Up next in 10
π₯ What is Generative AI? Everything You Need to Know.
Welcome to our deep dive into Generative AI β the powerful technology behind tools like ChatGPT, Midjourney, and more! In this video, we break down everything you need to know, even if you're new to AI.
π Agenda:
π What is Generative AI and how it differs from traditional AI
π How Generative AI actually works behind the scenes
π The role of data in training and powering these models
π What are Large Language Models (LLMs) like GPT, and why they're important
π How prompts impact the quality and accuracy of AI responses
π A look at popular tools and models in the AI space
Key pros and cons of using Generative AI in real life
Whether you're a beginner, a student, or just curious about the future of AI, this video is your complete guide.
πΊ CSharp TV - Dev Streaming Destination http://csharp.tv
π C# Corner - Community of Software and Data Developers https://www.c-sharpcorner.com
#CSharpTV #csharpcorner
Show More Show Less View Video Transcript
0:02
Hello and welcome to another episode of
0:04
Hello and welcome to another episode of
0:04
Hello and welcome to another episode of Shar TV. CV brings you latest tech
0:08
Shar TV. CV brings you latest tech
0:08
Shar TV. CV brings you latest tech updates, news, community
0:13
updates, news, community
0:13
updates, news, community updates. My name is Mahes Chan. I'm
0:15
updates. My name is Mahes Chan. I'm
0:15
updates. My name is Mahes Chan. I'm founder of Corner.
0:18
founder of Corner.
0:18
founder of Corner. Today we are going to talk about
0:20
Today we are going to talk about
0:20
Today we are going to talk about generative AI. I'm sure most of you are
0:23
generative AI. I'm sure most of you are
0:23
generative AI. I'm sure most of you are familiar with the term generative AI but
0:26
familiar with the term generative AI but
0:26
familiar with the term generative AI but today I'm going to discuss more get into
0:29
today I'm going to discuss more get into
0:29
today I'm going to discuss more get into more details what generative AI is all
0:31
more details what generative AI is all
0:31
more details what generative AI is all about.
0:34
Sarak welcome to the show. Rohit welcome
0:36
Sarak welcome to the show. Rohit welcome
0:36
Sarak welcome to the show. Rohit welcome to S Js JSI Pass welcome to the show.
0:40
to S Js JSI Pass welcome to the show.
0:40
to S Js JSI Pass welcome to the show. Thank you everybody for joining.
0:44
Thank you everybody for joining.
0:44
Thank you everybody for joining. Some of you are joining in the evening.
0:45
Some of you are joining in the evening.
0:46
Some of you are joining in the evening. So good evening to you. Some joining in
0:47
So good evening to you. Some joining in
0:48
So good evening to you. Some joining in the US good morning to you. It's 11:00
0:50
the US good morning to you. It's 11:00
0:50
the US good morning to you. It's 11:00 a.m. Saturday.
0:52
a.m. Saturday.
0:52
a.m. Saturday. I'm here from Philadelphia, outside of
0:54
I'm here from Philadelphia, outside of
0:54
I'm here from Philadelphia, outside of Philadelphia. Nice weather. Summer is
0:56
Philadelphia. Nice weather. Summer is
0:56
Philadelphia. Nice weather. Summer is here, by the way. So, Philadelphia
0:59
here, by the way. So, Philadelphia
0:59
here, by the way. So, Philadelphia summertime.
1:03
Looks like we got a good bunch of people
1:04
Looks like we got a good bunch of people
1:04
Looks like we got a good bunch of people joining us. Very exciting stuff. Genai
1:07
joining us. Very exciting stuff. Genai
1:07
joining us. Very exciting stuff. Genai is the new future. Generative AI. We are
1:10
is the new future. Generative AI. We are
1:10
is the new future. Generative AI. We are going to talk today. every Saturday
1:14
going to talk today. every Saturday
1:14
going to talk today. every Saturday um I'm going to bring this new show
1:17
um I'm going to bring this new show
1:17
um I'm going to bring this new show focused around AI geni prompts anything
1:22
focused around AI geni prompts anything
1:22
focused around AI geni prompts anything else around those topics.
1:26
else around those topics.
1:26
else around those topics. So today's our first show on this. So
1:28
So today's our first show on this. So
1:28
So today's our first show on this. So feel free to ask your questions, chime
1:30
feel free to ask your questions, chime
1:30
feel free to ask your questions, chime in. If you have any feedback, feel free
1:34
in. If you have any feedback, feel free
1:34
in. If you have any feedback, feel free to share that. Share on your social
1:36
to share that. Share on your social
1:36
to share that. Share on your social media. Let people know if you like this
1:38
media. Let people know if you like this
1:38
media. Let people know if you like this show. If you learned something from it,
1:41
show. If you learned something from it,
1:42
show. If you learned something from it, let your friends, co-workers know about
1:43
let your friends, co-workers know about
1:44
let your friends, co-workers know about it.
1:48
So
1:51
So
1:51
So before I start the show, I would like to
1:53
before I start the show, I would like to
1:53
before I start the show, I would like to ask and if you can comment, post comment
1:55
ask and if you can comment, post comment
1:55
ask and if you can comment, post comment on this. How many of you are using Gen
1:57
on this. How many of you are using Gen
1:57
on this. How many of you are using Gen AI in your day-to-day
2:00
AI in your day-to-day
2:00
AI in your day-to-day task for your day-to-day tasks for
2:02
task for your day-to-day tasks for
2:02
task for your day-to-day tasks for example chat GPT? And I think everybody
2:04
example chat GPT? And I think everybody
2:04
example chat GPT? And I think everybody almost everybody using some kind of jai
2:08
almost everybody using some kind of jai
2:08
almost everybody using some kind of jai these days right chat GPT is popular
2:10
these days right chat GPT is popular
2:10
these days right chat GPT is popular obviously
2:12
obviously
2:12
obviously Gemini bunch of other tools. So if you
2:15
Gemini bunch of other tools. So if you
2:15
Gemini bunch of other tools. So if you are using regularly one of the tools
2:17
are using regularly one of the tools
2:17
are using regularly one of the tools please post that you are using it and
2:19
please post that you are using it and
2:19
please post that you are using it and which tool you are using it.
2:23
which tool you are using it.
2:23
which tool you are using it. So just post in the comment which tool
2:26
So just post in the comment which tool
2:26
So just post in the comment which tool you're using it. Um and also have your
2:29
you're using it. Um and also have your
2:29
you're using it. Um and also have your questions ready too because first we are
2:32
questions ready too because first we are
2:32
questions ready too because first we are going to talk about JNAI maybe about 15
2:35
going to talk about JNAI maybe about 15
2:35
going to talk about JNAI maybe about 15 minutes or so 15 20 minutes maybe less
2:38
minutes or so 15 20 minutes maybe less
2:38
minutes or so 15 20 minutes maybe less depends and then then we will take some
2:42
depends and then then we will take some
2:42
depends and then then we will take some questions.
2:43
questions.
2:43
questions. Amazing. Uh Ro
2:46
Amazing. Uh Ro
2:46
Amazing. Uh Ro Rud says voice AI. Very good. Amazing
2:49
Rud says voice AI. Very good. Amazing
2:49
Rud says voice AI. Very good. Amazing chats pretty Google AI. Okay great let's
2:51
chats pretty Google AI. Okay great let's
2:51
chats pretty Google AI. Okay great let's get started. So let's load my slides
2:55
get started. So let's load my slides
2:55
get started. So let's load my slides here.
2:59
All right. So, this today's Gen AI for
3:01
All right. So, this today's Gen AI for
3:01
All right. So, this today's Gen AI for everyone.
3:06
How do I make it full screen? Let me
3:08
How do I make it full screen? Let me
3:08
How do I make it full screen? Let me make it. There you go.
3:11
make it. There you go.
3:11
make it. There you go. All right. So, what we will cover today
3:14
All right. So, what we will cover today
3:14
All right. So, what we will cover today in this webinar, we will cover what is
3:17
in this webinar, we will cover what is
3:17
in this webinar, we will cover what is generative AI.
3:20
generative AI.
3:20
generative AI. We will talk about how it works.
