is kind of operating in the ecosystem. Why is it the largest package repository in the world?

Well, because it's open source. Cause we have a huge community, lots of humans coming together and doing this. Um, I think the o the impacts of open source are two-sided, I think, uh, sorry, of AI on open source.

There's two sides to it. On one side, I'm rather hopeful to see how AI can further force multiply the whole concept of open source, right? Um, it's like Aisha, Iran have like this really cool idea and they wanna make this open to the, to the general public so that everyone can start building up on top of this.

It would be really awesome if AI can help them get there, right? Remove all the menial tasks, get that started, and then help them force multiply their efforts going down the line. Um, so I think a lot of the things you mentioned are around the fact that open source is exactly that.

It's open. So technically AI has access to everything there. Um, and it can be everything from reviewing code, right?

So empowering contributors I think is a huge area for AI to come into. Um, and then also helping new people come on board. There's many areas, I think those are the two areas I'm the, I'm most keen on.

Uh, because I think those are the things that at the end of the day, sustain that flywheel, that open source relies on right contributors and empowering them and then bringing new people in so they can use what we've built and leverage it and have it impact them, but then also having these new folks become contributors. Uh, yeah, That's, I i I I would like to share my perspective on how AI and open source working together. So I think right now it's pretty clear that open source as a player in the AI is, is much bigger player than the commercial product.

So you, you can, uh, talk about like, uh, the LLMs that are open source are evolving much faster than the stuff that open AI introduced and the stuff that Google, uh, uh, introduced and also the tooling that l allowing us as users to gain. Um, the value from AI is much broader in the open source tool. So you can see, you know, it, it's going very fast.

So one of the recent project that I see, I think two weeks ago was the, um, engineer G P T or something like that, or developer, G P T, I'm not sure which one of them, but it's actually amazing. You, you can build an entire project and push it into GitHub with only a set of, um, instructions that the user is giving. So it's not just an AI model that can generate code, it can do the end, end, End declarative model.

Yes. It's like, think about it, the entire project, all of the structures, all the models, he asking you question, you are, uh, providing back the answers, like is basically interview you about a project that you like to build, and, uh, G P T engineer or whatever the name is, it's a totally open source project that delivering a full functioning project, the development project for you. So I think open source eventually will be the place to get the value of AI, either in the tools or the models itself.

Okay. I wanted to come to this a little bit later, but since we're already on the topic of open source, I would love to tap into your thoughts on, so, you know, like I, I think that, uh, it's well known that even GitHub was surprised by, uh, by the success of co-pilot, but there was a lot of controversy when co-pilot was first released and now also G P T around kind of copyright and, and rights and open source code and, and what's legit and what's not legit. And we talked about the moral machine and about the ethics, like what about the creators?

What, what, how do you feel about like kind of how, you know, chat chip PT that eventually is going to be a for-profit company or I believe in open source and I'm, I'm the first believer of open source. I'm an open source advocate. I built the cloud native and open source community in Tel Aviv, but I feel like some of the, um, organizations that exploit kind of the openness of open source can hurt that entire kind of, uh, value system that we believe in.

So I, I shall let you, um, go for it. I should take it away. Sure.

Yes. So I think it's a time where there's going to be a co-pilot for everyone from the financial analyst to the city planner, to the art history, um, student. Uh, we will have an assistant that act as a force multiplier for all of our productivity.

Um, and I can't wait for it. And from a, uh, open source perspective, I have always believed that the future is written in code and it's open source. And I do believe that AI is only going to enhance that further.

That said, I get a lot of questions around, uh, I know the question is about creators, but, uh, let me take a step back and talk about open source for, for just a second here. Uh, the groups of open source at human scale, not, uh, affected by AI generated influx of corn, right? Um, the NPM ecosystem alone, uh, there are more than 30 million, uh, packages.

And, uh, I was looking at some studies that said there's, there are 1 million new packages are being added to the system, to the ecosystem npm ecosystem alone every month right now. So that's the pace. And, um, yeah, we talk a lot about security review of open source code.

Um, and I know there are differing opinions on this, but even if we took the top, uh, 10,000 packages out there in the NPM ecosystem alone, um, there was some calculations on napkins, uh, you know, on the back of a napkin saying that we will need a thousand researchers working on this for 3000 years to be able to understand what's going on underneath. Now with ai, um, you can already say that there'll be more code than ever before. It is, uh, you know, we are witnessing a surge in the volume of code being produced, and, uh, it's not just humans anymore with AI agents.

