Welcome to a special edition of The Cloud Collective Podcast, where we explore the transformative world of artificial intelligence (AI) and focus on practical strategies for success.
In this episode, host John Kundtz is joined by Kyndryl Vice President and Distinguished Engineer Greg Pruett, a multi-patent holder and expert in infrastructure and cloud architecture.
Greg begins by clarifying the differences between traditional predictive AI and the emerging field of generative AI, which has the unique ability to create new content.
He then outlines the five crucial steps for getting started with AI:
Analyze Business Needs and Desired Outcomes: Conduct a structured assessment to identify potential use cases and quantify the return on investment from AI adoption.
Define Your Data Strategy: AI systems are only as good as the data they're trained on. Greg emphasizes the importance of data quality, inclusivity, and representativeness and addresses privacy, security, and bias concerns.
Choose an AI Software Platform: Explore the growing ecosystem of AI platforms that offer pre-trained models and accelerate adoption by enabling organizations to use rather than invent AI solutions.
Select AI Infrastructure: Decide whether to run AI workloads on public cloud, private cloud, or a hybrid model, considering factors like data sovereignty and compliance requirements.
Establish ModelOps Processes: Implement robust processes for monitoring, updating, and testing AI models to ensure responsible deployment and maintain high-quality outputs over time.
The show wraps up with insights into overcoming common challenges in AI projects, ensuring data quality, and operationalizing AI within your organization.
Want to dive deeper into the world of AI and explore how it can revolutionize your business?
Don’t forget to subscribe to this podcast on your favorite podcast
***
Engage, Share, and Connect!
Spread the Word: Valuable insights are best when shared. Share this episode with peers who may benefit from it if you find it insightful.
Your Feedback Matters: How did this episode resonate with you? Share your thoughts, insights, or questions. Your engagement enriches our community.
Collaborate with The Disruptor and connect with John Kundtz.
Quick Connect Call:Dive deeper into the discussion. Book a 15-minute chat with John Kundtz -> Schedule here. Stay Updated: Don't miss out on further insights. Subscribe to our YouTube Channel and our Blog
Welcome to a special edition of The Cloud Collective Podcast, where we explore the transformative world of artificial intelligence (AI) and focus on practical strategies for success.
In this episode, host John Kundtz is joined by Kyndryl Vice President and Distinguished Engineer Greg Pruett, a multi-patent holder and expert in infrastructure and cloud architecture.
Greg begins by clarifying the differences between traditional predictive AI and the emerging field of generative AI, which has the unique ability to create new content.
He then outlines the five crucial steps for getting started with AI:
Analyze Business Needs and Desired Outcomes: Conduct a structured assessment to identify potential use cases and quantify the return on investment from AI adoption.
Define Your Data Strategy: AI systems are only as good as the data they're trained on. Greg emphasizes the importance of data quality, inclusivity, and representativeness and addresses privacy, security, and bias concerns.
Choose an AI Software Platform: Explore the growing ecosystem of AI platforms that offer pre-trained models and accelerate adoption by enabling organizations to use rather than invent AI solutions.
Select AI Infrastructure: Decide whether to run AI workloads on public cloud, private cloud, or a hybrid model, considering factors like data sovereignty and compliance requirements.
Establish ModelOps Processes: Implement robust processes for monitoring, updating, and testing AI models to ensure responsible deployment and maintain high-quality outputs over time.
The show wraps up with insights into overcoming common challenges in AI projects, ensuring data quality, and operationalizing AI within your organization.
Want to dive deeper into the world of AI and explore how it can revolutionize your business?
Don’t forget to subscribe to this podcast on your favorite podcast
***
Engage, Share, and Connect!
Spread the Word: Valuable insights are best when shared. Share this episode with peers who may benefit from it if you find it insightful.
Your Feedback Matters: How did this episode resonate with you? Share your thoughts, insights, or questions. Your engagement enriches our community.
Collaborate with The Disruptor and connect with John Kundtz.
Quick Connect Call:Dive deeper into the discussion. Book a 15-minute chat with John Kundtz -> Schedule here. Stay Updated: Don't miss out on further insights. Subscribe to our YouTube Channel and our Blog
AI Insights with Greg Pruitt a five-step blueprint for success. Hi everybody, my name is John Kuntz and welcome to this special edition of the Cloud Collective podcast. In today's episode, I am excited to welcome back Kindrel Vice President, distinguished engineer, multiple patent holder and infrastructure cloud architect, greg Pruitt, as he shares his valuable advice on how to get started with artificial intelligence. If you are looking for help to frame your thinking around AI in your AI projects, then you have come to the right place. I want to welcome back Greg Pruitt. Hi, john, thanks for having me back on the show. Hey, it's great to see you again. For those that weren't around or would like a refresh, why don't you share a little bit about your background with our listeners and just feel free to start anywhere you want?