3:24
Data and generative AI work together
3:26
Data and generative AI work together
3:26
Data and generative AI work together without data generative AI can't do
3:28
without data generative AI can't do
3:28
without data generative AI can't do anything and also LLMs we'll little bit
3:31
anything and also LLMs we'll little bit
3:31
anything and also LLMs we'll little bit little bit talk about data role of data
3:33
little bit talk about data role of data
3:33
little bit talk about data role of data and generative AI in LLMs
3:36
and generative AI in LLMs
3:36
and generative AI in LLMs I think Deepak this is a old this is not
3:39
I think Deepak this is a old this is not
3:39
I think Deepak this is a old this is not the latest slide I think I had another
3:42
the latest slide I think I had another
3:42
the latest slide I think I had another slide I uploaded
3:45
slide I uploaded
3:45
slide I uploaded um yeah
3:49
um yeah
3:49
um yeah it's fine we'll take from there I
3:50
it's fine we'll take from there I
3:50
it's fine we'll take from there I thought I had new slide nobody they talk
3:52
thought I had new slide nobody they talk
3:52
thought I had new slide nobody they talk about generative AI importance of
3:54
about generative AI importance of
3:54
about generative AI importance of prompts and models and pros and cons. So
3:57
prompts and models and pros and cons. So
3:57
prompts and models and pros and cons. So let's go through this right. Yeah, this
3:59
let's go through this right. Yeah, this
4:00
let's go through this right. Yeah, this is an old slide. So what is generative
4:03
is an old slide. So what is generative
4:03
is an old slide. So what is generative AI? As you know the name already says
4:06
AI? As you know the name already says
4:06
AI? As you know the name already says generative AI. Anything AI that
4:09
generative AI. Anything AI that
4:09
generative AI. Anything AI that generates, right? AI that generates
4:12
generates, right? AI that generates
4:12
generates, right? AI that generates generates any kind of content
4:15
generates any kind of content
4:15
generates any kind of content that's called generative AI. Generative
4:16
that's called generative AI. Generative
4:16
that's called generative AI. Generative AI is a part of AI. So when you look at
4:18
AI is a part of AI. So when you look at
4:18
AI is a part of AI. So when you look at AI there's AI is just a big word for all
4:21
AI there's AI is just a big word for all
4:21
AI there's AI is just a big word for all kind of artificial intelligence.
4:24
kind of artificial intelligence.
4:24
kind of artificial intelligence. Then AI has different you know it has
4:27
Then AI has different you know it has
4:27
Then AI has different you know it has bunch of machine learning is a part of
4:28
bunch of machine learning is a part of
4:28
bunch of machine learning is a part of it and deep learning is a part of it and
4:30
it and deep learning is a part of it and
4:30
it and deep learning is a part of it and generative AI is a small part of AI and
4:33
generative AI is a small part of AI and
4:34
generative AI is a small part of AI and today we are talking about generative
4:35
today we are talking about generative
4:35
today we are talking about generative AI. So when you think of these tools
4:38
AI. So when you think of these tools
4:38
AI. So when you think of these tools like chat, GPT, voice AI, JAI,
4:41
like chat, GPT, voice AI, JAI,
4:41
like chat, GPT, voice AI, JAI, midjourney, Sora,
4:44
midjourney, Sora,
4:44
midjourney, Sora, they all come under generative AI and
4:46
they all come under generative AI and
4:46
they all come under generative AI and there's hundreds of tools like that and
4:48
there's hundreds of tools like that and
4:48
there's hundreds of tools like that and some are general tools, some are
4:50
some are general tools, some are
4:50
some are general tools, some are specific tools.
4:52
specific tools.
4:52
specific tools. So when any tool is creating text, code,
4:56
So when any tool is creating text, code,
4:56
So when any tool is creating text, code, image, voice, videos, any kind of
4:58
image, voice, videos, any kind of
4:58
image, voice, videos, any kind of content using AI, that all comes under
5:02
content using AI, that all comes under
5:02
content using AI, that all comes under generative AI.
5:04
generative AI.
5:04
generative AI. So quick oneline definition what is
5:06
So quick oneline definition what is
5:06
So quick oneline definition what is generative AI? AI that generates new
5:09
generative AI? AI that generates new
5:09
generative AI? AI that generates new content
5:11
content
5:11
content that is generative AI.
5:14
that is generative AI.
5:14
that is generative AI. Uh I'm not going to say show any demos
5:17
Uh I'm not going to say show any demos
5:17
Uh I'm not going to say show any demos because you looks like you guys almost
5:19
because you looks like you guys almost
5:19
because you looks like you guys almost everyone is using generative AI anyway.
5:21
everyone is using generative AI anyway.
5:21
everyone is using generative AI anyway. So I'm not going to go and show you how
5:24
So I'm not going to go and show you how
5:24
So I'm not going to go and show you how to do a write a prompt in chat GPT
5:28
to do a write a prompt in chat GPT
5:28
to do a write a prompt in chat GPT right? You already know that unless you
5:29
right? You already know that unless you
5:29
right? You already know that unless you have some specific questions feel free
5:31
have some specific questions feel free
5:31
have some specific questions feel free to ask those questions.
5:34
to ask those questions.
5:34
to ask those questions. So this episode is first episode it's
5:37
So this episode is first episode it's
5:37
So this episode is first episode it's going to be almost like basic then next
5:40
going to be almost like basic then next
5:40
going to be almost like basic then next episode we're going to be a little bit
5:41
episode we're going to be a little bit
5:41
episode we're going to be a little bit more advanced and we're going to
5:42
more advanced and we're going to
5:42
more advanced and we're going to continue to deep dive into more advanced
5:45
continue to deep dive into more advanced
5:45
continue to deep dive into more advanced topics as we go along with this.
5:49
topics as we go along with this.
5:49
topics as we go along with this. Looks like new people are joining us.
5:51
Looks like new people are joining us.
5:51
Looks like new people are joining us. Welcome to the show. Everybody joining
5:52
Welcome to the show. Everybody joining
5:52
Welcome to the show. Everybody joining us welcome to the show. Feel free to
5:54
us welcome to the show. Feel free to
5:54
us welcome to the show. Feel free to post your comments. Say hi to everybody.
5:58
post your comments. Say hi to everybody.
5:58
post your comments. Say hi to everybody. So how does generative AI work?
6:00
So how does generative AI work?
6:00
So how does generative AI work? generative way I work when you think of
6:02
generative way I work when you think of
6:02
generative way I work when you think of this and we will not spend too much time
6:04
this and we will not spend too much time
6:04
this and we will not spend too much time here because in my following episodes I
6:07
here because in my following episodes I
6:07
here because in my following episodes I will get into more depth. So everything
6:11
will get into more depth. So everything
6:11
will get into more depth. So everything start with the data in world of
6:13
start with the data in world of
6:13
start with the data in world of generative AI everything starts with the
6:15
generative AI everything starts with the
6:15
generative AI everything starts with the data. So these AI companies for example
6:19
data. So these AI companies for example
6:19
data. So these AI companies for example open AI if you're not familiar with open
6:22
open AI if you're not familiar with open
6:22
open AI if you're not familiar with open AI is the company that created chat GPT.
6:26
AI is the company that created chat GPT.
6:26
AI is the company that created chat GPT. So company like open AAI what they do
6:29
So company like open AAI what they do
6:29
So company like open AAI what they do and they did is they went to the
6:32
and they did is they went to the
6:32
and they did is they went to the internet and they collected all kind of
6:35
internet and they collected all kind of
6:35
internet and they collected all kind of data they almost copied all data. Okay.
6:37
data they almost copied all data. Okay.
6:37
data they almost copied all data. Okay. They copied from all kind of websites.
6:41
They copied from all kind of websites.
6:41
They copied from all kind of websites. They copied data from all textbooks.
6:44
They copied data from all textbooks.
6:44
They copied data from all textbooks. They copied data from all music right
6:47
They copied data from all music right
6:47
They copied data from all music right libraries. So all those millions of
6:50
libraries. So all those millions of
6:50
libraries. So all those millions of millions of books, images and videos
6:52
millions of books, images and videos
6:52
millions of books, images and videos they have copied. Uh funny thing is lot
6:55
they have copied. Uh funny thing is lot
6:55
they have copied. Uh funny thing is lot of them content they copied they have
6:58
of them content they copied they have
6:58
of them content they copied they have not even ask us to you know they just
7:00
not even ask us to you know they just
7:00
not even ask us to you know they just copy this almost like a store right.
7:04
copy this almost like a store right.
7:04
copy this almost like a store right. So
7:06
So
7:06
So they copied all this data and then what
7:08
they copied all this data and then what
7:08
they copied all this data and then what happens is then data is first cleansed.
7:13
happens is then data is first cleansed.
7:13
happens is then data is first cleansed. Cleans means you know some data comes in
7:15
Cleans means you know some data comes in
7:15
Cleans means you know some data comes in whatever format then they put in a right
7:18
whatever format then they put in a right
7:18
whatever format then they put in a right format
7:20
format
7:20
format and then they train these AI models
7:24
and then they train these AI models
7:24
and then they train these AI models from this data and it learns patterns.