And it is a double-edged sword, right? You know, there's just opening up new, uh, possibilities, but new, uh, security risks for enterprises. At R rsa.

A couple, uh, months ago, I was a speaker and I realized that the enterprise security teams, uh, teams are not, uh, thrilled about the possibility that anybody who can prompt, can generate code. They probably do not like their developers to generate that much code anyways. So there's that, that, that piece there.

Uh, but I'm also seeing, um, seeing, uh, cybersecurity, uh, ally, uh, you know, type of work for AI too. So as AI grows more sophisticated, um, a lot of security researchers are taking advantage of ai, uh, finding the proverbial needle in the haystack fast, faster, um, bad actors doing the same, weaponizing AI writing more malicious code faster. So all of these like system dynamics of the issue, uh, at the core here, but I don't think that if we want to get above and beyond what we have generated so far, we do need to help.

It's already, like, you know, I have done a ton of research on container security alone. Uh, you are familiar with our reports, uh, Sharon, the container security landscape reports. It's already obvious that we are never ahead of the curve, like month over month.

We are looking into how the security dynamics are changing. And for every vulnerability that we remediate for new cvs are being added to the system. The only packages that has the most popular packages has the most number of CVEs because there are more security resources looking at them.

So the humans are the bottleneck in our systems right now. So AI is definitely going to help with that, um, before maybe creating some issues and we understand how to control it more and better for our purposes and for creators. I think it is going to be a force multiplier.

As I said at the very beginning, everybody is going to benefit from this. Um, I'd like the, uh, the coach, um, I don't know the, the, the, the person who said it, maybe Jeff Bezos again, Jerome, but I don't think AI is going to take away jobs. Uh, it's, uh, going to add a lot of new jobs into the ecosystem.

Uh, but a, a person who is, uh, going to use AI might be able to take your job from you. So for, for everybody, I think we need to be understanding the technology as much as possible. Um, empowering ourselves, taking advantage of it, as opposed to being overwhelmed with the, uh, the, the information load.

Um, we can turn into, um, these, uh, these, uh, humanoids where we have multiple security resource superhuman data analysts as our, at our disposal, uh, conversing with multiple AI agents, uh, in our jobs every day. Yeah, I agree. I mean, I, I'm sure that the value is gonna be substantially larger, but I do have that, you know, kind of still that sentiment for the folks that are powering AI and, and prompting it and giving it all the information and putting the information openly out there for it to leverage in order to make us all, you know, super humans as you, as you say.

Um, I, I'm not sure you have the STA points, Sharon, uh, for Aisha, um, perspective, it's like, it's really fascinating for me to think about the advantages of AI helping, uh, against cyber crimes versus, uh, uh, cyber criminals will take advantage over, um, the stuff that j AI can help with, like, you know, finding exploit and generating code to exploit them. So we will see more of that cat and mouse kind of fight with, with now a new weapon in the arsenal. Right, Exactly.

Actually what's been happening to date, but like, yeah, exactly. Sure. I did, I did wanna quickly just go back to the, to the open source question cuz I think it is a really critical one because I think it's just such a, it, it's a topic right now that open source has been such an explosive and impactful thing inside of the world because of the fact that it was open.

And I think there's a lot of concerns on the open source side to that question, right? Should I stay open? What does that mean for me?

Will these huge and like enormous companies just make, you know, make me relevant and take all and kind of steal, right? Everything that I've put up there. Um, and on the other side, right?

Like, we do want AI to succeed and help us keep doing great things. So there's just that topic on that contention I think is a really important one. Not that I think that we have any clear answers to it, but just quick 2 cents on that.

I think one open source kind of has this concept of licensing and I think that is something that is kind of situated well to help with this, uh, right? Because the core of what makes open source exist is the contributors. Mm-hmm.

Without them, we don't have open source without them. We don't have a s**t ton of things that exist today in the, in the engineering space. Um, and we really need to empower them and make the contributors be able to keep contributing.

So I think that when we're gonna start seeing just more seriousness around, okay, what does a license mean? What does it imply? And I think we're gonna see an evolution of licenses in open source move to take into account the fact that there are giant beasts of AI scraping and looking through everything, and we need to find the symbiotic relationship with it, right?

