14 hours 28 minutes
Hello, Siberians. Welcome to Lesson 4.3. Off model for off Discuss titled Is it true? Is there one Microsoft Azure architect design?
So here, the leading objectives that will be covering in this video.
We'll start out by discussing the actually I landscape when we talk about artificial intelligence and machine learning algorithms. Sometimes the assumption is that we need to create our own mission learning algorithm models. And that's not always the case. So we'll do a quick review off the island. Scapin has yet to statue. It
wouldn't introduce me as your cognitive service is. What exactly is the service? What does they provide to us?
Then we'll go in for the details around the AP highs that are included as part of the service and gives a brief description off some of the FBI's
wouldn't cover. How has your cognitive service's works, what we need to do if we want to start making use of the service for her applications
and then finally will cover some design considerations for the service? Let's get into this.
So when we talk about the actual here in landscape, as I mentioned, sometimes we can incorrectly assumed that we need to start
beauty in Mission Lenin models. We need to become data scientist. That's not the case. Actually, the first place that we want to start is making use off a high.
Because Microsoft Azure platform itself already makes use off mission than an alcove redeems spot off. The different service is that we consume.
A good example of that is the dynamic treasure, the letting in *** Monitor, which leverages had fast mission Lenin capabilities. So lend a historical behavior off our service on. Then it's gonna use that information that it gathers to hide and fight Parton's on a normal least, that indicates possible service issues.
And this is going to save us from being the one manually configuring, treasured for our lights.
Another good use cases. Security scenarios like that frustrate protection, which is this make mission than an algorithm. So the techno animal ease all suspicious activities.
The next option that we once look as is including hey hi
and that is talking about service is like cognitive service is off, but from work and what that means is, for example, cognitive service is pre trained Mission Lenin algorithm models that we can simply consume without having to be dead. A scientist creating our models essentially mission learning or artificial intelligence.
As a service,
we do not need to create every single artificial intelligence model that we want to implement. We can take advantage of some off. The service is to simply include them in our existing applications.
Then we have the option off creating. Hey,
and as your has multiple service is, that allows did a scientist to be able to create their own mission? Lenin models. And that's referring to,
for example, as your mission Lenin surfaces, which enables the building and the deployment off mission Lenin models using piped on tools and libraries
on even thinks like As your data bricks, so does options that are available Towers on off course. All this is on the paint by the Azure Solutions Gallery or the Ajo Aye, aye gallery, which your cases here and mission that in our providence and just cases for them.
now, when we talk about cognitive service is the what cognitive itself. It's an adjective that comes from the now cog cognition, which means the mental action or process off acquiring knowledge
on understanding that knowledge to taught experience and the senses, in other words, been able to interpret knowledge to it. Thought button or to experience
on dhe. That's what as your cognitive service's makes available to us. It provides Mission Lenin capabilities to solve general problems like analyzing texts, emotional, sentimental, analyzing images.
And we don't need special mission. Lenin ordered a science knowledge to be able to use service. Is it simply an AP I call?
Cognitive service's itself is apparent service that includes AP, eyes that allows us to be ableto hard. This cognitive features and that includes things like visions, peach language, knowledge, search and we can even interact with come in Mission Lenin models and makes off the working, which two laps or research?
Let's look at the FBI's in Monte Toes.
For example, when we talk about the vision feature off, a cognitive service is
so This includes AP eyes like computer vision, which we can use to analyze the content in provided images, for example, were exactly than this image. It includes FBI's like costume vision service, which allows us to be able to customize the image recognition toe feet our business needs,
even though this up retrained models,
there's availability in said, in situations for house to be able to train those models for their with our own specific data,
we have the face a p I, which enables face actually Butte detection and recognition. In other words, within this image, what emotions are they display? I'll show you a demonstration of that.
We have finished like the farm. Recognize our in quick organizer like form. Recognize allows us to be able to extract text and key value ps from tables and from document. Then we have a speech. AP ICE, which includes FBI's like speech service, is like pitched text. Another what? Transcribing, audible speech into available and searchable text
or text to speech or even speech. Translation. Being able to integrate view times
peach translation into applications on that also includes speech recognition. AP. I've been able to identify and verify the people that are speaking based on the hardier and this kind of neighbor situations, like being able to use voice recognition for authentication. Then we have our language AP heists,
and within that we have the language understanding a P I, which allows us to be able to hide in the fire user intense in their own words.
We have the cure and a maker, which allows Austra created conversational question and answer layer over data. For example, we can't on existing knowledge based articles into Bart's. The users can ask questions. We have the text analytics ap High, which allows us to analyze provider text for scenarios like sentiment analyses.
In other words,
is it positive or the negative boys in neutral face Like Kiefer, his extraction on language detection is the English is a Chinese is another language we have. The such a P eyes on such a pea eye's include AP. Eyes like the being into the safe with helps users to be able to complete queries faster. But I had an intelligent
type. I had capabilities, in other words,
being able to recognize and classified named entities and find such results based on them.
We have the being gotta say J P High, which allows us to be able to send a special search, quit being on guest, get a list off suggested queries. In other what Otto suggestions we have the big Web search with enables us to be able to implement save on ad free location aware, such for our uses
on Han in Han. We have all this different options,
the one I have on the rights there, which is called Labs Awesome in some cases refers to as research. This is where we can get tau experience, upcoming AP eyes that the makes off with such team are working on and you can see different projects. They're like projects. Guess Chur Project entered the Lincoln This a
different capabilities that we can start
plane around wheat before they're released.
Let's have a quick look at how the agile competitive service's works, how we can start making use of this. This first win that one needs to do is we need to create on as your cognitive service. We creates that for the azure pato. When we create this, we get an AP ikey that's available to us. Get a primary key, and we get a second Ricky that we can use interchangeably.
It also exposes all these different AP I hand points to us,
then within our code or within any clients that can make an http arrest a p I call
we can provide. You are well on the endpoint key
for our cognitive service and then we can make an A P I call providing those information
which would then be processed, and then we get a response back. It's a simple is that it's a simple as an AP Aiko. Let's review some design considerations before we hand this lesson.
The first design consideration. I want you to know about it around the number off cause that we can make a second. There is a limit to the number of calls that we can make a second on. That depends on the price interior that were selected.
We have the free tier which allows us to be able to make 20 calls for me needs on 5000 cost a month for free.
But then we have the SS civility, which allows us to able to make 10 calls per second if we're gonna be starving, dying information or 20 cost per second. If we're gonna be starting the data that we ascended,
there is also a limit in the image size that we can upload to the service.
For example, we're gonna be doing prediction is a four megabyte limits for gonna be during training. There's a six megabytes limit, so the image that we can upload to this service.
Also, let's talk about continuation supports because, as your cognitive service is, actually supports the implementation off the same AP eyes that wages and in the azure cloud as containers that we could implement anywhere we want, including on premises
on what that Mrs did allows developers to be able to use the same reach FBI capabilities where they had an advantage off. They have the flexibility off where to deploy and lost the service's. And this can be very useful in situations when when you wants this capabilities. But you have, like data governance requirement
that prevents you from being able to upload your data to Asia.
Want information about disease that it's currently on? Lee supported for its upset off as your cognitive service's AP eyes, and that's the ones that you're seeing on your screen there
on then. Finally, there's virtual network integration capabilities, which is currently in preview.
Onda also has limited support for its upset off AP highs in a few regions at the moment because it's it's in preview, but one that went once it's released, it's going to allow us to be able to
implement as your cognitive service's with virtual network design on be ableto limit cause of them made over the Internet. It brings me to the hand off this lesson. Thanks very much for watching, and I'll see you in the next lesson.