7:28
from this data and it learns patterns.
7:28
from this data and it learns patterns. Patterns means if there's a book in book
7:30
Patterns means if there's a book in book
7:30
Patterns means if there's a book in book it says
7:32
it says
7:32
it says capital of
7:34
capital of
7:34
capital of Capital of India is New Delhi. Right? So
7:37
Capital of India is New Delhi. Right? So
7:37
Capital of India is New Delhi. Right? So that's a sentence.
7:39
that's a sentence.
7:39
that's a sentence. So there's another book saying capital
7:41
So there's another book saying capital
7:41
So there's another book saying capital of India is Delhi. There's online
7:44
of India is Delhi. There's online
7:44
of India is Delhi. There's online document says capital of India is Delhi.
7:47
document says capital of India is Delhi.
7:47
document says capital of India is Delhi. So now AI finds this pattern that every
7:50
So now AI finds this pattern that every
7:50
So now AI finds this pattern that every time there's capital of India, there's a
7:52
time there's capital of India, there's a
7:52
time there's capital of India, there's a capital of India is New Delhi. Right? So
7:56
capital of India is New Delhi. Right? So
7:56
capital of India is New Delhi. Right? So now you got a kind of almost like a
7:59
now you got a kind of almost like a
7:59
now you got a kind of almost like a fact. Yeah. Capital of India is New
8:01
fact. Yeah. Capital of India is New
8:01
fact. Yeah. Capital of India is New Delhi. So anyway, same thing applies to
8:04
Delhi. So anyway, same thing applies to
8:04
Delhi. So anyway, same thing applies to the to the images, same thing applies to
8:09
the to the images, same thing applies to
8:09
the to the images, same thing applies to the sound, all that. So this process
8:12
the sound, all that. So this process
8:12
the sound, all that. So this process goes on and you know cost millions of
8:14
goes on and you know cost millions of
8:14
goes on and you know cost millions of dollars and lot of supercomputers and so
8:16
dollars and lot of supercomputers and so
8:16
dollars and lot of supercomputers and so on so forth. So now the other part of
8:20
on so forth. So now the other part of
8:20
on so forth. So now the other part of JAI. So now we got the data, we got the
8:23
JAI. So now we got the data, we got the
8:23
JAI. So now we got the data, we got the patterns, we got these models which I
8:25
patterns, we got these models which I
8:25
patterns, we got these models which I will talk in a second. So when you go to
8:28
will talk in a second. So when you go to
8:28
will talk in a second. So when you go to chat GPT and says what is the capital of
8:32
chat GPT and says what is the capital of
8:32
chat GPT and says what is the capital of India that's called prompt
8:36
India that's called prompt
8:36
India that's called prompt the chat GPD sends this prompt to the
8:38
the chat GPD sends this prompt to the
8:38
the chat GPD sends this prompt to the open AI's library which behind the scene
8:41
open AI's library which behind the scene
8:41
open AI's library which behind the scene is the LLM's running and says oh capital
8:45
is the LLM's running and says oh capital
8:45
is the LLM's running and says oh capital came India came it must be New Delhi so
8:47
came India came it must be New Delhi so
8:47
came India came it must be New Delhi so it goes in reply New Delhi so that's
8:49
it goes in reply New Delhi so that's
8:49
it goes in reply New Delhi so that's like a simple example of simple example
8:53
like a simple example of simple example
8:53
like a simple example of simple example of
8:54
of
8:54
of how this ji works
8:57
how this ji works
8:57
how this ji works And then this gets complex and complex
8:59
And then this gets complex and complex
8:59
And then this gets complex and complex complex. We'll get a little bit more
9:01
complex. We'll get a little bit more
9:01
complex. We'll get a little bit more into as we go through this.
9:04
into as we go through this.
9:04
into as we go through this. So here's another example. Let's take
9:06
So here's another example. Let's take
9:06
So here's another example. Let's take one more example in details, right? So
9:08
one more example in details, right? So
9:08
one more example in details, right? So when DAL is the the model for generating
9:12
when DAL is the the model for generating
9:12
when DAL is the the model for generating images and it has you know trained on
9:15
images and it has you know trained on
9:15
images and it has you know trained on millions of images. So when it trens on
9:17
millions of images. So when it trens on
9:17
millions of images. So when it trens on millions of images when we copy an image
9:20
millions of images when we copy an image
9:20
millions of images when we copy an image it also has a description like what this
9:22
it also has a description like what this
9:22
it also has a description like what this image is. So somebody means data
9:25
image is. So somebody means data
9:25
image is. So somebody means data scientist and data engineers and data
9:27
scientist and data engineers and data
9:28
scientist and data engineers and data analyst have to
9:30
analyst have to
9:30
analyst have to take the image create content what this
9:32
take the image create content what this
9:32
take the image create content what this is all about what this image all about
9:34
is all about what this image all about
9:34
is all about what this image all about right you take a for example a horse
9:38
right you take a for example a horse
9:38
right you take a for example a horse um then you have to put all the
9:39
um then you have to put all the
9:39
um then you have to put all the description there somehow
9:42
description there somehow
9:42
description there somehow so
9:43
so
9:43
so so there is a dial is a basically a data
9:46
so there is a dial is a basically a data
9:46
so there is a dial is a basically a data set
9:48
set
9:48
set that that teach models about lot of
9:52
that that teach models about lot of
9:52
that that teach models about lot of shapes colors sizes text what they are
9:54
shapes colors sizes text what they are
9:54
shapes colors sizes text what they are all about more data better obviously so
9:58
all about more data better obviously so
9:58
all about more data better obviously so when you create an image sometime you
9:59
when you create an image sometime you
9:59
when you create an image sometime you will see that image you can create in
10:01
will see that image you can create in
10:01
will see that image you can create in co-pilot image you can create in Gemini
10:04
co-pilot image you can create in Gemini
10:04
co-pilot image you can create in Gemini image you can create in midjourney they
10:06
image you can create in midjourney they
10:06
image you can create in midjourney they will all generate different images their
10:08
will all generate different images their
10:08
will all generate different images their quality is different and quality of
10:11
quality is different and quality of
10:11
quality is different and quality of those images being generated is
10:13
those images being generated is
10:13
those images being generated is depending on the how much data they have
10:16
depending on the how much data they have
10:16
depending on the how much data they have draw t train t train t train t train t
10:16
train t train t train t train t train t trained them what kind of image quality
10:19
trained them what kind of image quality
10:19
trained them what kind of image quality may they're generating and so on so
10:20
may they're generating and so on so
10:20
may they're generating and so on so forth
10:22
forth
10:22
forth So think about this. So first thing is
10:24
So think about this. So first thing is
10:24
So think about this. So first thing is data right data is trained this then
10:28
data right data is trained this then
10:28
data right data is trained this then there's something called neural networks
10:30
there's something called neural networks
10:30
there's something called neural networks and transformers. So neural networks is
10:32
and transformers. So neural networks is
10:32
and transformers. So neural networks is a is there like algorithms created the
10:36
a is there like algorithms created the
10:36
a is there like algorithms created the way our human mind works right our human
10:38
way our human mind works right our human
10:38
way our human mind works right our human minds how it works. So as a child we
10:41
minds how it works. So as a child we
10:41
minds how it works. So as a child we look at something and you know first
10:43
look at something and you know first
10:43
look at something and you know first thing we look at is mom dad right? So
10:46
thing we look at is mom dad right? So
10:46
thing we look at is mom dad right? So somebody's teaching a child a baby start
10:48
somebody's teaching a child a baby start
10:48
somebody's teaching a child a baby start talking and parent will say oh mom oh
10:53
talking and parent will say oh mom oh
10:53
talking and parent will say oh mom oh dad oh dog. So when baby every time in a
10:57
dad oh dog. So when baby every time in a
10:57
dad oh dog. So when baby every time in a baby's brain baby's taking the picture
10:59
baby's brain baby's taking the picture
11:00
baby's brain baby's taking the picture of a dog putting in the brain and saying
11:02
of a dog putting in the brain and saying
11:02
of a dog putting in the brain and saying oh this looks like this shape it's
11:05
oh this looks like this shape it's
11:05
oh this looks like this shape it's moving it's a dog
11:07
moving it's a dog
11:07
moving it's a dog oh this person is mom this person is
11:10
oh this person is mom this person is
11:10
oh this person is mom this person is dad. So this is how
11:14
dad. So this is how
11:14
dad. So this is how these AI models are trained
11:17
these AI models are trained
11:17
these AI models are trained and then there is next thing is called
11:20
and then there is next thing is called
11:20
and then there is next thing is called transformers. Transformers are the one
11:22
transformers. Transformers are the one
11:22
transformers. Transformers are the one they transform the
11:25
they transform the
11:25
they transform the the text, process the text and then go
11:28
the text, process the text and then go
11:28
the text, process the text and then go and recognize the words and compare them
11:31
and recognize the words and compare them
11:31
and recognize the words and compare them and it you know that's how they matches
11:33
and it you know that's how they matches
11:33
and it you know that's how they matches right so now let's take an example when
11:37
right so now let's take an example when
11:37
right so now let's take an example when you type something in chat GPT or or you
11:41
you type something in chat GPT or or you
11:41
you type something in chat GPT or or you say midjourney or any Deli front end
11:44
say midjourney or any Deli front end
11:44
say midjourney or any Deli front end that's using Deli they will say okay
11:48
that's using Deli they will say okay
11:48
that's using Deli they will say okay create a dog or or create a dog running
11:53
create a dog or or create a dog running
11:54
create a dog or or create a dog running or whatever. You just say that. So what
11:55
or whatever. You just say that. So what
11:56
or whatever. You just say that. So what they'll do is they take this prompt,
11:58
they'll do is they take this prompt,
11:58
they'll do is they take this prompt, right? Prompt breaks down. Prompt is
11:59
right? Prompt breaks down. Prompt is
11:59
right? Prompt breaks down. Prompt is like let's say draw a dog. That's
12:02
like let's say draw a dog. That's
12:02
like let's say draw a dog. That's perfect example. Draw a black dog. Draw
12:04
perfect example. Draw a black dog. Draw
12:04
perfect example. Draw a black dog. Draw a black dog.
