AI-3018: Fundamentals of Generative AI – Complete Beginner's Guide!

0:01[Music]

0:08hey friends good morning good afternoon

0:10or good evening thank you so much for

0:12joining us today today we are going to

0:14focus on fundamentals of generative AI

0:17the agenda of this particular video is

0:19going to goes like this first we are

0:21going to talk about what is generative

0:22Ai and then after that we'll talk about

0:25different language models which are

0:26available we'll talk about large

0:28language models small language models

0:30also we are going to talk about

0:32co-pilots and AI agents and how we can

0:35build or customize them we'll also talk

0:37about adopting generative AI in your

0:39business so what kind of benefits you

0:41will get if you adopt generative AI in

0:43your business we'll talk about Microsoft

0:46copilot which are available in almost

0:48every product of Microsoft nowadays and

0:50then we'll talk about considerations for

0:52prompt how we can exactly use generative

0:54AI for generating desired content using

0:58different techniques of prompt

0:59engineering

1:00and then after that last but not the

1:02least we'll talk about extending and

1:04developing generative AI apps so let's

1:07get started first of all what is

1:10generative AI if that is your question

1:12let me start with what is AI artificial

1:15intelligence definition says that AI is

1:17going to imitate human behavior by using

1:20machine learning to interact with

1:22environment and execute task without

1:24explicit directions on what to Output

1:27basically that's going to imitate a

1:29human kind of behavior with your

1:31software applications now this kind of

1:34AI enable software application will

1:36imitate human behavior so that you can

1:38feel like they're able to understand

1:40what kind of an input you're providing

1:42on the other hand when we talk about

1:44generative AI generative AI is basically

1:47a subset of AI in which an AI model is

1:51going to create original content in the

1:53response of your natural language prompt

1:56so basically user will provide some kind

1:58of a natural language input

2:00by either saying something or typing

2:02something and then based on that the

2:04generated result is going to be an

2:07original content it's not something

2:09which is coming from some existing data

2:11and search it's not something which is

2:13copied from somewhere it's a newly

2:15generated original content and that's

2:17what generative AI is all about now

2:19generative AI application is going to

2:22take this kind of natural language input

2:24and it's going to return an appropriate

2:27response in variety of formats as you

2:29can see in this Slide the formats are

2:31actually three different formats natural

2:33language Generation image generation and

2:36code generation basically natural

2:38language generation is going to give you

2:40a response in a human specific language

2:43and in this case you might submit a

2:45prompt like give me three ideas for a

2:48healthy breakfast or maybe you can ask

2:51for write a cover letter for my resume

2:53and when you give this kind of an input

2:55prompt it's going to give you a response

2:57in the natural language text on the

2:59other hand when you're using things like

3:01image generation you're going to give a

3:03prompt create an image of an elephant

3:06eating a burger now when you do this

3:08thing it's going to generate a new image

3:10with the same kind of requirement or

3:12maybe you can try this one which is

3:14create a logo for a florist business and

3:16it's going to give you a logo associated

3:18with

3:19that third but important one code

3:21generation is also the third thing which

3:23is associated with generative AI in

3:26which you can actually generate code if

3:28you a developer in this case your AI

3:31applications are designed to help

3:33software developers to write a code for

3:35example you could submit a request like

3:38show me how can I code a game of a Tic

3:40Tac Toe with python or maybe you can

3:43just put a python code to add two

3:45numbers kind of a prompt and it's going

3:46to generate a logic associated with that

3:49remember it can generate logical code

3:51with multiple languages like python

3:53JavaScript C and few more but if you

3:57want to understand generate evi properly

3:59the next thing which you need to

4:00understand is language models now while

4:03the mathematical principles behind

4:05language models can be very complex a

4:07basic understanding of the architecture

4:09which is used to implement them can help

4:12you to gain a conceptual understanding

4:14of how they work now in today's Cutting

4:17Edge technology large language models

4:19are based on Transformer architecture

4:21this is one of that famous architecture

4:24based on which you will find all the

4:26latest language models like gpt3 GPT 4

4:29or some other models from other

4:31companies like meta and Microsoft this

4:34kind of models are Transformer models

4:36and these Transformer models are trained

4:38with large volume of text which are

4:40enabling them to represent the semantic

4:42relationship between words and use those

4:45relationship to predict probable

4:47sequence of the text to make sense about

4:51that hey guys sorry for Interruption my

4:54name is maruti and I'm here to make an

4:56very important announcement I hope you

