Introducing Machine Learning Workloads

0:00[VIDEO LOGO] Welcome to the content on the AI 900 certification exam. I'm really excited to be sitting down and recording this content because my passion, my sweet spot has always been leaning towards the cutting edge of technology. Whether we like it or not or whether we realize it or not, machine learning and AI is now relevant and prevalent in nearly every single tool that we're using behind the scenes.

0:32Even the devices that we interact with every single day are now being designed and built specifically for AI workloads. Hey, Siri. What is AI. See. This is AI. This is machine learning in action. You've just witnessed it right there. The actual phone device that I'm working with right there in the spot have the ability to process the words that I was actually using, the language, the speech, the tones, the dialect.

1:06It was able to detect that. I was asking a question and then specific details about that question, and then respond to it and reply with it. How they got Siri to do that with so much accuracy, that's really where machine learning comes in. So in this set of videos and throughout the rest of this course, we're primarily focused in on how Azure, the cloud, makes working with AI tools and machine learning and building out all of our processes to run in the cloud responsibly, safely, reliably, and cost efficiently.

1:43That's what Azure is bringing to the table. And that's why they have a certification track dedicated to just their AI tools. So right now, we're focusing on what can we even build with AI? Where is AI being used? And then we'll wrap that up with where does Azure come in? Then throughout the rest of this set of videos, we're actually going to focus in on what the individual tools are that Azure brings to the table and how they actually work with building your AI workloads. Let's take a look.

0:06Nowadays, most of the professionals or thought leaders in the AI or machine learning space, they've all come up with different terms for what they want to use to describe their workloads, their algorithms, or what it is in general that they're doing. So what is the difference between AI, and machine learning, and data mining, and knowledge mining, and deep learning, all of these different terms, cool buzzwords that you see get tossed around? Really, at the end of the day, it all boils down to three things.

0:37AI, which is the all encompassing artificial intelligence and then underneath that is what makes AI work? Basically, what is it trying to solve? What is it trying to do? And that's machine learning or data mining. These are two different ways of training AI to think intelligently. So let's explore this just a little bit more at the concept level before we start talking about what the different workloads are that use AI. Let's go.

1:08So I've got VS Code brought up here on the screen on my computer. And yeah. You're going to see me work a lot more with VS Code in an upcoming set of videos where we do a little bit of a machine learning crash course. But the first thing I wanted to do before we dive into this, is of course talk about what you can do with AI machine learning, data mining, and even more. But that's kind of the point is we hear all of these buzzwords get tossed around AI. Machine learning, machine mining, or knowledge mining, or data mining, or deep learning.

1:44And it's like, what is all of this? Are they that different from each other? Sort of. They all have their own little flavors. They all have their own little styles or dialects. But if you were to look at it as a tier, you would say all of it, all of these words, they all belong to the family of artificial intelligence. The idea with artificial intelligence when it was first surmised by IBM was could we teach a machine, a machine, to play checkers and eventually learn from all of the games that it plays.

2:24And it improves every single time it learns about moves to the point to where could the machine ever learn enough about checkers to beat the person who created the algorithm to train the machine. Ultimately, the machine its all purpose here is to learn like a human learns from experience and then derive predictions, make predictions about what should happen next.

2:56So if we're thinking in the game of checkers. If I'm hopping my guys around, and they make it to the other side, and you get to king them or whatever. You know what I'm talking about when we're playing checkers, right? Eventually, the machine is going to be learning from its own moves, as well as my moves. It's going to be collecting data on every single move that I make until it gets so good at understanding the game as well as my behaviors, that it can accurately predict what I'm going to do next and more or less front run me and basically beat me at my own game.

3:32So the idea here is we can take machines and artificially teach them how to take and ingest data and learn from that data in order to make guesses on what would happen next. So artificial intelligence is using different learning techniques to ultimately make the predictions in the applications that we use every single day. So that's where the rest of these things come in. Now, if I were to really break it down, you could pretty much safely say the rest of all of these buzzwords really boils down to machine learning, and split it in half, and data mining.

4:14So this kind of begs the question, OK, well, what are the differences of machine learning and data mining? Ultimately, what we're trying to do here is we're trying to find ways for the machine to learn. So that way, we're talking about if it's playing checkers again, and it's AI that we're talking about here, when it collects all of this data about my moves and its move, how is it going to figure out when something was good or when something was bad? And this is what machine learning and data mining really bring to the table.