12:06
a black dog.
12:06
a black dog. So when you say draw a black dog, this
12:08
So when you say draw a black dog, this
12:08
So when you say draw a black dog, this whole sentence is broken down to draw a
12:11
whole sentence is broken down to draw a
12:11
whole sentence is broken down to draw a black dog. And these are you know these
12:14
black dog. And these are you know these
12:14
black dog. And these are you know these are the tokens and they find the context
12:16
are the tokens and they find the context
12:16
are the tokens and they find the context in the back end behind the scene. Say
12:19
in the back end behind the scene. Say
12:19
in the back end behind the scene. Say they look for the dog, they look for
12:21
they look for the dog, they look for
12:21
they look for the dog, they look for black and then drawing means they need
12:23
black and then drawing means they need
12:23
black and then drawing means they need to draw something. So that's what it
12:25
to draw something. So that's what it
12:25
to draw something. So that's what it works and then comes and then they
12:27
works and then comes and then they
12:27
works and then comes and then they generate the images behind the scene
12:31
generate the images behind the scene
12:31
generate the images behind the scene and now once the output is generated
12:34
and now once the output is generated
12:34
and now once the output is generated you can give a feedback is it good? Is
12:36
you can give a feedback is it good? Is
12:36
you can give a feedback is it good? Is it bad? I don't like it. I don't use it
12:39
it bad? I don't like it. I don't use it
12:39
it bad? I don't like it. I don't use it or you're asking more questions about it
12:41
or you're asking more questions about it
12:41
or you're asking more questions about it if the image is bad. For example, you
12:43
if the image is bad. For example, you
12:43
if the image is bad. For example, you know, you say draw a dog, black dog, and
12:47
know, you say draw a dog, black dog, and
12:47
know, you say draw a dog, black dog, and the black dog is great, but it's not
12:48
the black dog is great, but it's not
12:48
the black dog is great, but it's not cute. Then you'll say draw a black dog
12:51
cute. Then you'll say draw a black dog
12:51
cute. Then you'll say draw a black dog uh with healthy whatever. So, anything
12:53
uh with healthy whatever. So, anything
12:53
uh with healthy whatever. So, anything input you gives him, AI keeps adding
12:56
input you gives him, AI keeps adding
12:56
input you gives him, AI keeps adding that behind the scene into its model,
12:58
that behind the scene into its model,
12:58
that behind the scene into its model, into it database, into a data set and
13:01
into it database, into a data set and
13:01
into it database, into a data set and this stop keep learning and get better.
13:03
this stop keep learning and get better.
13:04
this stop keep learning and get better. In other words, more you use, more data
13:06
In other words, more you use, more data
13:06
In other words, more you use, more data you feed it, more data it has, more it
13:09
you feed it, more data it has, more it
13:09
you feed it, more data it has, more it learn and train on, better it gets.
13:12
learn and train on, better it gets.
13:12
learn and train on, better it gets. So hopefully you guys learning
13:14
So hopefully you guys learning
13:14
So hopefully you guys learning something. Hopefully I'm making sense
13:15
something. Hopefully I'm making sense
13:15
something. Hopefully I'm making sense here. U but if you think anything is
13:18
here. U but if you think anything is
13:18
here. U but if you think anything is confusing feel free to ask questions
13:21
confusing feel free to ask questions
13:21
confusing feel free to ask questions after this I will take those questions.
13:24
after this I will take those questions.
13:24
after this I will take those questions. So one thing comes in discussions is LLM
13:28
So one thing comes in discussions is LLM
13:28
So one thing comes in discussions is LLM large language model and large language
13:30
large language model and large language
13:30
large language model and large language model is very important part or I would
13:32
model is very important part or I would
13:32
model is very important part or I would say the most important part of
13:34
say the most important part of
13:34
say the most important part of generative AI.
13:36
generative AI.
13:36
generative AI. large language models you can think of a
13:37
large language models you can think of a
13:38
large language models you can think of a brain right when you think of a
13:40
brain right when you think of a
13:40
brain right when you think of a generative AI there are three components
13:42
generative AI there are three components
13:42
generative AI there are three components to it one is front end front end of
13:44
to it one is front end front end of
13:44
to it one is front end front end of generative AI could be for example chat
13:46
generative AI could be for example chat
13:46
generative AI could be for example chat GPD chat GPT is a front end means all
13:48
GPD chat GPT is a front end means all
13:48
GPD chat GPT is a front end means all you do is put a prompt in there and
13:51
you do is put a prompt in there and
13:51
you do is put a prompt in there and that's all it does it's it's take the
13:53
that's all it does it's it's take the
13:53
that's all it does it's it's take the prompt it takes the input shows you the
13:55
prompt it takes the input shows you the
13:55
prompt it takes the input shows you the output that's what chat GPT is right
13:57
output that's what chat GPT is right
13:57
output that's what chat GPT is right chat GPT doesn't do any work its own
14:01
chat GPT doesn't do any work its own
14:01
chat GPT doesn't do any work its own what chat DB does it passes
14:04
what chat DB does it passes
14:04
what chat DB does it passes the that input text called prompt to the
14:08
the that input text called prompt to the
14:08
the that input text called prompt to the LLM large language model. Large language
14:10
LLM large language model. Large language
14:10
LLM large language model. Large language model is the is the brain behind that
14:13
model is the is the brain behind that
14:13
model is the is the brain behind that does everything. It it has the data set
14:15
does everything. It it has the data set
14:15
does everything. It it has the data set behind that works with the data set is
14:18
behind that works with the data set is
14:18
behind that works with the data set is there a lot of it has algorithms right
14:21
there a lot of it has algorithms right
14:21
there a lot of it has algorithms right it's processing it runs on
14:23
it's processing it runs on
14:23
it's processing it runs on supercomputers
14:25
supercomputers
14:25
supercomputers so when you pass something say draw a
14:28
so when you pass something say draw a
14:28
so when you pass something say draw a dog that LLM
14:31
dog that LLM
14:31
dog that LLM listens to that draw a dog and then does
14:34
listens to that draw a dog and then does
14:34
listens to that draw a dog and then does all the what we explained just now token
14:35
all the what we explained just now token
14:36
all the what we explained just now token and context and draw that so most of
14:38
and context and draw that so most of
14:38
and context and draw that so most of these large language models are trained
14:40
these large language models are trained
14:40
these large language models are trained on billions of parameters so means
14:43
on billions of parameters so means
14:43
on billions of parameters so means there's billions of parameters like one
14:45
there's billions of parameters like one
14:45
there's billions of parameters like one parameter is a dog and then there are
14:48
parameter is a dog and then there are
14:48
parameter is a dog and then there are multiple millions of millions of
14:50
multiple millions of millions of
14:50
multiple millions of millions of combination of those like cat jumping
14:53
combination of those like cat jumping
14:53
combination of those like cat jumping dog jumping right dog eating all that
14:57
dog jumping right dog eating all that
14:57
dog jumping right dog eating all that this is becomes like a combinations
15:01
this is becomes like a combinations
15:01
this is becomes like a combinations um as I said so so you probably heard
15:04
um as I said so so you probably heard
15:04
um as I said so so you probably heard this say oh open AI is sp is in
15:06
this say oh open AI is sp is in
15:06
this say oh open AI is sp is in investing hundreds of millions of
15:08
investing hundreds of millions of
15:08
investing hundreds of millions of dollars each month which they are so
15:11
dollars each month which they are so
15:11
dollars each month which they are so what they have done is they have taken
15:13
what they have done is they have taken
15:13
what they have done is they have taken entire our world's database store in it
15:16
entire our world's database store in it
15:16
entire our world's database store in it these big big databases giant databases
15:19
these big big databases giant databases
15:19
these big big databases giant databases and they are running these models on
15:20
and they are running these models on
15:20
and they are running these models on supercomputers
15:22
supercomputers
15:22
supercomputers uh that's why they partnered with
15:24
uh that's why they partnered with
15:24
uh that's why they partnered with Microsoft because initially they didn't
15:26
Microsoft because initially they didn't
15:26
Microsoft because initially they didn't have billions of dollars so Microsoft
15:28
have billions of dollars so Microsoft
15:28
have billions of dollars so Microsoft invested billions of dollars for them
15:30
invested billions of dollars for them
15:30
invested billions of dollars for them and gave them this
15:33
and gave them this
15:33
and gave them this Azure supercomputing powers
15:36
Azure supercomputing powers
15:36
Azure supercomputing powers and that's where GDP GP runs so GPT4 is
15:40
and that's where GDP GP runs so GPT4 is
15:40
and that's where GDP GP runs so GPT4 is open AAI's latest large language model.