4:58are liking our video and you're doing a

5:00continuous learning with us on an Azure

5:03cloud and AI related topics if you are

5:05enjoying this thing I'm going to

5:07announce skill tech. club which is our

5:09upcoming website which is going to be

5:11launched very soon we are here to tell

5:13you one thing that everyone who is a

5:16subscriber of this particular channel

5:17will get aure cloud and aure AI related

5:20certification courses free of cost in

5:23Skil tech. Club so you will be a part of

5:25the Skil tech. Club kind of a membership

5:28automatically free of cost

5:30and everyone who's a subscriber of this

5:32particular channel will get those

5:33benefits which are available in that so

5:36what are you waiting for I request you

5:37to please subscribe to this Channel and

5:39share it with your friends and families

5:41if they are also interested in Azure

5:43cloud and AI learning that's it from my

5:46side now you can carry on with your

5:48learning thank you so if you're

5:50interested in learning Transformer model

5:52in depth you can just comment down in

5:54this particular video and we will create

5:56a separate video for you on that but as

5:58of now this slide is actually trying to

6:00show you that in the left side first

6:02section you're going to provide your

6:04training text now remember this is a

6:06training text which is a text based raw

6:08content you're going to provide that

6:10thing and Transformer model is going to

6:12first take it in the encoder Bo this

6:15encoder block is actually going to

6:17encode the text is going to convert the

6:19text into tokens and then in the vector

6:22and once this tokens and vectors are

6:24generated with that is going to

6:26understand what kind of an input prompt

6:28you have provided the generation of the

6:30new text content is going to happen in

6:32the second block which is known as

6:34decoder Block in the decoder block they

6:36are going to understand the relationship

6:38between each word with the other words

6:40in that particular statement now

6:42obviously this is a complex process but

6:45this incoder and decoder blocks are

6:47actually going to have multiple

6:48attention blocks inside which this

6:50encoder block and decoder block are

6:52going to have multiple attention blocks

6:54in between which will actually help this

6:57process in the stepbystep way at the end

6:59of the decoder block you're going to get

7:01your output which is generated based on

7:03your input which you have provided now

7:05this diagram is just trying to give you

7:07a very very high level bird's eye view

7:10of how exactly Transformer model works

7:13but in order to understand this thing in

7:15depth you can search for Transformer

7:17model architecture and that actual

7:19architecture is actually going to take

7:20few hours to understand that in depth

7:23but yes it's going to be worth of your

7:24time it's very important if you are a

7:27data scientist then you understand a

7:29Transformer architecture if you're not a

7:31data scientist you're just a business

7:33user or a developer maybe this won't be

7:35that much useful for you so you can just

7:37focus on generative AI applications more

7:40than anything else now let's talk about

7:42some language models which are available

7:44in today's time the first thing which we

7:46want to discuss right now is foundation

7:48models remember organizations and

7:51developers can train their own large

7:53language models from scratch but in most

7:56cases it's very practical thing to use

7:59an existing Foundation model and then

8:02optionally if you want you can fine-tune

8:04those Foundation models with your own

8:06training data when you do this thing

8:08you're not only going to save a lot of

8:10time you're also going to save a lot of

8:12cost which is involved in creating large

8:15language models on Microsoft Azure and

8:18the Azure openi service which is

8:20available inside that it's actually

8:21including curated set of models from

8:24open AI which are hosted on Microsoft

8:26Azure Cloud this is actually going to

8:28offer a benefit of cutting age language

8:31models like generative pre-trained

8:33Transformer models like GPD models what

8:35we call and you also have some kind of

8:38image generation models like d e which

8:41is helping you to generate images with

8:42that now all these models are actually

8:45available in the model catalog of your

8:47Azure open AI service and your Azure AI

8:50Foundry portal is also having this if

8:52you don't know about Azure AI Foundry or

8:54Azure open AI then stay tuned we are

8:57going to learn this thing in this

8:58particular course after some

9:00modules then stay tuned we are going to

9:02learn this kind of models and this kind

9:05of portals very soon in this particular

9:07course if you ask me what kind of models

9:10are available as of now in the model

9:11catalog of the services well some of the

9:14names are mentioned on this slide you

9:15can found Microsoft models openi models

9:18in that hugging phas models mistal even

9:21meta models are also available in this

9:23the very recently some of the new models

9:26like you have deep seek models are also

9:28available in the model catalog of azure

9:30AI Foundry portal so this is really cool

9:33thing and it's going to take a lot of

9:35time once you decide which model you

9:37want to go on with you can fine tune you