Greg Pruett:
Thanks, john, I appreciate it. My background is actually primarily in system design. I spent many years designing systems for IBM and Lenovo, so I have a deep engineering background as well as a software engineering background. More recently, I've been part of Kindrel working on services and helping our customers to adopt new technology, so I'm part of the CTO office here in Kindrel and recently doing some very exciting work with artificial intelligence.
John Kundtz:
That's spectacular. I'll tell you what. Every time I get to talk to any of our distinguished engineers, it's always a fascinating topic and always learn a lot. You can't go anywhere today without having something in the news about artificial intelligence, or AI. There's a ton of hype around it, particularly around gen AI or generative AI or chat GPT, but AI has been around a pretty long time and before this hype around generative AI came along, there was predictive AI, which is sometimes referred to as traditional AI. I wonder if you could just spend a moment, greg, explaining the differences and maybe give some examples of each.
Greg Pruett:
That's a great point, John. Artificial intelligence is around in a lot of different forms, and generative AI is just one field of artificial intelligence. But there are numerous other places where artificial intelligence has made an impact on our lives. You think about natural language processing, Think about your home devices, your Google, your Alexa, your Siri. Those are devices that can understand speech and respond to speech requests, typically using something like a recurrent neural network in R&M. Also tremendous advances in the area of computer vision over the years typically convolutional neural network, CNNs that are very good at recognizing objects or being able to recognize objects in video, used in transportation, for self-driving vehicles, used in retail for self-checkout or loss prevention. A lot of these, what you call more traditional AI technologies, are also maturing and becoming much more widely used in production environments. It's not just about looking at generative AI. We're helping our clients with all types of AI. Now, generative AI, like you were saying, is quite different. While it's based on pre-trained data, generative AI actually has this unique ability to generate or create new content. It provides some I'll say randomization to the pre-trained data and allows it to create summaries of documents, write Q&As based on a large volume of data that actually create blogs or even presentations. Now we're seeing some exciting new areas of generative AI creating pictures and even generating videos.
John Kundtz:
Greg, many companies are reading about AI, especially large enterprises. They're trying to figure out how to leverage some of these technologies to innovate, to improve productivity, to help their decision-making, gain competitive advantage, reduce time to market, stuff like that. But I think a lot of new technologies people struggle on how do you get started?
Greg Pruett:
With AI. I think there's multiple steps to it. Maybe I can organize that answer into a few steps. First, like you were saying, there are lots of ways that you could apply AI, but we really encourage our customers to do a very structured assessment so that they can look at different business needs, different desired outcomes, and go through and score what kind of return on investment they could get out of these different scenarios. I really think it starts with business needs and how you can achieve business differentiation. Then, I would say, probably the second most important thing is thinking about true data and your data strategy. Ai can do amazing things in terms of consolidating data or producing insights out of data or making predictions from data, but the whole system is only as good as the data. So most of us need to spend considerable time thinking about data science and thinking about what kind of data quality, data cleanup, data reliability needs to be implemented, as the data sources are certainly very important, but along with that and last time, of course, john, we talked about sovereign cloud Another very important thing that we're seeing more and more in the news these days is, you know, making sure you deal with anonymizing data, keeping data private and ensuring fairness and avoiding bias and all of those things.
John Kundtz:
So there's a lot of aspects to your data strategy in terms of making sure it's a good data set, it's an inclusive data set, it's a representative data set, and so that's another very important aspect of, I would say, preparing and that's probably a key step, right, if you think about it, if you think about these large language models, and you don't want to train them on somebody else's data, and you certainly don't want to expose your data to somebody outside of your organization, particularly if you're trying to build a competitive advantage or do something innovative. It's the garbage in garbage. Have to sort up your data, your strategy, where it is, who has access to it, who doesn't have access to it, before you can actually, I think, really do something unique within your organization, particularly a large enterprise. Does that make sense?