15:46
So here's an example, right? This is a
15:48
So here's an example, right? This is a
15:48
So here's an example, right? This is a how how what what role LLM plays in
15:51
how how what what role LLM plays in
15:51
how how what what role LLM plays in generative AI. So there's input, right?
15:53
generative AI. So there's input, right?
15:53
generative AI. So there's input, right? You just put input, it becomes a ji
15:56
You just put input, it becomes a ji
15:56
You just put input, it becomes a ji prompt. Prompt goes to LLM. LLM does its
16:00
prompt. Prompt goes to LLM. LLM does its
16:00
prompt. Prompt goes to LLM. LLM does its own thing and generates the output and
16:02
own thing and generates the output and
16:02
own thing and generates the output and shows to the UI. That's pretty much like
16:04
shows to the UI. That's pretty much like
16:04
shows to the UI. That's pretty much like a simple mechanism under generative AI.
16:09
a simple mechanism under generative AI.
16:09
a simple mechanism under generative AI. AI generative AI actually is not even
16:12
AI generative AI actually is not even
16:12
AI generative AI actually is not even though it's AI behind the scene as AI
16:14
though it's AI behind the scene as AI
16:14
though it's AI behind the scene as AI but as a user you don't use much you
16:18
but as a user you don't use much you
16:18
but as a user you don't use much you don't really need to use and do much
16:20
don't really need to use and do much
16:20
don't really need to use and do much about learn about AI a lot at all or as
16:23
about learn about AI a lot at all or as
16:23
about learn about AI a lot at all or as a user right but when you're building
16:25
a user right but when you're building
16:25
a user right but when you're building these ji models or ji systems then you
16:30
these ji models or ji systems then you
16:30
these ji models or ji systems then you need to really dig deep into how these
16:33
need to really dig deep into how these
16:33
need to really dig deep into how these large language models are developed what
16:35
large language models are developed what
16:35
large language models are developed what algorithms are used what hardware used
16:38
algorithms are used what hardware used
16:38
algorithms are used what hardware used So you in if you are just a user and
16:41
So you in if you are just a user and
16:41
So you in if you are just a user and most people are right 95% of 98% of
16:45
most people are right 95% of 98% of
16:45
most people are right 95% of 98% of people are just users of geni they're
16:47
people are just users of geni they're
16:47
people are just users of geni they're consumers they're not building geni
16:49
consumers they're not building geni
16:49
consumers they're not building geni tools so you will probably if you're not
16:52
tools so you will probably if you're not
16:52
tools so you will probably if you're not really a data scientist or working for a
16:55
really a data scientist or working for a
16:55
really a data scientist or working for a company that's building it own large
16:56
company that's building it own large
16:56
company that's building it own large language model or small language models
16:59
language model or small language models
16:59
language model or small language models you probably won't even need all the
17:01
you probably won't even need all the
17:01
you probably won't even need all the details but it's good to know how it
17:02
details but it's good to know how it
17:02
details but it's good to know how it works behind the scene it's always good
17:04
works behind the scene it's always good
17:04
works behind the scene it's always good to know that
17:06
to know that
17:06
to know that so that's a little bit here um on this
17:08
so that's a little bit here um on this
17:08
so that's a little bit here um on this screen is what is role of data science
17:10
screen is what is role of data science
17:10
screen is what is role of data science in AI models. Well data scientist when
17:13
in AI models. Well data scientist when
17:13
in AI models. Well data scientist when we remember we just talked about the
17:15
we remember we just talked about the
17:15
we remember we just talked about the data initially you just get all kind of
17:18
data initially you just get all kind of
17:18
data initially you just get all kind of data data is coming from database coming
17:21
data data is coming from database coming
17:21
data data is coming from database coming from internet coming from web pages PDFs
17:24
from internet coming from web pages PDFs
17:24
from internet coming from web pages PDFs books scanned all those so data
17:26
books scanned all those so data
17:26
books scanned all those so data scientists what they do is they curate
17:29
scientists what they do is they curate
17:29
scientists what they do is they curate all the data put in the right format
17:31
all the data put in the right format
17:31
all the data put in the right format cleans it make sure there's no junk data
17:34
cleans it make sure there's no junk data
17:34
cleans it make sure there's no junk data also make sure that data accuracy is
17:37
also make sure that data accuracy is
17:37
also make sure that data accuracy is there right without accuracy there's
17:39
there right without accuracy there's
17:39
there right without accuracy there's going to be problem so that's what data
17:41
going to be problem so that's what data
17:42
going to be problem so that's what data science scientists do. So when you think
17:43
science scientists do. So when you think
17:43
science scientists do. So when you think of a what job is growing, one of the
17:46
of a what job is growing, one of the
17:46
of a what job is growing, one of the fastest growing job is the data
17:48
fastest growing job is the data
17:48
fastest growing job is the data scientist job. Why? Because of this
17:50
scientist job. Why? Because of this
17:50
scientist job. Why? Because of this reasons because AI heavily rely on data
17:54
reasons because AI heavily rely on data
17:54
reasons because AI heavily rely on data chain AI especially right jai heavily
17:57
chain AI especially right jai heavily
17:57
chain AI especially right jai heavily rely on this data
17:59
rely on this data
17:59
rely on this data and without the data and data is if data
18:02
and without the data and data is if data
18:02
and without the data and data is if data is bad data it's a problem. So data
18:04
is bad data it's a problem. So data
18:04
is bad data it's a problem. So data scientists play a major role and this
18:07
scientists play a major role and this
18:07
scientists play a major role and this role will continue to grow in coming
18:09
role will continue to grow in coming
18:09
role will continue to grow in coming time.
18:11
time.
18:11
time. Um then what happens these advanced
18:15
Um then what happens these advanced
18:15
Um then what happens these advanced models are trained on billions of
18:16
models are trained on billions of
18:16
models are trained on billions of parameters. So there's models are
18:18
parameters. So there's models are
18:18
parameters. So there's models are created which are algorithms and data
18:20
created which are algorithms and data
18:20
created which are algorithms and data and there are AI models then as we
18:22
and there are AI models then as we
18:22
and there are AI models then as we talked about recently that they need
18:24
talked about recently that they need
18:24
talked about recently that they need supercomputers to process data and
18:26
supercomputers to process data and
18:26
supercomputers to process data and generate output and that's why we just
18:28
generate output and that's why we just
18:28
generate output and that's why we just talked about GPT GPT4 is something if
18:31
talked about GPT GPT4 is something if
18:31
talked about GPT GPT4 is something if you want to learn more on this you can
18:32
you want to learn more on this you can
18:32
you want to learn more on this you can just search GPT AR4 architecture and
18:35
just search GPT AR4 architecture and
18:35
just search GPT AR4 architecture and you'll get a lot of stuff in there.