9:39can customize and you can use it now

9:42let's talk about a little in depth about

9:44large language models and small language

9:47models now as the name suggest there are

9:49many language models which are available

9:51that you can use to power generative AI

9:54applications and in general we always

9:56have two different differentiations

9:58large language models and small language

10:01models as the name suggests large

10:03language models are trained with the

10:05vast quantity of text that represents

10:07wide range of subject matter related

10:09data now in this data you are actually

10:11going to have billions and even

10:13trillions of parameters inside that on

10:15which those large language models are

10:17trained basically if your model is

10:19trained with the higher number of

10:21parameters and more amount of data then

10:23it means that your large language model

10:26is more capable compared to your small

10:28language model but but processing that

10:30particular data and then generating the

10:32responsive text from that is also going

10:35to be more expensive because it has to

10:37go through that whole bunch of data

10:39whenever you are going to use

10:40fine-tuning or other customizations with

10:42the models so basically large language

10:44models will be more capable more

10:47accurate but it's also going to be more

10:50expensive on the other hand when you're

10:52focusing on small language models these

10:54are the train with more smaller or some

10:57subject Focus data sets it's not going

10:58to have a wide variety of data sets and

11:01parameters on this but it's going to be

11:03focusing on some particular focused

11:05subject on that this focused vocabulary

11:08is going to make them very effective in

11:09a specific conversational topic but it's

11:12going to make them less effective at

11:14more General language discussions so

11:17basically this is going to be something

11:18like that small language models will be

11:20expert of a specific Focus topic but not

11:23about some general discussion kind of a

11:25thing the smaller size of slms are going

11:28to provide more options for deployment

11:31including local deployments to devices

11:33and even on on premise computer and that

11:36is going to make them much faster and

11:38easier to fine tune compared to large

11:40language models well if you ask me right

11:42now which one is better well do not come

11:45to this conclusion that this is better

11:46for me because in your case depends upon

11:49your project depends upon your client

11:51and depends upon your company you maybe

11:53have to choose small language models or

11:56large language models and based on that

11:58you have to decide how you're going to

11:59use them how you're going to fine-tune

12:01them or where exactly you're going to

12:03deploy them now let's talk about

12:05co-pilot and AI agent now generative AI

12:09apps are often integrated into

12:11applications as a chat interface and

12:13this is something which most people know

12:15because of the famous application called

12:17Chad GPD they always provide contextual

12:20support for common task in those

12:22applications you ask a question and it's

12:24going to give you a response Microsoft

12:27copilot is also a generative AI based

12:29app that is integrated into a wide range

12:31of Microsoft products if you are a

12:34business users then business users can

12:36use generative eii to boost their

12:38productivity and creativity with AI

12:41generated content and automation of task

12:44developers can also extend Microsoft

12:46co-pilot by integrating them into their

12:49business process and data and even they

12:51can create a co-pilot like custom agents

12:54into incorporate a generative AI

12:56capabilities into their own applications

12:59and Services which they are developing

13:01this slide is actually showing you

13:02Microsoft co-pilot which is enabling you

13:05to summon a generative eii chat app

13:08where you are working on a Windows or a

13:10Microsoft 365 application like Microsoft

13:13Outlook or Microsoft Word and then using

13:15generative AI you will be able to chat

13:18with this pop alert and you can ask

13:20questions and you can generate related

13:22content with that now if you have a

13:24question that where exactly geni can

13:26help me in businesses well the answer

13:29answer is there are three levels of

13:30generative AI adoption in organizations

13:33the first one is you can use

13:35off-the-shelf generative AI apps like

13:38Microsoft 365 copilot to empower your

13:40users and increase their productivity

13:43this is one of the way by which most

13:45organizations are going to use

13:46generative AI the second one is you can

13:49actually extend Microsoft copilot to

13:51support custom business processes or

13:53task basically in this case you are

13:56going to use your own data to control

13:59how your co-pilot is going to respond

14:01and you can also use your own customized

14:04user prompts in the organization

14:06basically organizations can add their

14:08own data and they can control the

14:10behavior of the Microsoft co-pilot with

14:12this kind of extension and the third and

14:14the final usage of this is you can build

14:17your own co-pilot like agents to

14:19integrate generative AI into a business

14:22apps or to create a unique experience

14:24for your customers so think about you

