AI for the Rest of Us: A High-Level Overview

Kick off Season 3 of Inside MySQL: Sakila Speaks as leFred and Scott welcome Matt Quinn for an engaging introduction to the world of Artificial Intelligence. In this episode, we step back from the database and explore what AI really is, how it’s shaping society and technology, and why it matters to anyone in tech today. Whether you’re just curious about AI or eager to understand its key concepts, join us as we break down the basics and set the stage for a season of discovery.

------------------------------------------------------------

Episode Transcript:

00:00:00:00 - 00:00:31:22
Welcome to Inside MySQL: Sakila Speaks. A podcast dedicated to all things MySQL. We bring you the latest news from the MySQL team, MySQL project updates and insightful interviews with members of the MySQL community. Sit back and enjoy as your hosts bring you the latest updates on your favorite open source database. Let's get started!

00:00:32:00 - 00:00:58:22
Hello and welcome to Sakila Speaks, the podcast dedicated to MySQL. I am leFred and I'm Scott Stroz. Join us today. It's Matt Quinn, vice president and head of AI at Orracle. Matt leads how Oracle Cloud Infrastructure's AI services are adopted by customers in EMEA. Matt brings deep expertise in enterprise software strategy and a passion for making AI both powerful and its adoption practical.

00:00:59:00 - 00:01:21:03
Today he is here to help us unpack what GenAI really means for the organizations we work for and buy from, and what it means for developers, data professionals, and MySQL users everywhere. Matt, welcome to Inside MySQL: Sakila Speaks. It's great to have you with us to kick off season three of our podcast. Thank you very much, Fred,  Scott, great to be with you.

00:01:21:08 - 00:01:43:21
Looking forward to, to an interesting conversation and getting us going for season three. Awesome. Matt, thanks for being here with us. So right off the bat, when most people hear the term AI, they probably think of chat bots. But that's just one form of AI. Can you help provide us with like a high overview of the different types of AI that exist?

00:01:43:23 - 00:02:15:10
Absolutely. And I think AI and itself is a broad church, right? There's a number of different, kinds of AI. The term actually dates back to the 1950s as a concept for you know, machine thinking. It's had a couple of false dawns over the time when compute and data to train. I wasn't really quite ready for this, but as we got into the 90s and the early noughties, as compute power grew, as storage grew, a confluence of internet accessibility, lots of data becoming available, and then we time fed forward.

00:02:15:12 - 00:02:33:12
We found that organizations could do the fundamentals of what we know of AI today things like machine learning. So learning a trend and a pattern, looking at what happened in the past and do a statistical regression on that to predict some future outcome based on what happened in the past. And we use examples of this today without even knowing it.

00:02:33:12 - 00:02:52:11
You know, is this email that's coming into my email system, is this spam or not spam? Those kind to classifier types of AI have been prevalent for the last ten, 15, 20 years, and we're moving forward to where AI has this more kind of human interaction. It's surfacing and it's suddenly popped into the zeitgeist, for for conversation.

00:02:52:15 - 00:03:14:03
So it has multiple facets. We have machine learning trained something to do, something very specific, show it, something that it's seen before and enable it to predict the future based on what it's learned. But we're starting to see this wave of generative AI do more advanced, more nuanced, more humanlike things, and I think that's a really powerful kind of inflection point that we've seen in the last two, three years.

00:03:14:05 - 00:03:39:02
Thank you. So because in your first, answer, you said you said about the 70s and 90s, but why is I having such a huge moment right now? So what changed since that time? I think that the real inflection point is the the kind of conversational nature of it. You can speak human to it, and it can speak human back to you.

00:03:39:04 - 00:04:01:13
If I think about how compute evolved, you know, it used to be I had to type cryptic commands on the green screen in order to be able to use a computer, which meant the audience of people who could use computer to do something was very limited. In the 80s is the GUI. The graphical user interface kind of emerged suddenly it was a keyboard in a mouse, and the population of people who could interact with the computer was much broader.