18:37
you'll get a lot of stuff in there.
18:37
you'll get a lot of stuff in there. Maybe in our coming episodes I will
18:39
Maybe in our coming episodes I will
18:39
Maybe in our coming episodes I will spend more time on this technical side.
18:43
spend more time on this technical side.
18:43
spend more time on this technical side. So when you think of a GPT4 um like what
18:47
So when you think of a GPT4 um like what
18:47
So when you think of a GPT4 um like what kind of hardware it needs. So it has one
18:49
kind of hardware it needs. So it has one
18:49
kind of hardware it needs. So it has one trillion parameter. This is a little bit
18:51
trillion parameter. This is a little bit
18:51
trillion parameter. This is a little bit details on GPT4. It runs on these Nvidia
18:56
details on GPT4. It runs on these Nvidia
18:56
details on GPT4. It runs on these Nvidia supercomputers and GP and this GPD4 has
19:00
supercomputers and GP and this GPD4 has
19:00
supercomputers and GP and this GPD4 has 25,000 or more GPUs running right. there
19:03
25,000 or more GPUs running right. there
19:03
25,000 or more GPUs running right. there 25,000 of supercomputers sitting in
19:06
25,000 of supercomputers sitting in
19:06
25,000 of supercomputers sitting in somewhere they have 20 million GPU hours
19:09
somewhere they have 20 million GPU hours
19:09
somewhere they have 20 million GPU hours it's trained on that and it has it has
19:11
it's trained on that and it has it has
19:11
it's trained on that and it has it has stores dozens of pabyte storage so when
19:14
stores dozens of pabyte storage so when
19:14
stores dozens of pabyte storage so when you think of pabytes go do and research
19:16
you think of pabytes go do and research
19:16
you think of pabytes go do and research how many zeros are in pabyte storage
19:20
how many zeros are in pabyte storage
19:20
how many zeros are in pabyte storage and um it yeah it cost 100 100 million
19:24
and um it yeah it cost 100 100 million
19:24
and um it yeah it cost 100 100 million plus GPU cloud compute that was that's
19:27
plus GPU cloud compute that was that's
19:27
plus GPU cloud compute that was that's Microsoft it cost almost $120 million
19:30
Microsoft it cost almost $120 million
19:30
Microsoft it cost almost $120 million just to train GPT4 model right so you
19:34
just to train GPT4 model right so you
19:34
just to train GPT4 model right so you you get the idea it's a massive massive
19:36
you get the idea it's a massive massive
19:36
you get the idea it's a massive massive undertaking
19:38
undertaking
19:38
undertaking so let's think about some of the talk
19:40
so let's think about some of the talk
19:40
so let's think about some of the talk about some of the popular JNAI models
19:43
about some of the popular JNAI models
19:43
about some of the popular JNAI models what are they are what they do so one of
19:45
what are they are what they do so one of
19:45
what are they are what they do so one of the most popular GPT4 as you can see it
19:48
the most popular GPT4 as you can see it
19:48
the most popular GPT4 as you can see it generates text to text so you say this
19:51
generates text to text so you say this
19:51
generates text to text so you say this can you write me a poem can you create
19:53
can you write me a poem can you create
19:53
can you write me a poem can you create code for me can you write a sorting
19:55
code for me can you write a sorting
19:55
code for me can you write a sorting algorithm in Python all those things so
19:58
algorithm in Python all those things so
19:58
algorithm in Python all those things so that's it takes tax generates tax these
20:00
that's it takes tax generates tax these
20:00
that's it takes tax generates tax these are popular ones Right.
20:03
are popular ones Right.
20:03
are popular ones Right. Dali is in med journey. Those are two
20:05
Dali is in med journey. Those are two
20:05
Dali is in med journey. Those are two most popular models for generating text
20:08
most popular models for generating text
20:08
most popular models for generating text to image. So if you're trying to
20:09
to image. So if you're trying to
20:09
to image. So if you're trying to generate text to image, if you're trying
20:11
generate text to image, if you're trying
20:11
generate text to image, if you're trying to generate text to movies, Sora is a
20:14
to generate text to movies, Sora is a
20:14
to generate text to movies, Sora is a model from OpenAI that's focuses on
20:17
model from OpenAI that's focuses on
20:17
model from OpenAI that's focuses on generating text to movies. You already
20:20
generating text to movies. You already
20:20
generating text to movies. You already heard of GitHub Copilot. It also uses
20:22
heard of GitHub Copilot. It also uses
20:22
heard of GitHub Copilot. It also uses behind the scene with the with another
20:25
behind the scene with the with another
20:25
behind the scene with the with another model. It generates text to code. And
20:27
model. It generates text to code. And
20:27
model. It generates text to code. And there's Adobe Firefly that generates
20:29
there's Adobe Firefly that generates
20:29
there's Adobe Firefly that generates image to image. Sono is a another model
20:32
image to image. Sono is a another model
20:32
image to image. Sono is a another model that generates sound to music. So if
20:34
that generates sound to music. So if
20:34
that generates sound to music. So if your interest is you know depending on
20:36
your interest is you know depending on
20:36
your interest is you know depending on what your interest is you can focus on
20:39
what your interest is you can focus on
20:39
what your interest is you can focus on these different models.
20:42
these different models.
20:42
these different models. Looks like more people join everybody
20:44
Looks like more people join everybody
20:44
Looks like more people join everybody welcome to the show. Today we are
20:45
welcome to the show. Today we are
20:45
welcome to the show. Today we are talking about generative AI. If you are
20:47
talking about generative AI. If you are
20:47
talking about generative AI. If you are late and you miss the show, you can this
20:50
late and you miss the show, you can this
20:50
late and you miss the show, you can this video will be recorded and available to
20:53
video will be recorded and available to
20:53
video will be recorded and available to download immediately after after this
20:57
download immediately after after this
20:57
download immediately after after this show on on YouTube on LinkedIn
21:01
show on on YouTube on LinkedIn
21:01
show on on YouTube on LinkedIn um even Twitter. It will also be
21:03
um even Twitter. It will also be
21:03
um even Twitter. It will also be available on Corner's
21:06
available on Corner's
21:06
available on Corner's YouTube account and also on C# Corner
21:08
YouTube account and also on C# Corner
21:08
YouTube account and also on C# Corner website and video section as well.
21:12
website and video section as well.
21:12
website and video section as well. So, generative AI is not perfect. It has
21:16
So, generative AI is not perfect. It has
21:16
So, generative AI is not perfect. It has some issues. it has problems. So when
21:18
some issues. it has problems. So when
21:18
some issues. it has problems. So when you think about when you're thinking
21:20
you think about when you're thinking
21:20
you think about when you're thinking about using generative AI, you have to
21:22
about using generative AI, you have to
21:22
about using generative AI, you have to make sure that whatever um content
21:25
make sure that whatever um content
21:25
make sure that whatever um content you're getting from it, it's it depends
21:28
you're getting from it, it's it depends
21:28
you're getting from it, it's it depends on what you like. If you're generating
21:30
on what you like. If you're generating
21:30
on what you like. If you're generating an image and you like it, doesn't
21:31
an image and you like it, doesn't
21:31
an image and you like it, doesn't matter, right? It's fine. If you are
21:35
matter, right? It's fine. If you are
21:35
matter, right? It's fine. If you are saying, oh, write me a letter, you know,
21:37
saying, oh, write me a letter, you know,
21:37
saying, oh, write me a letter, you know, some kind of legal letter, you want to
21:39
some kind of legal letter, you want to
21:39
some kind of legal letter, you want to make sure that somebody's checking that
21:41
make sure that somebody's checking that
21:41
make sure that somebody's checking that for the for the accuracy, for the proof.
21:44
for the for the accuracy, for the proof.