14:26are a developer who's already developing

14:28application for your clients and

14:29customers you can integrate generative

14:32AI into those custom applications using

14:35your own co-pilot like agent and then

14:37you can provide generative AI facilities

14:39into your own applications as well now

14:41let's talk about different Microsoft

14:43co-pilots which are available in various

14:45Microsoft products now first let's talk

14:47about Microsoft co-pilot which is

14:49available at the URL copilot

14:52microsoft.com this is going to provide

14:54Microsoft co-pilot's home on the web

14:56browser page you can go there you can

14:59ask questions you can generate content

15:01such as text and

15:02images next you can go for Microsoft

15:05co-pilot which is integrated into your

15:07Bing search engine while searching you

15:10can use bing.com and then you can just

15:13use a co-pilot which is available in the

15:15chat interface of your Bing search

15:17engine when you do this thing is going

15:19to help you in getting very specific

15:21into your search results and the task

15:23which you want to achieve with that when

15:26you browse with Microsoft age browser

15:28they're also pilot is available on the

15:30right top corner you can just click on

15:32that and you can get co-pilot pan in

15:34your Microsoft age browser this is going

15:37to help you to research a specific topic

15:40it can help you to generate a new

15:42content for example you can maybe

15:44publish a blog post using that while all

15:46of these co-pilots options are signing

15:48in working with the work or a school

15:51account is enable you to use co-pilot in

15:53the context of your organization data

15:55and services with all of these co-pilot

15:58options you you have an option to

15:59signing in with your work and school

16:01account which basically allows you to

16:03incorporate your work and school related

16:07information data with your co-pilots and

16:09then it can work very well with your own

16:11configurational organizational data also

16:14next we have Microsoft 365 co-pilot this

16:17co-pilot can be an AI assistant for your

16:20information workers Microsoft 365

16:23co-pilot integrates co-pilot into a

16:25productivity applications that

16:27information workers use every day for

16:29example you can use calot in Microsoft

16:32Word to generate a new document based on

16:34the natural language prompt you can also

16:37refine summarize and improve the

16:39document with the few prompts you can

16:41use co-pilot same way with Microsoft

16:43PowerPoint also where you can generate

16:45presentations based on the content of a

16:47document or an email you can add

16:50graphics you can reform slides or maybe

16:53you can improve your presentation with

16:55some of the edit animations with that

16:57you can also use micro of Outlook

17:00co-pilot where it can help you to

17:02summarize your email threads it can help

17:04you to configure your schedule and you

17:07can even find relevant emails and

17:09documents to prepare for a specific

17:12meetings basically all these tools are

17:14actually going to help you to save your

17:16lot of time and giving you very specific

17:19Insight which you are looking for a

17:21particular meeting or a document or a

17:23client next we have co-pilot in

17:26Microsoft Dynamics 365 now maybe if you

17:29have never used Microsoft Dynamics then

17:31you don't know how exactly this is going

17:32to work but in this case also Microsoft

17:35Dynamics 365 is actually a suit of

17:37business tools that is helping your

17:39users into a specific role to perform a

17:42business processes co-pilot in Microsoft

17:45Dynamics 365 is going to provide

17:48contextual assistance into these tools

17:50and it's going to help users to be more

17:53efficient and effective for example

17:56let's say you are a sales professional

17:58and you can use cop aler to quickly find

18:01relevant customer and Industry

18:03information by integrating with the

18:05company's CRM when you do this thing

18:08your CRM database and some other data

18:11which you can associate with calot can

18:13help you to get the desired data from

18:15that even in the other case let's say

18:17customer service agents can use co-pilot

18:19in Dynamic 365 for a customer service to

18:22analyze support ticket or maybe to

18:25research similar issues or maybe to find

18:27a resolution of a specific issue which

18:30is happening on a day-to-day basis they

18:32can communicate with the end users with

18:34just a few clicks and prompts which are

18:36available in Microsoft co-pilot if you

18:39are someone who is already comfortable

18:40with Microsoft Dynamics 365 then you can

18:43check out how you can use generative AI

18:46with this by following this particular

18:48link next but important one which I use

18:50on a daily basis which is co-pilot for

18:53Microsoft Azure Cloud I hope you all

18:55know that Azure cloud is available on

18:57portal. azure.com

18:59this Azure portal is basically allowing

19:01you to create everything in Azure Cloud

19:04when you're using co-pilot with Azure

19:05portal it can actually assist you in the

19:08infrastructure Administration related

19:10work basically that works with all your

19:13cloud services and it can help you to do

19:15infrastructure based deployment you are

19:17going to manage your it infrastructure

19:19with the help of co-pilot to check and