00:04:01:15 - 00:04:19:02
Mobile did the same for us, but you still had to learn things. You had to take the human to interact in a way that made sense to the computer. With generative AI, I think what's happened in the last 2 or 3 years is actually the computer is coming to meet the human. Suddenly it's able to interact with us in our language.

00:04:19:07 - 00:04:37:19
I can have a conversation with it. I can ask a question in natural language. Now I might need to engineer my prompts to get the right kind of outcomes to guide it. Actually, the computer understands what I say. It can meet my language and understand that interact with me in a very human way. And I think that's caught the imagination of people.

00:04:37:19 - 00:04:59:18
They've suddenly had this 'aha' moment and that then has gone from, you know, an academic or data or IT kind of problem. It's broken out of it and gone into the board to say, well, actually, what does this mean? How will this work? And as people start to imagine what it could do beyond, you know, asking a question about, you know, what recipe do I have?

00:04:59:18 - 00:05:20:13
Or how can I find an answer to a question I could historic could use a search engine for, but save me some heavy lifting organizations to look at it and say, oh, hang on a minute. What manual processes in my organization...What low value repetitive tasks are happening in my organization that this might help me change? So suddenly AI has gone from being an IT conversation to being a business conversation.

00:05:20:13 - 00:05:48:15
It's it's got the opportunity. It's got the ear of the board. And suddenly that's just pivoted the demand and the interest in AI I think in the last couple of years. That is quite insightful. So because I has become the big thing in the world and everybody is talking about AI, there's got to be some, some common myths or misconceptions about AI out there that you've heard give us one or a couple that you've you've heard that you need to clear up and be like, that's not actually the case.

00:05:48:17 - 00:06:11:19
So there's a couple of things that I think, reoccurring in the conversations I have with customers, with, with engineers, with particularly people outside of IT. And one of those is around privacy. And I think that the challenge that we have with AI is the first services that really burst this into the public domain. There's kind of ChatGPT services.

00:06:11:19 - 00:06:31:04
There's first, opportunity where you could just go to a website, sign up for free, try something for free, engage with it and have a human like conversation. But that spread like wildfire, like 100 million users in a crazy amount of time. The interesting thing there is that free service, and I always like the phrase if something's free, you are the product.

00:06:31:06 - 00:06:54:21
That's those kinds of public sites where it's, you know, it's a consumer-grade service. There's no charge. The huge costs sitting underneath those models, like running the infrastructure, running the applications, having train the models. So the reality is in that environment, the value exchange it was happening is the prompts that I give that free service are available to be used to retrain the model to extend it, to make the product better.

00:06:54:23 - 00:07:24:01
So you're giving access to the data that you provide through a prompt to the service provider that is running that service. That's the value exchange. Now that's created this perception in people's minds that AI isn't private or safe or secure. And I think the reality is, when you do this in an enterprise context, you can absolutely run those models in a ring fenced way, the same way you'd run a database platform where it's isolated.

00:07:24:07 - 00:07:41:09
It's not sending data back to the model provider, it's secure and it's yours. And that enables you to do things. Bring your private data combined with the intelligence that the model has been trained on with public data. And that's what builds builds a system. But it doesn't have to be a system where you're losing control of that data.

00:07:41:13 - 00:08:02:22
So I think there's a lot of FUD around that fear, uncertainty and doubt. And it's up to us as technologists to help dispel the myths and separate where that might be happening in certain domains. That free service is public services. Maybe that is happening, but in an enterprise it scenario, you absolutely can put the security and privacy guardrails around it to meet the kind of enterprise controls that you'd expect.

00:08:03:00 - 00:08:36:10
Whilst reaping the benefits of the AI productivity gains, that you could have. So I think that that to me is the big one. Awesome. Thank you. So, because you said that, you you talked about AI, in industries, and how it's used. And I really like the analogy with the database. So for us, with MySQL, we really enjoy, the databases, could you, paint a picture of how AI is being used across the industries, or is it just specific, or can we use it, in different ways?