Because open source is about empowering others and AI kind of helps that, but we don't want to get it to the point where it's taking advantage of others. And I think the other topic that is an interesting one is that the line between learning something and copying something has always been fuzzy, but AI just makes it even fuzzier since we don't really have that full understanding of what that means, right? So if like I take, uh, if I take a book that Sharon wrote or, um, code that Iran wrote on DevOps a long time ago and it inspired me to write something on top of it, maybe you can call that learning and not copying.

But when AI is doing it, is it copying? Is it learning? Right?

Like there's, there's still things that we have to figure out. E Even with the example of gener generating art, when you are asking Dali or me journey to generate some, uh, picture with the style of, I don't know, Leonardo da Vinci for example, you are basically stealing the, the style from that artist. I saw a whole Twitter thread on that.

It's true. It's, Yeah, I guess we will see a lot of, uh, legal aspects and, um, compliance that related to AI that will be formed in the up upcoming years. The tool will really set the boundaries of what AI can do and what is illegal or stealing property, uh, considered to be stealing in a property of someone.

Yeah, I mean, I've spoken about this historically in other contexts of like kind of the big corpse like wrapping managed services around open source pro products and like kind of selling them. Uh, and, and really, um, and this brought on a whole kind of trend of like locking down licenses and things like that. Uh, the next bigger, more massive step would be these big corpse that run AI that would do that at a much greater scale.

And it, it has yet to be seen how this will be handled. But I, I do, I do feel a little bit of round sentiment there of where this could also be, it's a double-edged sword. It's scary and we need to see how like we manage it.

We're coming to the end of the, this was a really fun conversation to went by really, really fast. But there is one last thing I would wanna tap into your excellent minds here on, um, which is a little bit on the quality side. So we've seen it really, really rapidly adopted things being used and code being generated, like you said, tons of code and tons of things.

Um, and at least on the cloud native engineering side, which is extremely complex. I mean, in the 10 years that like, kind of, you know, we've become a lot more, um, seasoned and in, in cloud operations, the complexity and scale has grown immensely. Um, and while I think that AI is learning and is very capable, even today I saw that it made a mistake in something that I, um, that I prompted it, it like, it was completely wrong about like kind of the essence of, uh, of the issue.

And I'm wondering how much junk is going to be, uh, found in our code, um, how this is gonna impact kind of future code, what, how like these abstraction layers are gonna impact future engineers. But a lot of stuff that like kind of, will we lose the foundations? Will we even know how to run these machines once all we do is, uh, is create like kind of machine generated code.

Um, and I'd love to start with you Iran, cause this is actually your sweet spot, but, uh, I'd love to hear what you're, what you think, Again, this is forecasting what I think it, we will see, uh, better models, we'll be introduced with more accuracy and more, uh, context window, meaning they can, uh, output and get input of more data and mm-hmm. So the, the output will be much, uh, better. And we also, so going to see that skill of prompt engineering will be one of the main skill of any employee.

Um, and if you get something wrong, uh, from chatting with G P T, you might be getting, um, the answer you are, you would like to with a different kind of pro. So we will improve as an individual with that skill of pro prompt engineering to get better results from the ai. And in parallel, we will see the models will be improved over time.

This is totally makes sense. So you have no fear, you're not afraid that there's gonna be tons of junk and unusable code and things that are gonna make our systems explode in, in a few years, like race conditions and bugs And, uh, I think human errors are more likely to be than AI errors over time. I guess right now a human being can write better code than ai, but it's a temporary state eventually, and eventually it's not.

10 years from now, it's maybe a few months from now AI will do better job in writing code. Uh, so I'm totally not afraid. I'm more afraid about, you know, human beings will continue to write buggy code, right?

AI going to help us with that. I'll keep learning that buggy code you're saying, uh, actually I'd love to take it to the security side. Tell me a little bit about how you kind of envision this on the cloud native security side or, and, and things in that area And maybe data science too.

Um, because here's what I think, um, it creates this, um, confidence and sometimes a false sense of confidence in people that I have seen. Because here, here's the thing, and I'm, I'm totally for it, that the learning curve has changed drastically. I think Sharon, um, I, to give you a very concrete example, what we have done at Slim a couple months ago is we did an AI hackathon and I took a group of people, a, a lead software developer, a brand designer, a social media manager, and yours truly a data scientist.