19:21learn more about aure Cloud specific

19:23co-pilot you can follow this particular

19:25link and remember co-pilot is available

19:28inside asure your Cloud as a separate

19:29product itself next we can use Microsoft

19:32co-pilot for security also which will

19:34provide an assistant for Security

19:36Professionals as they are going to

19:38assess mitigate and respond to security

19:41threats which are happening in that the

19:43screenshot is actually showing you how

19:45we can use cop aler in Microsoft 365

19:48Defender when you do this thing the

19:50defender for the cloud and Defender for

19:52your organization is easily able to

19:54integrate with your copilot and you can

19:56dig deep into the specific Security One

19:58lities and issues with the help of

20:00co-pilot props not only that if you are

20:03someone who's going to analyze data and

20:06you want to use AI in your data

20:08analytics you can do that also your data

20:10analyst need to work with the code and

20:12visualization Tools in order to analyze

20:14data and Report inside in this case your

20:17Microsoft fabric is going to have a

20:19coiler associated with that which

20:21enables you to analyze your data

20:24automatically and it's going to generate

20:26the code that needs to analyze manipul

20:29and visualize the data in the spark

20:30notebook so as you can see in this first

20:33screenshot we are actually having a

20:35notebook which is available inside

20:37Microsoft fabric that notebook is

20:39generated with the help of python code

20:41and this code is generated with the help

20:42of copilot so basically as a data

20:45analyst you can quickly generate this

20:47kind of a notebook code all you have to

20:49do is run the notebook analyze the data

20:51see the data frames or the charts or

20:53dashboards which are generated and

20:55associated with that next if you are are

20:58dealing with powerbi and you want to

21:01manage powerbi dashboards reports

21:03co-pilot can actually analyze your data

21:06and then they can suggest and create

21:08appropriate data visualizations from it

21:10and that is what which is exactly

21:12visible in the second screenshot last

21:14but not the least we have GitHub cop

21:16aler which is one of my favorite

21:18undoubtedly because this is going to

21:20help you to generate code yes GitHub is

21:23the world's most popular place for

21:25developers to manage their code and the

21:27repositor to develop applications and to

21:30collaborate with their team members

21:32undoubtedly GitHub has the largest

21:34community of developers associated with

21:37each other using GitHub copilot these

21:39developers are going to maximize their

21:41productivity by analyzing and explaining

21:44code by adding code documentation with

21:46EAS by generating new code based on the

21:49natural language prompt by refactoring

21:52and optimizing code by generating test

21:54cases for their existing functions and

21:58by in integrating into their existing

22:00development tools like Microsoft Visual

22:02Studio code and many more if you are

22:05interested in developing applications

22:07with the help of GitHub copilot I

22:09strongly recommend you to check out this

22:11link this link is actually giving you

22:13step by-step process of GitHub co-pilot

22:16with a lot many tutorial Labs with that

22:19now after all this variations of

22:20multiple Microsoft co-pilots let's talk

22:23about considerations for prompt because

22:25ultimately when you are going to provide

22:27a proper prompt then only you're going

22:30to generate desired result so while some

22:33generative AI applications are going to

22:35provide buttons and some kind of a

22:36visual tools to interact with the

22:38language model most of the time you have

22:41to depend on this prompt which you are

22:43going to enter into the chat bot while

22:46typing or maybe by speaking and then if

22:49your prompt is proper then only you're

22:51going to get desired result now we know

22:53that our large language models are very

22:55smart enough to understand our inputs

22:58but

22:59there are some common prompting

23:00techniques which you can apply to get

23:02the best out of your generative AI app

23:05and some of the techniques are actually

23:06mentioned here you can see that in this

23:08slide there is one prompt which is

23:10showing you that summarize the key

23:12considerations for adopting co-pilot now

23:15with this there is a number one which is

23:17mentioned in this prompt which is

23:18telling you that we have just followed

23:21the first guideline which is start with

23:23the specific goal for what you want the

23:25generative AI app to do then we are

23:28going to the second one which is provide

23:30a source to ground the response in the

23:32specific scope of information where we

23:35are saying that describe in this

23:37document so basically you are mentioning

23:39that in this document we want to

23:40describe that thing then we are going

23:42with the third one add a context to

23:44maximize response appropriateness and

23:46relevance in which we are saying for a

23:48corporate executive so we are basically

23:50strictly saying that we want this thing

23:52for corporate Executives only that's the

23:54context which we are specifying then we

23:56have set a clear expectation for the

23:58response in which we are seeing that

24:00format the summary as no more than six