00:08:36:10 - 00:08:58:07
And, now it's a great question. I think, like most technological innovations, the thing that is most disruptive about AI is it has an opportunity to be a general purpose technology. And so if I think about things like the internet and electricity before it, electricity is a general purpose technology, right? It's one thing that it is it's ubiquitous in society.

00:08:58:09 - 00:09:21:11
But it's used for many things. Right? It's used for the lights in my room, for the microphone, the router that's routing this, this conversation to you. It's also used to heat my house. It's it's used to generate, to run factories. It's a general purpose technology. The beauty of that is it's power and it's ubiquity and it's up is only constrained by the imagination of people who take electricity and think about what problems could I solve with it?

00:09:21:13 - 00:09:46:16
I think I will be very similar to that. But it's up to us in industry, in technology to invent ways to use this, that are productive, that deliver value for our organizations or for society at large. And the real opportunity there is is boundless, is captured only by our imagination. What I am seeing is there at the very specific first mover type, use cases that are happening.

00:09:46:20 - 00:10:08:10
And they might be things like, you know, drug discovery and protein folding, like highly academic, data science led things. They're moving really fast because those are things where data scientists were already doing lots of work, they were already doing machine learning. They were already up and running with AI. What I'm noticing is that enterprise adoption is a different kind of material, right?

00:10:08:10 - 00:10:30:00
It's a different kind of IT problem to go solve. So what we're seeing is enterprises are experimenting. They're doing lots of pilots. They're they're kind of engaging in, you know, the art of the possible. How could we use this in our organization? What things do we not know about how to do this? We haven't trained our organizational muscle to be able to go from idea to pilot to production yet.

00:10:30:02 - 00:10:51:11
So what I'm seeing is organizations look at human in the loop scenarios. So they're starting with applications where AI is helping a task that already happens to happen a bit more efficiently, a bit more effectively, or drive more coverage. And my favorite example of this was, is in regulated industries where actually, you know, organizations are a bit fearful of upsetting the regulator.

00:10:51:11 - 00:11:10:22
And, you know, we're using AI. And what's the governance challenge with this? I work with a few organizations. You've actually turned that on his head. And what they've said is, how can we use AI to improve our compliance, and regulatory frameworks. So they were looking at this and saying, well, you know, today we have a contact center and we have a team that listen to all the recordings.

00:11:10:22 - 00:11:30:19
Well, actually, they listen to 5%. They sample the recordings and they look for compliance challenges. And then they use that to inform how they educate people and report with compliance, status. So they said, well, actually, why don't we have I listen to all of the calls and then the team that were previously only listening to 5% can go and mark the AI's homework.

00:11:30:21 - 00:11:47:08
And this creates value because now I've improved my compliance perspective on screening all of my phone calls. And the people who are listening to those calls and marking the AI's homework, they can improve and iterate on the model and make it better over time. So we have that human in the loop. So it's augmenting the capability of a team to do something and improving the outcomes for the organization.

00:11:47:14 - 00:12:09:11
I think when you start with use cases that are in that kind of domain, the organization can learn, can adapt and then understand how do I apply this to other problems. And it really has to come from what's the biggest problem in my organization? What's my strategic objective? How does that relate to a data strategy, to an AI strategy to go solve those business problems I want to solve?

00:12:09:14 - 00:12:31:19
And that's the real connective tissue here. It's not a science experiment. It's not AI for the sake of it, just like it wasn't data for the sake of it. It's about data to solve a business problem, help us take action in our organizations. That's awesome. So the three of us obviously work for Oracle, and there's been a lot of news about what Oracle wants to do in terms of AI.