4:48With machine learning, we typically start off with inputs and outputs. So for instance, if I'm going back to my checkers analogy, my input would be the moves that I'm making, and the output would be, were they good moves or not? Like what was the result of those moves? Did I end up jumping over one of their little guys, or did I get a king, or whatever the case. Or was it a bad? Did I make a mistake and set myself up for failure?

5:20So with machine learning, we have inputs, and we have outputs, and we usually already know or have a little bit of an idea of the correlation between these. If something came in and we know the output, we can determine if that was good or if that was bad. So we can train the machine using statistics and all sorts of other algorithms that we'll talk about. We can train the machine based on all of these inputs and these outputs. The more and more examples of this that we give you, the better or more accurately you'll be able to predict it.

5:56With data mining, it's a little bit different. In machine learning, we know the correlation between inputs and outputs. But what if we didn't? What if we had inputs and outputs and it was really at the end of the day just a humongous amount of data? And we don't really know because it's just so much data that we've collected, and it's maybe unstructured, we don't really know if this data came in, and that data came out. We don't really know how it got from this point to that point.

6:27So with data mining, we're training our machines to comb through the data and find relationships or paths to various outputs. And as we get more and more inputs, we get more and more results, we're able to find more and more correlations between the inputs and the outputs. So as we start to learn the correlations between the inputs and the outputs, then we can judge if this is going well.

6:59We can judge if this is a good correlation or a bad correlation and use that in machine learning. See? It kind of goes full circle there. So this is just really setting up a very short video to talk about what are all of the pieces that make up this puzzle? The idea is we're training machines to learn from data based on inputs and outputs and make judgments or predictions based on those inputs and outputs. And sometimes, we don't know the correlation between our inputs and outputs.

7:30And that's why we lean on data mining. And once we've got a good correlation, a good sample set of inputs and outputs and their correlation, we can train the machine to learn more accurately how to predict what comes next. So now in the next few videos, we'll take a look at real world use cases of where AI is being used. I hope this has been informative for you, and I'd like to thank you for viewing.

0:00[AUDIO LOGO] Language processing is the one that you just witnessed me do in a couple of videos ago when I was actually introducing this and I was talking to my phone, asking it a question. The idea here is that we have a machine, like an application that's just running, and we want it to be able to understand questions, maybe coming from a customer who's chatting to a bot for help, or maybe coming from my voice to my phone device, or my watch, or whatever the case is.

0:33The core concept of it is that it understands questions. It understands the language. It understands the tone. And it understands the task at hand. Is it a question? Is it a command that's asking here? So let's talk more about what language processors bring to the table and where you interact with these daily. In particular, we got a fun one coming up, talking about ChatGPT. Let's go. So I don't want to say something like the most popular form of AI that's being used, but I will say something like the most mainstream form of AI that's being used is language processors.

1:10Now, what are language processors? Language processors can listen or read languages. That's kind of the idea at the end of the day. And from these language processors, we can learn a lot. It's not just translation or interpretation that we're talking about. We can actually look at language or listen to language and detect things like the tone. Can we tell the sentiment of what is being said?

1:42We see this a lot on Twitter. Let's jump on Twitter for a second here and see what's going on on Twitter today. See, I see some pretty cool things. I see someone's auctioning some artwork off. I see someone is telling you to go outside, move your body, and enjoy nature. We see someone's advertising a video game here. It's kind of what's going on. And then we've got some jokes from like Barstool, or whatever the case is. I don't know. The point is if I were to put this into Azure's-- Azure has a tool here.

2:14I'm spoiling it for you-- hang on, I hit a button-- Azure's Cognitive reasoning or learning tools. I can tell Azure, parse this tweet. And tell me if this person is happy, sad, angry. We can ask them, what type of mood is coming from this tweet? Now, take it a step further. What if I wanted to say, hey, I'm running a campaign that tracks hashtag #DataKnox right here. And everybody's going to say something about hashtag #DataKnox.

2:48And I want to figure out if all of these tweets that are hashtagging #DataKnox, I want to figure out if they're happy or mad or sad or euphoric or whatever. I want to figure out what the sentiment is for this particular trend, for this particular thing that's happening in the world. And that's where, again, Azure tools like Cognitive reasoning and learning come into play. And this is actually where you see like these trends and what's happening right now and understanding what the sentiment is around them.

3:22All of this behind the scenes is AI and machine learning. In real time, effectively, Twitter is analyzing the amount of mentions for Mac mini, and hashtag #MATIC, and whatever Antonio Brown is into now. It's determining what is happening in its own feed, in its own data. And then it even determines, why is it trending? What is the sentiment around the things that are trending? And who should I be promoting this to?