Greg Pruett:
Yeah and John, I joke with people sometimes that the biggest winners in AI may actually be the lawyers. We're starting to see more and more public lawsuits against AI companies, either for training data that may have included trademark material, or due to bias and fairness or openness. We're seeing various types of lawsuits now, so that data strategy is really key. What are your data sources? What are your privacy controls, your responsibility controls? Cool, the number three. In the first two you notice I didn't really mention any AI terms like creating models or any of that. I was just talking about basics business needs, return on investment and data sources and privacy. I would say after that, you do want to think about what type of AI, like you're asking, is generative AI correct or are there other AI technologies that can help solve that problem? One thing I normally recommend is to think carefully about adopting what I call an AI software platform. There's a lot of very good AI platforms emerging. Of course, there's the Google Vertex platform, the NVIDIA AI Enterprise platform. There's Red Hat OpenShift AI platform. There's a number of very good software platforms that are emerging that can really help you adopt AI more quickly in terms of using it, not inventing it. These AI platforms can help you by providing you pre-trained models. Maybe you don't even need to go through the training step, or maybe you do, but maybe you can start with a pre-trained model that's pretty good and then do some customizing or tuning on that model with your own specific date. Then after that, after you understand and pick an AI software platform, then it's when I would start thinking about infrastructure. Where do you want to run this? Where do you want to run this in terms of a production environment? Do you want to run it in public cloud? Do you have data sovereignty requirements that make you want to run it in private cloud? All of those are possible. You're going to have to make decisions, choosing a software platform and choosing a private or public infrastructure for hosting your production. Ai, I think, or steps three.
John Kundtz:
That makes sense, like she's describing. Your first two steps are basically I would call it basic business strategy blocking and tackling. Now you're ready to put the foundation. Now you're ready to build the walls. You've got to pick your platform and then figure out where you want to store your data and run the software. Now we've got the first four down. I believe there's one more step that you recommend.
Greg Pruett:
So for the fifth step, I think you really have to think about the operational model, the process for running a production AI environment, and maybe this isn't as intuitive Like any other software application. You will need to think about the lifecycle of that application. How do you update the content in the model? Maybe if you have a generative AI model that's trained on all the user manuals for automobiles, what if a new generation of automobiles comes out, a new model years? You need to be able to update the model and add more data. Whenever you update a model, though, there's a number of things that should be done. We talked about, with the data strategy, making sure that your data is clean and accurate and representative, but also your model needs to be tested to see if there's bias, to see if it produces bad results, and so anytime you retrain, you want to retest and look at those responsible AI processes to ensure that what you're putting out is of high quality.
John Kundtz:
Makes sense. If you look at some of the generative AI companies like OpenAI or Claude from Anthropics if you just pick on OpenAI GPT-4, it was around for almost six or eight months before they let it loose into the world. I believe, from what I've read, they spent a lot of time on security, on accuracy, on privacy and bias and all those things you just talked about. Hit the market like with a bang. I think you're right. The operational processes are going to be your guardrails. How is it used? We learned of anything is you can't just take everything right out the gate as gospel truth right, at least from my experience with generative AI. It's another smart person in the room I normally look forward to that fifth step, as is ModelOps or Machine LearningOps, modelops Basically the five steps and, by the way, this is based on your blog posts that you wrote and will obviously include the link in the show notes to it, the summary. You analyze your business needs, your desired outcomes, the requirements. You define your data strategy, pick an AI software platform and then you select how you're going to run the infrastructure public cloud versus private cloud and then you establish your ModelOps processes. That is, I think, excellent and very valuable information. Any concluding thoughts or parting words of wisdom on this topic?
Greg Pruett:
I guess I'll say two things. One I hope I didn't make it sound too daunting. I think there's a fairly simple formula and we have a process that we're doing with our clients that's repeatable and ensuring successful AI projects. The second thing, though, I'll say is that there's some studies out there that say that a large percentage of AI projects right now are failing. A lot of them get to a successful proof of concept, but they fail to actually make it into production or to produce the results that are requiring. The two areas that we see where companies are struggling one is data quality. We harped on that earlier in the discussion. The data is key. It doesn't matter how good your model is if your data is poor. Ai can't make up poor quality data. So that data strategy and data quality is important part of the process, and that's why, whenever we talk to our clients about AI, we're also talking to them about the fundamentals of data science. The other aspect is just the operational nature, and again you may not initially think about that when you're thinking about AI technology, but really designing an operational model for how to get to production and how to ensure a responsible AI systems.
John Kundtz:
Excellent when I read your article, and there are so many other topics and things we just don't have time to cover today which I certainly would love to talk about. I think other people would want to hear about as well. So I would actually love to have you come back on the show and talk more about those at some point in time, if you're open to it. But in wrapping it up, I want to thank you, Greg, for a great interview. One more question, and that's how can people learn more about what we're talking about today and how might they get started in developing their AI strategy?
Greg Pruett:
You gave a plug already for my blog on LinkedIn. I'd love to see feedback from folks on LinkedIn on that, and Kendra published a lot of other blogs and very useful information to folks to help learn about this and learn how we can help.
John Kundtz:
Excellent, so we'll definitely include the link to your blog post on LinkedIn in the show notes. Of course, I recommend people connect directly with you as well. That way, they can also check out your article. Five Steps to Getting Started with AI. So wrap this up, greg. Thanks for joining us. It's been a great discussion and super exciting topic, and I just want to thank you for joining us on this edition of the Cloud Collective Podcast. Have a great day, everybody. Thanks.