00:12:31:19 - 00:12:58:02
And, you know, are we currently a significant player in the AI world or are we going to get there eventually, do you think or, you know, is it is there is there some other path for Oracle in terms of AI? I think Oracle has a unique position in a number of ways. So if we think about the news that you're talking about yeah, there's lots in the press today about the huge investments that we're making, the giant partnerships that we're doing.

00:12:58:04 - 00:13:21:13
These are about the industrial scale infrastructure that will be needed by organizations, both to train the next generation of these models, but equally to run and inference them. So if you're an organization that wants to consume AI, you want to do that scale. You need that bulletproof, high performance, low latency infrastructure that is secure and robust in order to run the workloads that are powered by AI.

00:13:21:15 - 00:13:43:20
If you're going to do this in an enterprise fashion, you're going to want to do that in a robust, secure, resilient fashion. So building out that infrastructure, the Oracle Cloud infrastructure that we have today, the strong partnerships we have with, GPU providers and software vendors like Nvidia, these are the kind of raw foundational capabilities at absolutely epic scale that are critical to this success

00:13:43:21 - 00:14:21:20
in AI and Oracle's right at the front of that. Interestingly, though, it's not just about tin in data centers. It's about the software stack. It's about the ability to take that raw compute and augment it securely with robust data practices, bring data into the world, bring AI to where that data lives today. That's where I see Oracle being really powerful between huge database platforms from Oracle to relational database platform to MySQL, these are key capabilities and your key software assets that will help organizations unlock the power of that infrastructure and bring it to life in their organization.

00:14:22:00 - 00:14:53:00
And then at the other end of the spectrum, you have, SAS applications at fusion. These are the business process tools, the systems of record, the organizations trust to do work for their organization. They have key elements of data, and they operate they run business processes in your organization. So the ability to surface the outputs of the AI and applications that business users use so they can understand it, use the, you know, interact with the data, glean insights from it, leverage the power of AI to take action for and with them.

00:14:53:02 - 00:15:13:21
That that combination right across the stack, I think is where Oracle is uniquely positioned. And and hence,I am here. Excellent. Yeah. Very nice to see that Oracle actually is a big player in the AI and had the opportunity to see plenty of, of stuff on that tool like the data centers, who we created the, for it.

00:15:14:03 - 00:15:44:00
And, so and yeah, in MySQL, we, we already also see with MySQL HeatWave what brings to to AI there. But, so with that position of of Oracle, going on lot on AI, do you think it will impact, the, the product portfolio, of Oracle, like some stuff to, like, MySQL we know about it, but for other products, do you think that will also impact them?

00:15:44:02 - 00:16:08:19
This, this role of Oracle in the industry? I think it will. I think it will it will bring a new gravity to, to the solutions that we offer. I think the other component, and you're seeing it with MySQL, you see it with Oracle, you know, actually, how do we take the greatness of the database platforms that we have and extend that to simplify organizations use of new technologies?

00:16:08:23 - 00:16:31:13
And you know, my favorite example of this is how do you enable the existing database to do more of the tasks you need in an AI world? So with that, I'm thinking about vectorization, storage of vectors, the ability to run inferencing close to the data. I don't have to pull all my data out of the database just to then run some inferencing over it.

00:16:31:18 - 00:16:58:15
How do I bring that AI capability directly to where the data lives? So I think we're seeing that with lots of the product innovations. And we're also thinking about like what does this mean to governess. You know, if you have a solution where, you know, you've become used to as an organization governing and managing a relational database, how do I then work in a world where I have unstructured data, structured data I have now vectors, it's these are all living in different store data stores.

00:16:58:17 - 00:17:13:12
How do I govern and control that? How do I make sure that I'm keeping that data in in sync? How do I make sure that I've got my GDPR compliance correct? A customer wants to be forgotten. I've now got more places that I need to forget. The customer. I, you know, update that data because it has to be correct.