3:53Based on predictions, based on things like-- hey, this guy Knox, he really likes these types of tweets. So we see that this type of tweet is trending right now. And we know that Knox likes those types of tweets. So we predict that this should be in his feed. All of this comes from processing language and just understanding how end users interact with that language. But in recent history, language processors have become so, so much more thanks to a tool called ChatGPT.

4:27Honestly, I say, thanks to a tool called ChatGPT. Tools like this have existed for a long time. But ChatGPT has really gotten it to a place where it's really special. Oh, no, it looks like it's down right now. Hang on a second. OK, I got in. It took me a little bit, but I ended up going to the actual beta version of ChatGPT. That's chat.openai.com/chat. That's how I end up here on ChatGPT. Now, the idea here is that I can effectively tell ChatGPT to do anything for me.

4:59And it's not just answer my questions like looking up things in Wikipedia. No, I can tell ChatGPT to create things for me and give it really specific details that I want it to use. So down here in the little chat box right here, you tell it what you want it to do, like explain how Microsoft Azure benefits users who work with artificial-- got to spell "artificial" correctly-- artificial intelligence, something like that, just on the fly.

5:37And I press Enter here. It types it in, and here comes my results. Look at this. Do you see? So it's just looking this up. That's a very basic use case of how this comes in. Look at this. I mean, it gives you your answer right there on the fly because it understood exactly what it was that I told it to do. So I'll let it get done generating the response here because it's going to type out a list of really cool things that Azure brings to the table for people who work with AI. And then I'll show you how this goes even further.

6:08What's interesting about this, now that it's done, I can ask follow-up questions. Even cooler, I can even challenge AI. I could say things like, are you sure that's right? Or I don't think that's right. And it will actually learn from that and correct its mistakes. I can follow it up with, tell me more about GPU-enabled virtual machines in Azure. So we let it do some thinking, and here it comes back.

6:40And it tells me more about Azure-specific information of GPU virtual machines. And it even tells you how. This is parallel processing-- a high degree of parallel processing power required for deep learning and artificial intelligence workloads. Keep in mind here, it is keeping on topic for the initial conversation of AI. So now that it's done answering all of that, watch this. I'll say, now create a poem about Azure AI offerings.

7:16Watch what it does. Just watch. Here it comes. Do you see? Do you see what's happening right here on the screen? I didn't teach it how to do this, but it did. It learned how to do this. Look at this. [LAUGHS] It's kind of awesome. This is why I think language processing has taken on a new level. I mean, as of time of this writing, ChatGPT has really only been a mainstream tool for like a month at this point.

7:50So we're very early to ChatGPT as a tool. But it not only is able to follow up in a dialogue with follow-up. I can actually tell it be creative. I can tell it be creative with things. I can even tell it create full-blown applications for me. Now, that's a slightly different form of the language-processing tools that OpenAI brings to the table and offers. But it does actually have a code generator.

8:21You can tell it things like, using this programming language, using Python if you want, generate an application for me that tracks stars in the sky at night. I don't know. But you could tell it to do that, and it will literally spit out an application for you in real time. So it's kind of crazy. But where else do we actually see creativity and AI combined? Let's do AI art. AI Art Generator, look at this right here, hotpot.ai. So I can tell it to generate art on the fly.

8:53What should AI draw? How about a computer hacker in knight's armor? How about that? And then we'll scroll down and tell it Create. Boop, there we go. So let it do a little bit of thinking. And in roughly 10 to 15 seconds, we should get some form of artwork that generates a computer hacker in knight's armor. So let it spit and spin for just a second. And then once it's finally generated the artwork, I'll circle back to it.

9:24And there it is. After about five more seconds, this is my computer hacker. It's actually really cool, right? I'm going to save that. It's actually really cool artwork. So anyways, you could tell, using AI art generators, it can process the language of what I expect to see. And because it's also processed images, things that we're going to talk about in the next video-- because it's processed a boatload of images, it knows, based on these language inputs, what accurate outputs should look like.

9:57I would say it got this to a pretty good degree of accuracy. I mean, the armor here, this looks more like cloth down here. So I would kind of argue this isn't like a full knight's suit of armor, but it's got a cool looking helmet thing. And maybe this is like some shoulder pads or neck pads. I don't know. Maybe it's like just an armored hoodie. I don't know. But the point is, to a degree of accuracy, it was able, within a matter of seconds, to understand what it was that I was telling it to do and spit out some results.