24:03bullet points with a professional tone

24:06so basically you are specifying this

24:08fourth point that I want six bulleted

24:10points only and then finally we have an

24:13iterate based on previous prompts and

24:15responses to refine the result now this

24:18is a fifth one which we have given in

24:20the right side section where we are

24:22going to specify in the cop palot so in

24:24the calot you can actually have system

24:26message configuration conversation

24:28history which is maintained with that

24:30and your current prompt is going to be

24:32keep adding new new prompts inside that

24:34now when you're doing this thing at that

24:36time only you'll be able to understand

24:38this for you guys if you have never used

24:41co-pilot or if you have never used

24:44prompt engineering techniques I strongly

24:46recommend you to check out my videos

24:48which are mentioned in the description

24:50of this particular one we have a

24:51dedicated video on prompt engineering

24:53techniques and we have a separate video

24:55which is showing you how you can use

24:57Microsoft cop Al how you can use

24:59Microsoft copot in the age browser I

25:02surely recommend you to check out those

25:04two videos without fail now let's talk

25:06about the last and the final topic of

25:08this particular module which is

25:10extending and developing generative AI

25:13apps now obviously llms are capable to

25:15generate a generative content based on

25:17my prompt and if I want I can do

25:20fine-tuning and configurations with my

25:23own data also inside that but still

25:25there are many cases where you decide

25:27for your organiz ation you're going to

25:29develop your own co-pilot kind of an

25:31agent when you decide that we want to

25:34develop a generative AI apps you have

25:37two different options you can use

25:39co-pilot Studio or you can use Azure AI

25:42Foundry now this slide is actually

25:45showing you in which cases you're going

25:47to use co-pilot studio if you want a low

25:50code development tool for creating

25:53agents using power automate and you want

25:55to extend Microsoft 365 co-pilot then

25:58you're going to go for copilot studio if

26:00you want fully managed hosted software

26:03as a service kind of a SAS model then

26:05you will go with this you can have a

26:08dialogue and conversational

26:09orchestration so basically the flow of

26:12that particular dialogue conversation

26:13you can control you're going to have

26:16built-in analytics with security and

26:18governance control which is a part of

26:20the copala studio integrated with that

26:23and then you can deploy to Common chat

26:25channels like web apps social channels

26:29and teams and this can be associated

26:31with that now if your work is limited to

26:33this kind of features then we strongly

26:35suggest you go with the co-pilot Studio

26:37specifically if you're not a very

26:39hardcore developer using a programming

26:42languages then you are going to use this

26:43one on the other hand if you say no I

26:47want to develop generative AI

26:49applications I'm a person who is a pro

26:51code and I can develop any kind of

26:53logical code with my programming

26:55language knowledge then Azure AI Foundry

26:57is your thing this is a development

27:00environment which is a proc code

27:02development platform with full catalog

27:04of models and fine-tuning capabilities

27:07remember whatever model catalog which we

27:09have mentioned in the beginning of this

27:11particular video all are available this

27:14is not a software as a service it's a

27:16platform as a service with full control

27:18over Cloud infrastructure so obviously

27:21compared to copilot Studio this is going

27:23to be more expensive but it's going to

27:25give you more control over whatever

27:27deployment you want to do with that your

27:30prompt and model orchestration will be

27:32there you also have an evaluation engine

27:34to taste the performance reliability

27:37scalability and responsible AI safety

27:40with content filters all these

27:42additional features which you will get

27:43in Azure AI Foundry portal will not be

27:46available in your Microsoft copala

27:48Studio you can also deploy as an

27:50endpoint in Azure for use in custom apps

27:53and services basically that's going to

27:56make your deployed model as an API and

27:59anyone who knows how to consume API can

28:02actually make a request to your deployed

28:04model for your kind information there

28:07are so many videos on Azure AI Foundry

28:09Labs which are available on our YouTube

28:11channel so I strongly recommend you to

28:13check out that one of the video which I

28:16am mentioning right now in this side is

28:18very important for you to start with

28:20Azure AI Foundry portal with practical

28:23Labs step by step that's it for today's

28:25video I want to thank you for being with

28:27me till the end of this video and I have

28:30a request out of all this Microsoft

28:32co-pilots can you please comment which

28:34one is your favorite in the comment box

28:36of this particular video I will make

28:38sure that we will create a separate

28:40video on that particular Microsoft

28:42co-pilot and we'll tag you in that video

28:44so that everyone knows that you

28:45requested this I'll see you soon

28:47tomorrow this is maruti signing off

28:50thank you happy learning

28:53[Music]