00:17:13:14 - 00:17:42:02
So I think this concept and we see it across the MySQL platform, we see it across Oracle database. Actually by bringing the vectorization, the vector storage, the vector generation, the the ability to query right into the database engine, you simplify the operational management, you simplify the governance of that model. It makes it easier to secure, to manage access in ways that your organization is already familiar with, by managing a MySQL estate or by managing an Oracle platform.

00:17:42:06 - 00:18:04:09
So so suddenly you're able to expand the scope of the things that you do without it bringing extra operational and governance, complexity into your organization. So it's already influencing our product portfolio. It's already changing the way that we expand to help organizations take advantage of these new needs, these new demands and services, but bring those in a way that makes them part of the existing ecosystem.

00:18:04:09 - 00:18:22:15
They're using the Oracle. And of course, that will continue to evolve in ways that, you know, if I had a crystal ball, I would I'd be looking at what those might look like. But, you know, the key here is that we're moving early, we're moving fast, and we're learning from those, demands and evolving products to help organizations gain the value of that.

00:18:22:15 - 00:18:49:14
They don't have to invent all of these capabilities themselves. They can consume them baked into the products they already used. So I know from the MySQL side that we have customers who have terabytes or petabytes of data. What role is is that data going to play in building or benefiting from AI? And again, I'm talking particularly about like structured data that would be in a MySQL database.

00:18:49:16 - 00:19:12:13
Got it. So so if I think about that, that kind of structured data often that's going to be data that represents entities or processes in your organization. Right. It is the state of a process or of, of an entity, a customer, an order fulfillment, something that exists in the real world projected into a piece of data in that database.

00:19:12:15 - 00:19:54:03
And if we want AI to be a part of how business gets things done, runs a business process, it's going to need to have secure, robust access to well-trusted, grounded data that represents the real world. And I think that key is where AI, in the large language model, the kind of ChatGPT I can interact with it, I can have a conversation with a process that's that's trained and kind of sealed in its data set that it was trained on, but it brings in intelligence that helps it understand a question, interpret language to, perhaps reason over some of the, the, the assets that you've given it as part of that prompt where it becomes

00:19:54:03 - 00:20:16:23
really powerful is in the process is this commonly been referred to as RAG or resource augmented generation. This is the ability to take your private data and securely add it effectively to the prompt. So you add lots of context to the question that you asked the model. Now I can use its intelligence and the ability to understand based on the public data it was trained on.

00:20:17:01 - 00:20:37:21
And in response to the question that you've asked, it can also now answer that ground in your in your own data. So if that data is the structured data, you know, it's about an order, it's about a product description, it's about a fulfillment or an employee. Then suddenly you have the ability to look at that private data set and reason over it using the intelligence from the large language model.

00:20:37:23 - 00:21:04:08
So data will be fundamental because data represents the real world. Data represents the things that we want our business to do. So if you can bring that data, enable it to be composed with that large language model, with the AI, then the AI suddenly can do things in our organization. It can provide insights into our organization. Or if we think more about agenetic AI, it can start to take action, force or recommend actions enable things for us to be done on our behalf.

00:21:04:10 - 00:21:27:17
I think that's where we start to see the flywheel really turn structured data that represents business processes powered by large language models and simplifying the way that that kind of ecosystem, combines. That's where we'll unlock real business value. Enterprise value, versus helping me my homework. That's great. So thank you very much, for that information. It's very insightful.

00:21:27:19 - 00:22:04:15
So yeah, I'm very happy to, to, to, to start this, this third season, with you, Matt, about, AI and we will see in the future, episodes, also in the next one, everything related to and more in with MySQL, of course. But, I think it's very, very interesting time, for people to test AI and, for the people who will listen to us that they can play, on OCI with HeatWave there is a free, HeatWave that has also, AI capabilities.

00:22:04:18 - 00:22:27:11
There is also the, OCI, GenAI service that can be, useful to play with. I, play with the, with both of them. And it's very, very, very interesting and, surprising. Oh. It works. I don't know if you have something else to add for us, but, we were I was very happy to to to chat with you.