10:29So the thing about-- you know what, actually, I was about to get it worked up. I was about to jump ahead. We're saving it for another video. Just hang in there. All right, this has been understanding language processing. It's all about taking instructions or listening to language and knowing what the request is and how to respond to it. I hope this has been informative for you, and I'd like to thank you for viewing.

0:00[MUSIC PLAYING] The next fascinating thing that we are going to talk about is video object detection. At the surface level, you may look at this and go, oh, that's a cool gimmick, but I don't really see what the real world use case for this would be. Then, we'll talk a little bit more about where this is really being used today, and some of the business use cases for it. And this part will blow your mind. Let's talk a little bit more about it.

0:30So another way that you'll find some pretty fascinating things in AI is just right here on Twitter because people are doing really incredible things with language processing as well as video processing. So in the previous video, we took a look at language processing. I just wanted to circle back to. This is really cool. This particular artist is an artist, and actually a fascinating AI follow in general. You probably want to give them a follow. I asked AI to turn countries into women, and it actually created some pretty fascinating results here where it took the generic styles and imagery of people who are from those countries and turn them into pictures of, I guess, what women would look like from those particular countries.

1:13It was fascinating to look at what I thought. But now we focus in on what can I do with videos? And right here, you're looking at it. You see this? Isn't that wild? Basically, these are significant people throughout history, whether they're fictional or not, and we see how AI has the ability to track exactly what's happening on these individuals faces. As I was sitting here drawing, we see how AI has the ability to track exactly what's happening on these individuals faces, and alter the artwork or the video work or the actual images or videos themselves to mimic exactly what the original source person was doing.

2:04AI now has the ability to track every single one of these aspects of an end users face and understand exactly what is happening on that face, which I think is just so fascinating. We even see this in Teams, Microsoft Teams or Zoom. You've ever seen that blur out the background? How does it know that was the background? How does it know where my body is? It's because it's used enough machine learning from processing images to know what the outline of a human would be.

2:39It knows where the eyes are, and it knows, oh, well, this little portion here must be the nose. And it knows, oh, this little portion here must be-- I mean, it's the worst drawing of a face ever. But this must be the person's mouth. And it can detect where all of these objects are on the screen, and then draw an outline of it. And that's why it knows how to track the facial movements and the body movements and blur out everything else behind them. It's because it's looked at enough images and been told what is the outline that it should expect.

3:15But now, even more fascinating, sets up this particular conversation, which is about tracking objects in videos again. Oh, I want to show you this one. This is pretty fascinating here. This is not Tom Cruise, but it looks like Tom Cruise, isn't it? See? He's doing the Wednesday Addams dance. But Tom Cruise didn't actually do the Wednesday Addams dance on TikTok. But somebody made an head transmission, and that's like 40 years ago Tom Cruise too. That's Risky Business Tom Cruise.

3:47Or maybe in Mission Impossible 1 Tom Cruise. OK, I got distracted. This is the one that I really wanted to show you though. This is real footage of people, and what we're seeing here is AI has gotten so good at training what is happening on the screen, it knows when to outline a person and track exactly where a person is on the screen. It also shades people on the screen-- hang on one sec. Let's go back here for a second. So what we're seeing here is people doing parkour, they're jumping all over stuff, and it outlines the whole body of the person.

4:18But did you notice-- hang on one second. Let's see if I can get right back here, right there. Do you see? It didn't even outline a whole person. You just got a head. And it was still able to detect, oh, and instantly, it was able to detect, oh, this is a person. There's their face. We're going to outline the shape of this person and shade it red. Cars are going to be orange. We're certain that's a car. And it says right there, you can see of the little dialog box it says, car.

4:50Up here in the pink, it says potted plant and it's shading that pink. So let's keep playing here. Let's actually go full screen with this. Let's actually record this full screen. That way, you can see-- isn't this fascinating? The fact that in real time this video footage can absolutely track exactly what a person or an object is doing in this exact moment is-- look at this. This is really cool. Look at that. You see this? Boats way out there in the water. It's able to detect these boats in real time. I couldn't even tell that was a boat. [LAUGHS] It's able to detect these boats out there in the water.

5:25Now it's confusing maybe some train cars with boats, but still, we're still doing these things with a degree of accuracy in tracking exactly what they're doing in the footage. So what's the use case for this other than it's really, really cool? How is this actually used? And this is the part that's going to blow your mind. The part that blows your mind is now our cameras have built in chipsets. These become more than just cameras. They are really IoT devices. And they actually have the machine learning AI object detection built directly into the camera itself.