21:44
for the for the accuracy, for the proof. There's no mistakes in there. So think
21:47
There's no mistakes in there. So think
21:47
There's no mistakes in there. So think of Jen AI is your assistant which is
21:51
of Jen AI is your assistant which is
21:51
of Jen AI is your assistant which is just doing your work but they're not
21:53
just doing your work but they're not
21:53
just doing your work but they're not smart. That's why you got to think of
21:55
smart. That's why you got to think of
21:55
smart. That's why you got to think of them. They're generating content based
21:57
them. They're generating content based
21:57
them. They're generating content based on what they know what they have. It
21:59
on what they know what they have. It
21:59
on what they know what they have. It doesn't mean it's accurate. It also get
22:01
doesn't mean it's accurate. It also get
22:02
doesn't mean it's accurate. It also get hallucinations. Solutionations means if
22:04
hallucinations. Solutionations means if
22:04
hallucinations. Solutionations means if you ask DNI something and they can't
22:06
you ask DNI something and they can't
22:06
you ask DNI something and they can't find the they can't find the right data
22:11
find the they can't find the right data
22:11
find the they can't find the right data or they can't find the match it creates
22:13
or they can't find the match it creates
22:13
or they can't find the match it creates its own thing and uh that could turn
22:17
its own thing and uh that could turn
22:17
its own thing and uh that could turn into a problem right so you have to be
22:19
into a problem right so you have to be
22:19
into a problem right so you have to be make sure
22:21
make sure
22:21
make sure that the the you are proofreading the
22:25
that the the you are proofreading the
22:25
that the the you are proofreading the content generating sometime it conso it
22:28
content generating sometime it conso it
22:28
content generating sometime it conso it can also have the wrong text right so if
22:30
can also have the wrong text right so if
22:30
can also have the wrong text right so if for example if If a model is fed the
22:34
for example if If a model is fed the
22:34
for example if If a model is fed the wrong code
22:37
wrong code
22:37
wrong code and you're asking to generate the code,
22:38
and you're asking to generate the code,
22:38
and you're asking to generate the code, it's going to generate the wrong code,
22:40
it's going to generate the wrong code,
22:40
it's going to generate the wrong code, right? It's wrong maybe old code. It
22:43
right? It's wrong maybe old code. It
22:43
right? It's wrong maybe old code. It would be it could be in for example
22:45
would be it could be in for example
22:45
would be it could be in for example copilot, right? If you use copilot to
22:47
copilot, right? If you use copilot to
22:47
copilot, right? If you use copilot to generate code or you use chatgp to say,
22:50
generate code or you use chatgp to say,
22:50
generate code or you use chatgp to say, oh, write me code for this sorting an
22:53
oh, write me code for this sorting an
22:53
oh, write me code for this sorting an array uh using this this this right. So
22:56
array uh using this this this right. So
22:56
array uh using this this this right. So it what it's do is whatever code it has
22:58
it what it's do is whatever code it has
22:58
it what it's do is whatever code it has when it was trained, it's going to
23:00
when it was trained, it's going to
23:00
when it was trained, it's going to generate new code based on that. That
23:02
generate new code based on that. That
23:02
generate new code based on that. That code could be 6 months old, 8 months
23:04
code could be 6 months old, 8 months
23:04
code could be 6 months old, 8 months old. It could be an older C# language.
23:07
old. It could be an older C# language.
23:07
old. It could be an older C# language. It could be a um it could have issues,
23:09
It could be a um it could have issues,
23:09
It could be a um it could have issues, security issues, right? Depends on where
23:12
security issues, right? Depends on where
23:12
security issues, right? Depends on where the model was trained from. So you have
23:14
the model was trained from. So you have
23:14
the model was trained from. So you have to make sure that uh once the content is
23:18
to make sure that uh once the content is
23:18
to make sure that uh once the content is generated uh that it's accurate, it's
23:20
generated uh that it's accurate, it's
23:20
generated uh that it's accurate, it's the one you want, you've tested it,
23:21
the one you want, you've tested it,
23:22
the one you want, you've tested it, things like that.
23:25
So where are we going the future? Where
23:28
So where are we going the future? Where
23:28
So where are we going the future? Where is the future heading with Gen AI? I
23:30
is the future heading with Gen AI? I
23:30
is the future heading with Gen AI? I think it's pretty bright. There's going
23:31
think it's pretty bright. There's going
23:31
think it's pretty bright. There's going to be a lot of more applications and
23:33
to be a lot of more applications and
23:33
to be a lot of more applications and they're already being developed right
23:35
they're already being developed right
23:35
they're already being developed right now. So you know you're saying I think
23:37
now. So you know you're saying I think
23:37
now. So you know you're saying I think you'll have your own personal AI doctor
23:40
you'll have your own personal AI doctor
23:40
you'll have your own personal AI doctor for example. You'll be talking to some
23:41
for example. You'll be talking to some
23:41
for example. You'll be talking to some kind of AI tool. It could be a hardware.
23:44
kind of AI tool. It could be a hardware.
23:44
kind of AI tool. It could be a hardware. It could be even your phone and saying
23:46
It could be even your phone and saying
23:46
It could be even your phone and saying hey I I have a headache. Uh it's in
23:50
hey I I have a headache. Uh it's in
23:50
hey I I have a headache. Uh it's in brain. It's in the left side. I do this.
23:52
brain. It's in the left side. I do this.
23:52
brain. It's in the left side. I do this. It's going to start recommending you
23:54
It's going to start recommending you
23:54
It's going to start recommending you something say or you can say I have a
23:56
something say or you can say I have a
23:56
something say or you can say I have a pain in my stomach um stomach and what
23:59
pain in my stomach um stomach and what
24:00
pain in my stomach um stomach and what could I do so you just start chatting
24:01
could I do so you just start chatting
24:01
could I do so you just start chatting with this becomes you know there can be
24:03
with this becomes you know there can be
24:03
with this becomes you know there can be a devices where you can talk to them as
24:05
a devices where you can talk to them as
24:05
a devices where you can talk to them as a as a you know voice so that's coming
24:08
a as a you know voice so that's coming
24:08
a as a you know voice so that's coming soon right and who who knows in the the
24:11
soon right and who who knows in the the
24:11
soon right and who who knows in the the way hardware innovation is going um it
24:14
way hardware innovation is going um it
24:14
way hardware innovation is going um it could have a person literally
24:16
could have a person literally
24:16
could have a person literally um I have doctor kind of looking and you
24:19
um I have doctor kind of looking and you
24:19
um I have doctor kind of looking and you just talk to that person who knows AI AI
24:21
just talk to that person who knows AI AI
24:21
just talk to that person who knows AI AI tutor. You can definitely have AI tutors
24:23
tutor. You can definitely have AI tutors
24:23
tutor. You can definitely have AI tutors in coming time where you can say I want
24:26
in coming time where you can say I want
24:26
in coming time where you can say I want to learn this. I want to learn at low
24:28
to learn this. I want to learn at low
24:28
to learn this. I want to learn at low low speed. I want to learn in Hindi. All
24:30
low speed. I want to learn in Hindi. All
24:30
low speed. I want to learn in Hindi. All those things they're coming. Obviously
24:33
those things they're coming. Obviously
24:33
those things they're coming. Obviously you can create music and movies and
24:35
you can create music and movies and
24:35
you can create music and movies and video games. So in future
24:38
video games. So in future
24:38
video games. So in future I can just sit here and type and ask
24:40
I can just sit here and type and ask
24:40
I can just sit here and type and ask Sora to create a an action movie that
24:44
Sora to create a an action movie that
24:44
Sora to create a an action movie that will include these characters. So
24:46
will include these characters. So
24:46
will include these characters. So imagine your imagination
24:49
imagine your imagination
24:49
imagine your imagination is the is the key here and you can
24:52
is the is the key here and you can
24:52
is the is the key here and you can imagine things and most of the time this
24:55
imagine things and most of the time this
24:55
imagine things and most of the time this Sora can create the movie for you in
24:57
Sora can create the movie for you in
24:57
Sora can create the movie for you in high quality and you can watch your own
24:59
high quality and you can watch your own
24:59
high quality and you can watch your own movie. How cool is that? That's pretty
25:01
movie. How cool is that? That's pretty
25:01
movie. How cool is that? That's pretty cool, right? You can write your own
25:03
cool, right? You can write your own
25:03
cool, right? You can write your own story book. You can write your comics
25:04
story book. You can write your comics
25:04
story book. You can write your comics book. You can do all that by yourself.
25:07
book. You can do all that by yourself.
25:07
book. You can do all that by yourself. Um there are already AI systems being
25:10
Um there are already AI systems being
25:10
Um there are already AI systems being developed. They will run your business
25:11
developed. They will run your business
25:12
developed. They will run your business workflows. For example, AI will watch
25:14
workflows. For example, AI will watch
25:14
workflows. For example, AI will watch your all emails, will watch your work
25:17
your all emails, will watch your work
25:17
your all emails, will watch your work habits, will watch your tasks and slowly
25:20
habits, will watch your tasks and slowly
25:20
habits, will watch your tasks and slowly it will learn and it will start doing a
25:22
it will learn and it will start doing a
25:22
it will learn and it will start doing a task for you on your behalf. You can
25:25
task for you on your behalf. You can
25:25
task for you on your behalf. You can always confirm, you can always approve,
25:27
always confirm, you can always approve,
25:27
always confirm, you can always approve, you can always do all these things. But
25:29
you can always do all these things. But
25:29
you can always do all these things. But this is coming very soon in next year,
25:31
this is coming very soon in next year,
25:32
this is coming very soon in next year, next two years, next three years.