00:22:27:13 - 00:22:48:16
But I I'll challenge one thing, Fred, that you said, and I'll be slightly cheeky on it, but playing with it and experimenting, it is step one to learning what can be done, but will only really learn how to do this when we start to practically apply it to real world problems. So we need to move from this kind of experimentation and pilot phase.

00:22:48:20 - 00:23:05:16
That has to happen. And as individuals, as technologists, we will have to do that to learn and get to grips with this technology. But we do need to find ways as organizations to actually do this in anger. And I think, you know, I always use the phrase if you if you want to run a marathon, you start by getting up and running the five K, right.

00:23:05:20 - 00:23:23:02
And it's you do a real run and it hurts. It hurts like hell when when you do that first one. But it becomes easier as you do more of them and you start to expand scope. You start to get longer. You can do bigger runs. Organizations need to do that same piece and train the organizational muscle. You'll do it with real world projects.

00:23:23:04 - 00:23:42:13
We absolutely need to learn how to do this and experiment and learn. But the best way that an organization can, can learn to do this quickly is to find a real world problem to solve and work back from. Why does this organization need to use AI to do this? What problem can it solve for us? And then think about how can AI help us do it?

00:23:42:16 - 00:24:10:08
I think we can get that flywheel going. The playing with it will inspire us. But that's not the end game, right? That's just this chapter one. Yeah. The playing around I think, feeds the, the, the need for lack of a better word that, you know, a lot of times, like I come from a developer background and a lot of times, customers or clients didn't necessarily ask us for what they wanted.

00:24:10:11 - 00:24:36:05
They asked us for what they thought we could deliver. So the Henry Cole thing. Right. I want a faster horse. Exactly. So instead of, you know, saying, hey, this is the problem we have, they're like, well, this is how we think you can solve it. And I think AI is kind of the same thing where people don't really know what the capabilities are, or they're asking for capabilities that they think are the limits of the AI rather than the capabilities that they actually want.

00:24:36:07 - 00:24:52:02
And I think we're going to get to a point probably, sooner rather than later, that we're going to realize that AI can help us with a lot more stuff than what we think it can do right now. Definitely. And it will it will meet in the middle. Right. The business will be saying, I've got these problems I want to solve.

00:24:52:02 - 00:25:17:13
And I think AI, as part of the solution, and developers and technologists who've taken the time and invested the energy to go learn the technology, to see the are the possible, they can then be inspired by the kinds of things that happen when those two meet in the middle. That's where we'll see a real innovation coming in organizations doing really clever things and taking the great products and services that we've built on Oracle Cloud infrastructure in MySQL, in HeatWave, in Oracle database.

00:25:17:14 - 00:25:36:04
The main aim of those is to make it simpler for organizations to take those ideas very quickly, pilot them and prove value. But it's not about piloting them in isolation. We need no cliffs, right? We need to get to the point where when that pilot's ready, we can securely robustly, deliver that into production and we can scale it.

00:25:36:09 - 00:25:56:19
I think doing these, experiments in these enterprise scale, frameworks, in the tools that we provide that gives organizations a route from pilot to production. And that's the bit that I think organizations are really craving. And it's a we we're about to see a real inflection point on that fantastic. Matt, again, thank you for joining us.

00:25:56:21 - 00:26:10:14
I think this has been a great conversation, and I really think that our listeners are going to get a lot out of it, and hopefully it whets their appetite to learn more about AI in upcoming episodes. Thank you very much for having me. Great, great to talk with you. I look forward to listening to all of their story.

00:26:10:16 - 00:26:30:06
Thank you, thank you. That's a wrap on this episode of Inside MySQL: Sakila Speaks. Thanks for hanging out with us. If you enjoyed listening, please click subscribe to get all the latest episodes. We would also love your reviews and ratings on your podcast app. Be sure to join us for the next episode of Inside MySQL: Sakila Speaks.