6:01My first real experience with this was with demoing the Cisco Meraki line of cameras. I actually was able to get my hands on some Meraki cameras, and you can find training on how to use and implement Meraki cameras right here on CBT Nuggets. The fascinating use case for this with Cisco Meraki was not only can it detect people on the screen, it can then use that data to determine how many people do I have?

6:31And are those people close to objects? Does this create a problem? If it does, let's do something about it. The use cases that Cisco Meraki likes to brag about is let's say we've got a camera that's looking down on a checkout line at something like Target or Walmart or whatever. And it sees, OK, we've got two people. We've got the boxed in with purple and maybe they're shaded in with purple. We've got two people in the checkout line, not a big deal. But all of a sudden, that checkout line gets busier and busier and busier and busier.

7:05And now all of a sudden, people are mad because they have to wait in line. So what Cisco Meraki can do is it can use this object detection to detect people, and when too many people get concentrated in a specific space or maybe they're too close to this particular checkout line object, Cisco Meraki can now send an alert to an application to the manager that says, hey, you need more people working registers.

7:35It can alert them. Even more fascinating, we may have a dangerous piece of equipment. Something that might be prone to exploding, if things go particularly wrong. What I can do, what Cisco Meraki can do is it can do things like detect people in proximity to this dangerous piece of equipment. And if certain conditions or thresholds aren't met well because now too many people are getting too close to this equipment and they're not supposed to be close to this equipment right now, Cisco Meraki can send alert signals to those people, or it can even take actions to shut down the equipment.

8:20Now hopefully, when I circle back to this object detection video that we were just looking at here, now hopefully, you're seeing this in a different light. Oh, here's people moving in real time, and we can detect exactly what they're doing and maybe even predict where they're going next. And if danger lies in front of them, this is what makes Tesla's autopilot work, right here. This is what tells Tesla cars when they're on autopilot and they're driving on their own, if somebody runs in front of the street, this real time detection of human objects is what tells the Tesla to stop.

9:02See? This is really fascinating stuff. And even if it's not the area of AI or machine learning that you're interested or working in, you can now really appreciate what it means to be detecting these objects in real time, and making predictions about where they're going to go next and if danger lies in front of them. So this is the point of image detection and object detection and data processing on images. I hope this has been informative for you, and I'd like to thank you for viewing.

0:00[AUDIO LOGO] With machine learning, the goal is very purpose oriented. We're trying to teach the machine how to process a specific language or how to process and detect things in a specific video. What if we just have a tremendous amount of data and we don't really know what it is trying to tell us? At this point, what we're really focused in on is finding correlations in our data.

0:31What are the paths that our data can take us down? And that's what data mining is all about. So in this video, we are focused in on just elaborating a little more on what data mining brings to the table. Let's go. I'm not a big fan of just reading Google results from the screen, but this particular definition that comes from Rutgers is a really great definition. The process of using computers and automation to search large sets of data for patterns and trends, that's really it. That's where I'm going to leave it, because the idea here-- I want you to think about this-- is some businesses collect so much data, so much extraordinary amounts of data from it, that they don't even know what it means.

1:15They don't know what it's telling you. I want to tell you right now-- your credit card company, maybe it's Visa. Maybe it's Mastercard. Whether you like it or not, or whether you know it or not, Visa knows everything about you. They know not only the obvious things like where you live. They also know how much money you make. They know what kind of spender you are. They know where you like to go on vacation. Or they can predict with accuracy where your next vacation will be. They know how many kids you have and what gender are those kids, without you ever telling them.

1:50They look at every single detail about your spending habits. What specifically did you buy from Old Navy? What specifically did you buy from Amazon? Where did you go on this last vacation? And how much money did you spend on that vacation internationally? How much do you like to eat out? And in particular, what are your favorite genres of food? Are you a Mexican person or an Italian-style person? I like it all, personally. I've never met a piece of food I didn't like.

2:21So the idea here is that businesses collect just extraordinary amounts of data. In particular, like I'm saying, Visa, Mastercard, they're collecting so much data about you that if they were to map it all out, it would be thousands of data points about you that these places, these businesses have collected. You could extrapolate that towards social media. Heck, remember all of this language processing that we've been talking about, LLP, Language Processing, and I was mentioning things like Siri or Alexa?