Uh, we came together and we said, let's do a team that does AI role reversals. Basically I became the social media manager and we spent and our, I'm in the driver's seat and our brand designer became our data scientist. I gave her some data and we literally wash her.

And I provided some data, uh, tools, uh, including certain LLMs where she was basically querying the dataset using plain English. And we have done an end-to-end container security analysis, um, in a Python, in a Jupyter Notebook, basically just asking questions and sometimes asking questions as vague as what could I be missing in this data set? Can you help me create these visualizations?

Maybe a force directed network of all these vulnerabilities as tied to packages, right? Some interesting questions, some vague questions. But the end result was, um, literally in an hour.

I can tell you that it has done a week worth of a junior data scientist's analysis for me now because I know the dataset really well and I know the right tools. I was sometimes asking the questions and I sometimes seeing errors in the code that I was stepping in and changing those. But that took that, you know, brand designer and social media manager a couple minutes of just asking the right questions, visualizations, just like they were getting into the conclusions with the dataset really quickly.

Then our software, um, developers started doing these designs on our code, the, uh, logo using mid journey. And it was really fun. Wow, we couldn't even stop that for a while.

And with social media, I took our best performing content in social and um, literally created these storms using a couple of different tools, different LM models, and it was creating some really cool results that said, this is just a word like NR word of, uh, work, right? Then you try to take it to the next level. You start, you might start feeling that you've got this, now you can do the machine learning model.

K means clustering principle, uh, component analysis, isolated for sort of a, a value may not really understand what's going wrong, which happens very often, then you need that expertise to step in. But as Iran said, this is just a toddler technology. I know that it is going to get a lot better.

Does that mean that the experts need to take a step back? I dotted I think we can use our resources, our, um, our mind power on more complex stuff. Uh, the thing that I worries me is usually, uh, the people who are relatively new who with that, uh, full sense of confidence might go on creating more complexity, adding more complexity, not being aware of the, uh, the disadvantages, the, uh, downsides of the technology, again, which might be sold by, uh, better recursive self improvement of these technologies that we are dealing with.

All right, last thoughts from you Ron. Cause we're almost at time. Just, uh, gimme kind of your take on like, uh, all the things that I asked quality, uh, things that We asked, but a few things there.

And uh, I'll, I'll try and keep it short cuz I'm seeing this at flocks as well, right? Where we're trying to, not trying, we're leveraging AI both in order to enhance and empower our development, our velocity, and across many different teams. I think Aisha pointed on it, it's like if you're, you're gonna need to use AI in order to be better at your job.

And we're seeing that really optimize different efforts across the flocks and then we're using AI in order to optimize the experience for our customers, right? For our customers and for our users. And that's kind of, I said as an impact to it.

Um, I think a few of the questions you asked, right, was one about the legitimacy, like ai, I saw a thread that was kind of, kind of funny about AI inbreeding where AI was using AI images to create AI images and I was like, s**t, um, if I have AI doing DevOps practices learning from an AI DevOps practice, then we're gonna get Skynet in like five years with some backdoor to, to some concerning places. Um, I think, so to just keep it short, I think that AI is going to fall practically into two spaces things AI can do for me. So I touched on shift left earlier, can we shift all the way left with all the way left being the, the machine?

Um, and therefore can we shift things that we have full trust and reliability and can AI do the swell, we'll move it there and then again, to earlier points that were made here, we can then go and think about the more complex things, the innovation and all of that. And where AI comes in, like getup, copilot as like your buddy, right? Where it's like you still fact check it, uh, but it's there.

It's helping you, it's optimizing and enhancing you, not necessarily taking something away from, from your table. Um, but yeah, I think I, about every six hours I flipped from being an optimist to a pessimist on ai. So depending on when you catch your hair, This was such an incredibly interesting panel.

I am sorry that it's over and there was so much more I wanted to ask. But, uh, really some really great insights here. I think that that does, it does have to be seen what's coming and what's gonna happen in the cloud native space.

And I agree with you a hundred percent. Uh, sometimes I'm an optimist, sometimes I'm a pessimist, but there's no doubt that my mind is completely blown. And the short time that I have like really witnessed what it's capable of and some of the things that Iisha has shown me that she's done in her honor is on.

You guys have done. So thank you all for being here and sharing your thoughts and, uh, let's see what happens, uh, if the machines are gonna be taken over. Thank you everybody.

Thanks Rome. Thank you.