2:56Those devices, while they're primarily designed-- hang on, Alexa is listening to me right now. Alexa, stop. See? Ask me to play any song or-- Those devices are 1,000% listening to you so that they can collect data to more accurately predict what it is that you're saying. I mean, really, their end goal is to get better at listening and understanding the language and answering your questions more effectively. But at the same time, they're also collecting data points about you and learning things about you so that when this device hears you talking about, oh, I need to add this to the shopping list, well, the next time you log in to amazon.com, guess what's on the home page recommended for you, because they collected the data on you.

3:45The kicker here was at some point, all of these things, all of these businesses, all of these tools, they were all designed to collect data. And now that they've got thousands upon thousands, if not millions, of data points just about you on their hard drives, now they're left going, OK, well, what does that mean? What does this data mean? How can I take info about a person and map that to shopping habits and recommend things to them that they might actually like?

4:22It all started with data mining, where we have to take an input, which is details, all of this data that we've just collected, whether it's voice clips or spending habits or just like even medical history, things like that-- all of these businesses and these hospitals and everything out there is just now data. You're just ones and zeros on hard drives. Now what they're doing is they're taking all of this input data about you.

4:53And then they're looking at what they consider to be output data or results. You bought this. And they're trying to find the correlation between which inputs drove this output. How did we get to this point? And that is what data mining does. It searches large sets of data for patterns and trends, turning those findings into business insights and predictions. Based on the data that we mined, learning the inputs of you versus the outputs of your shopping habits, we now predict-- keyword here-- that you would be interested in buying this thing next.

5:37See how data mining comes to be? It's all about we don't really know how this input led to that output the way we might with machine learning. When we think about the language processor or the video object detection, with the language processor, we typed in language. We got an output, and we could then score that output and let it know if it was good or bad. With the video object detection, we were able to give it video footage.

6:07And then we were able to tell it this is a person. So the more persons that it identified, it was able to know, OK, well, this is accurate or not. With data mining, we don't really know this correlation. And that's why it's trying to find these trends. And this is particularly useful amongst gigantic data sets that are particularly unstructured. Data mining is one of the places, in my opinion, that Azure really shines because Azure can be a data engineers playground.

6:40It's all about aggregating data across a bunch of different sources. Data can come from CSVs. It can come from SQL databases. It can come from document databases, where they store JSON files. It can come from Parquet files. Azure is great at connecting to all of those data sources and aggregating them into a single data source that we could then process with data mining. So now that we're seeing this come full circle and we understand where data mining is, let's talk about the next thing that comes up, one of my favorite things, anomaly detection.

7:14I hope this has been informative for you, and I'd like to thank you for viewing.

0:00[AUDIO LOGO] Data mining can sometimes be used in conjunction with machine learning for a very important reason, and that's to detect outliers. Once we have a tremendous amount of data and we know what is the normal pattern of relationships for our data, when something really weird stands out, recall that an anomaly. This doesn't really fit in with what we know about our data.

0:30So what is happening in our environment that's particularly weird? Anomaly detection has now become one of the most important facets to actually managing today's enterprise networks. We see this from Cisco. We see this from Juniper. We see this from Palo Alto. And really, any provider that works with security at all is now leveraging machine learning data mining, and AI in general, to help them detect anomalies. So in this video, we're going to focus in a little more on what anomaly detection is all about.

1:02Let's go. So I'm getting logged in right now to Cisco DNA Center, courtesy of Cisco's D Cloud environment. This is really a data center cloud based environment that is designed for people to demonstrate, or demo, or just take some of their tools for a test drive. Now the DNA Center is a full enterprise network management solution. Basically, if you have-- think about a large college campus, all of the buildings, all of the wireless access points, all of the people, all of the user accounts, all of the different firewall rules and firewall settings.

1:41There's a lot that goes into managing an entire enterprise grade network. And then making sure they're mobile throughout the entire enterprise and throughout the entire campus-- these are people walking all over the place. They're going everywhere, right? But whenever person a here goes from point A to point B-- here they are not on point B-- they still need access to all of the resources and the rules that they use when accessing the network. So making sure those permissions follow them across this huge network and this huge campus, it becomes a bit of a pain.

2:17And that's what the entire point of Cisco DNA Center is. It's all about provisioning and managing and monitoring large scale networks, and making sure the users who connect to that network have the same permissions and security rules as they migrate through the entire network. Now one of the biggest things that DNA Center does is it actually collects data. It collects a humongous amount of data on the users and the network devices. And it's able to determine, with a certain percentage, first of all, what's the health of our network devices?

2:50How are things going? It's able to detect critical issues, and it's able to even detect security issues that pop up. It does all of this in one of the four workflows that it has here. In particular, once your network is online, you then monitor it and manage it under the Assurance section. Now this isn't a Cisco course. This is just to demonstrate some fascinating things for you. It presents you a bunch of different dashboards, like the health of your network, any issues or events that happened on your network, sensors that you have in your network, Wi-Fi that you have on your network, any rogue wireless issues that you may have, and even power over Ethernet.

3:31And it collects data on all of these things. Now here comes the crazy cool part. It uses machine learning under the hood, as it's collecting a humongous amount of data from thousands of users or devices across your entire campus. And it learns what's normal. It learns what's normal and establishes a baseline for your network. That way, when something weird happens, it immediately flags it as an anomaly and raises an issue or an event.

4:05So I want you to consider this for a second. If I jump into Assurance, and then I go into Issues and Events, I want you to think about how your IT ticketing systems might work currently in your IT infrastructure or your software development world. Typically, a person opens a ticket. It's a complaint or question. And then somebody has to address it. But what if I told you DNA Center has the ability to detect things before a person ever would, and maybe take remediation steps automatically, or maybe open up a ticket for you to investigate immediately.

4:40This is what you're looking at here when you start looking at the DNA automation and services here. And in particular, this works hand in hand with security appliances and implementations. So not only are we detecting anomalies just in the network health in general, we can also detect security based anomalies. Oh, you know what? This signature of this application looks off. Or we know this to be bad. So let's raise an issue and shut down these particular ports before contagion spreads throughout my entire network.

5:14Anomaly detection is now critical in every data center and every ISP around the world. And whether you've realized it or not, machine learning has now become a part of all of the hardware that we implement in our infrastructure now these days. Heck, even the MacBook that I'm using right now that I'm recording this on it has the M2 processor. And they've now dedicated a chunk of that M2 processor to just AI, neural networks, machine learning workloads.

5:45So that way, whenever I'm using XYZ app, maybe that app has some machine learning built into it. Well, it can leverage this processor to make my experience with XYZ app better. This is exactly what you're seeing here with something like Cisco DNA Center, where it's learning what is really going on in our network. How many users do we normally have? What are the workloads of these switches and these firewalls and these routers that we normally have? And if something on one of these looks really, really off, we know something is going on.

6:19Or rather, I should say DNA Center knows something is going on this particular part of the campus. So let's immediately create some sort of ticket or remediation steps to repair and heal something that is happening in our entire environment. This is also really, really popular on security information and event management tools. So these are the tools that are designed to collect log data or any type of data source that we can possibly send to it.

6:50This is specifically built to find anomalies in security issues. We're talking about things that seem off with how our Active Directory is behaving, or are OAuth requests seem off, or something like that. Or an application was downloaded that meets a certain signature, and we don't quite recognize that signature. But it looks bad. So SIEMs are all about detecting those. And these, again, are all about machine learning. But it gets even better.

7:20With all of this, the big thing that makes it work is it needs to be accurate. It needs to, when it makes a prediction, it needs to be accurate. And the only way it gets accurate is with more data. When you typically have something like a SIEM, while, of course, it's not going to leak any private information as it's learning of bad signatures or anomalies that detected attacks, it actually reports this stuff to a global security network so that we can all share and collect more data on security events and more proactively predict and negate security issues when they arise.

8:03So this is what anomaly detection is all about. And these are the types of tools where you see it in action today. I hope this has been informative for you, and I'd like to thank you for viewing.

0:00[MUSIC PLAYING] So this is all great, but where does Azure come in for all of these different things that we can do with machine learning and AI? Do they just host our application and it's up to us to figure out the whole machine learning and training the model part, or can they help with that too? That's what we're focused in on in this entire course. That's what the AI 900 is for. It's to teach people, hey, now that you've got a little bit of an understanding about machine learning, maybe working with our tools could make your experience even better.

0:38That's what the 900 is about. So it's introducing the tools and letting you know what their purposes are for. Then as you progress into your certification journey and you go for a higher level of certification, that's when they start to teach you, here's how you use it. Here's how you get started with these tools. So start off with the AI 900 just learning what the tool is and what it does, and then in the next certification exams, you'll learn here's how you actually implement it in your solution. Without further ado, let's see what the cloud and Azure brings to the table.

1:13So where does Azure come in? Well, Azure comes in and the fact that we are going to be collecting or working with data, with data. Lot of data. Potentially, tremendous amounts of data. Terabytes of data. Petabytes of data. And in order to process these large amounts of data in a quick amount of time, we need specialized computing workloads to process this data.

1:43Whenever I actually create a machine learning model, if I'm trying to process millions, millions of rows of data and I need to process them quickly, I can't use my desktop to do that. Heck I can't even use the servers I have downstairs to do that. I need something truly specialized, a true workhorse to power this type of workload, and I don't have hundreds of thousands of dollars to shell out to get it. This is where the cloud comes into play. The cloud is scalable.

2:16This means that we can effectively get more computing, and more storage resources on demand. And when we're done with them, we tear them down. That makes it cost efficient. In particular, we like to say that instead of spending a humongous capital expenditure or CapEx shelling out 100k at once on a server that now has to last me a few years, instead of using a CapEx expense, I can now use an OpEx or an operational expense.

2:51I only pay for the resources when I need them. And I pay for them on sometimes an hourly rate, sometimes a per minute rate, sometimes it might be per task write. The point is I only pay for what I use when I use it. When I'm done, I tear the resource down and I stop getting billed for it. So I can achieve significantly more resources. I can get as many resources effectively as I need. Not just one server, I could get 100 servers if that's what I want.

3:23And run them for an hour and then tear them down. And then I'll get billed for effectively 100 hours, but that's a lot different than buying one Hawking server that has to be on all the time, and it's a sunk cost. That's the idea. Now, of course, your accountant may say, yeah, but then we get to depreciate the asset. Well, that's an entirely different thing and you'll have to fight with your accountants on that. But here, with this case, we have a predictable cash flow. And we have the ability to get the exact amount of resources that we need when we need them.

3:56Azure is incredible at ingesting and processing and securing and moving data. The data is redundant. There will be at least three copies of your data in a data center. And you have the option of spreading those copies out across multiple data centers all around the world, if that's what you really want it to do to make sure the data was secure.

4:28But now we have the ability to do large scale ingestion of data, as well as processing and computing of that data, and then securing and making sure that the data is backed up and redundant. And then once we have the machine learning models and we want to use those models in applications in the same way Amazon promotes or shows you resources that it thinks you might want to buy, in the same way, we can host these apps that use our models in the cloud.

5:02The idea now is that we can get reliable, always on, always secure, always scalable, cost efficient services as opposed to building everything or self, being responsible for the network and the operating system and security patches and access controls. We can offload a humongous amount of maintenance and management to the cloud so that we can focus on our code or on our solution.

5:35We don't have to focus in on IP address space, or firewall rules, or the operating system needs to be patched. And we don't have to focus that much on, did I buy enough server? Did I buy too much server? Is it always on? How much electricity is it burning? Who is specialized to work this server? Do you see how the cloud is all about offloading management tasks so that you can focus in on your solution while it always remains available online because it's replicated and it's redundant?

6:11The cloud gives us the ability to do these things. But even more so, Azure has a specific AI and machine learning section. If you log into Azure, go ahead and create your free Azure account now. They do ask for a credit card to be on file, but you don't get build anything until you actually spin up some services. So don't worry about it. You're not going to incur a cost until you deploy a resource. So spin up your account now at Azure. And once you get logged in like so, you can go to all services right here.

6:42And it breaks it down based on categories. And right here, we see AI and machine learning. Oh, look at this. Do you see? Computer vision, custom vision, anomaly detector. Here's those cognitive services. That goes hand in hand with language understanding. We even have the ability to create a machine learning accurate search tool here. Then we even have our own machine learning studios where we can tell Azure, hey, we're going to create some Python notebooks directly in Azure, and you'll have the ability to connect your Azure storage services, like databases or database backups.

7:26Oh, look there's even face APIs over there. I missed that. That's really cool. So anyways, this is the idea, is now Azure has the ability to ingest all of our data into storage accounts, and then we can actually do our Azure development right here in Azure and let it handle the scaling and workloads needed to process our data and turn it into a data model, which we can then use in our applications. Azure is here for that. The cloud is an amazing way to do it.

7:56And now when you have specialized tools and Studios and workspaces to actually do your work directly in the cloud, it becomes even better. So this page right here and these resources that you're looking right here on the screen, this is why you're here. This is why you came to the AI 900. You want to learn what these do, and when you would use them. And how you would use them, how you would implement them? That's the next certification exam.

8:26So for now, let's start to understand when we might use these different resources. What is the solution and what is the application look like when it's done? I hope this has been informative for you, and I'd like to Thank you for viewing.