25:33
next two years, next three years.
25:33
next two years, next three years. There's going to be chain AI and this
25:35
There's going to be chain AI and this
25:35
There's going to be chain AI and this kind of AI is going to be everywhere. We
25:37
kind of AI is going to be everywhere. We
25:38
kind of AI is going to be everywhere. We can see
25:40
can see
25:40
can see um you can have your own digital AI
25:41
um you can have your own digital AI
25:41
um you can have your own digital AI digital tune. For example, if you're
25:43
digital tune. For example, if you're
25:43
digital tune. For example, if you're sleep sleeping and your AI tune could be
25:46
sleep sleeping and your AI tune could be
25:46
sleep sleeping and your AI tune could be talking to somebody, right? For example,
25:48
talking to somebody, right? For example,
25:48
talking to somebody, right? For example, I'm here running the show in future. I
25:51
I'm here running the show in future. I
25:51
I'm here running the show in future. I may be sleeping or doing something else
25:53
may be sleeping or doing something else
25:53
may be sleeping or doing something else while my twin is running probably this
25:55
while my twin is running probably this
25:55
while my twin is running probably this show. That'll be cool.
25:59
show. That'll be cool.
25:59
show. That'll be cool. Okay, that's cool guys. So next this was
26:02
Okay, that's cool guys. So next this was
26:02
Okay, that's cool guys. So next this was it for today's we'll take some questions
26:04
it for today's we'll take some questions
26:04
it for today's we'll take some questions after this but the next episode same
26:07
after this but the next episode same
26:07
after this but the next episode same time I'm going to talk about prompt
26:08
time I'm going to talk about prompt
26:08
time I'm going to talk about prompt engineering prompting
26:11
engineering prompting
26:11
engineering prompting is the key for jai if you want to get
26:14
is the key for jai if you want to get
26:14
is the key for jai if you want to get best out of it if you want to become a
26:16
best out of it if you want to become a
26:16
best out of it if you want to become a really productive prompt engineering is
26:19
really productive prompt engineering is
26:19
really productive prompt engineering is the key and don't worry about the name
26:21
the key and don't worry about the name
26:21
the key and don't worry about the name prompt engineering it's not complex as
26:23
prompt engineering it's not complex as
26:23
prompt engineering it's not complex as it sounds because when you think of
26:25
it sounds because when you think of
26:25
it sounds because when you think of engineering it always sounds like oh
26:27
engineering it always sounds like oh
26:27
engineering it always sounds like oh it's very complex topic prompt
26:29
it's very complex topic prompt
26:29
it's very complex topic prompt engineering is not complex topic at all.
26:31
engineering is not complex topic at all.
26:31
engineering is not complex topic at all. It's very simple. Um, so next week what
26:34
It's very simple. Um, so next week what
26:34
It's very simple. Um, so next week what we will do is we are going to talk about
26:35
we will do is we are going to talk about
26:35
we will do is we are going to talk about prompt engineering same time Saturday
26:39
prompt engineering same time Saturday
26:39
prompt engineering same time Saturday 11:00 a.m. Now let's take some
26:41
11:00 a.m. Now let's take some
26:41
11:00 a.m. Now let's take some questions.
26:46
Okay, looks like uh let's take some
26:48
Okay, looks like uh let's take some
26:48
Okay, looks like uh let's take some questions guys. Hopefully look a lot of
26:50
questions guys. Hopefully look a lot of
26:50
questions guys. Hopefully look a lot of people joined us. That's great. Thank
26:52
people joined us. That's great. Thank
26:52
people joined us. That's great. Thank you everybody for joining. It looks like
26:54
you everybody for joining. It looks like
26:54
you everybody for joining. It looks like some people joined late. Um
26:57
some people joined late. Um
26:57
some people joined late. Um let's take some question. Rajes is
26:59
let's take some question. Rajes is
26:59
let's take some question. Rajes is saying uh you can stop sharing my screen
27:01
saying uh you can stop sharing my screen
27:01
saying uh you can stop sharing my screen the book now let me see
27:05
the book now let me see
27:05
the book now let me see Rajes is saying so LLMs consist of data
27:08
Rajes is saying so LLMs consist of data
27:08
Rajes is saying so LLMs consist of data as well yeah so data is a part of these
27:11
as well yeah so data is a part of these
27:11
as well yeah so data is a part of these large language models the they're all
27:15
large language models the they're all
27:15
large language models the they're all behind the scene running the whole thing
27:17
behind the scene running the whole thing
27:17
behind the scene running the whole thing right um a data set is sitting somewhere
27:21
right um a data set is sitting somewhere
27:21
right um a data set is sitting somewhere right data set is sitting somewhere it's
27:24
right data set is sitting somewhere it's
27:24
right data set is sitting somewhere it's you know there algorithms and data
27:25
you know there algorithms and data
27:26
you know there algorithms and data that's what it is right there's a lot of
27:27
that's what it is right there's a lot of
27:27
that's what it is right there's a lot of algorithms running and there's a lot of
27:29
algorithms running and there's a lot of
27:29
algorithms running and there's a lot of data running eventually all the
27:31
data running eventually all the
27:31
data running eventually all the parameters going from input has to go go
27:34
parameters going from input has to go go
27:34
parameters going from input has to go go to the data sets and find the matching
27:37
to the data sets and find the matching
27:37
to the data sets and find the matching and then generate the output. So that's
27:39
and then generate the output. So that's
27:39
and then generate the output. So that's kind of the whole process. I will talk
27:41
kind of the whole process. I will talk
27:41
kind of the whole process. I will talk about LLMs more in my upcoming show how
27:44
about LLMs more in my upcoming show how
27:44
about LLMs more in my upcoming show how to create your own LLMs. We will also
27:46
to create your own LLMs. We will also
27:46
to create your own LLMs. We will also talk on that in details as well. By the
27:49
talk on that in details as well. By the
27:49
talk on that in details as well. By the way, we on C# corner we have already
27:51
way, we on C# corner we have already
27:52
way, we on C# corner we have already created our it's called it's not LLM
27:54
created our it's called it's not LLM
27:54
created our it's called it's not LLM small language model and the difference
27:57
small language model and the difference
27:57
small language model and the difference is because focus is on education in C#
27:59
is because focus is on education in C#
27:59
is because focus is on education in C# community only.
28:02
community only.
28:02
community only. Okay. So let's take another question.
28:04
Okay. So let's take another question.
28:04
Okay. So let's take another question. Sak says so do you think in the future
28:08
Sak says so do you think in the future
28:08
Sak says so do you think in the future AI models will be smart enough to figure
28:10
AI models will be smart enough to figure
28:10
AI models will be smart enough to figure out the best prompts on their own
28:13
out the best prompts on their own
28:13
out the best prompts on their own without us writing them? So prompting is
28:16
without us writing them? So prompting is
28:16
without us writing them? So prompting is really answer is yes. So there's already
28:20
really answer is yes. So there's already
28:20
really answer is yes. So there's already we I don't know if you know John Gorel
28:22
we I don't know if you know John Gorel
28:22
we I don't know if you know John Gorel he's part of our team he was also CEO of
28:25
he's part of our team he was also CEO of
28:25
he's part of our team he was also CEO of Alpine Gate AI we have created something
28:29
Alpine Gate AI we have created something
28:29
Alpine Gate AI we have created something called a smart prompting for C# corner
28:32
called a smart prompting for C# corner
28:32
called a smart prompting for C# corner it's coming on C# corner in next few
28:34
it's coming on C# corner in next few
28:34
it's coming on C# corner in next few months
28:35
months
28:35
months uh where you can input a little bit you
28:39
uh where you can input a little bit you
28:39
uh where you can input a little bit you can have your intent what you're trying
28:40
can have your intent what you're trying
28:40
can have your intent what you're trying to do and if you're not smart in writing
28:44
to do and if you're not smart in writing
28:44
to do and if you're not smart in writing a prompt it's going to write prompt
28:46
a prompt it's going to write prompt
28:46
a prompt it's going to write prompt behind on your behalf But the key again
28:49
behind on your behalf But the key again
28:49
behind on your behalf But the key again is it has to learn more about you, your
28:51
is it has to learn more about you, your
28:51
is it has to learn more about you, your habits. It has to know what exactly are
28:53
habits. It has to know what exactly are
28:53
habits. It has to know what exactly are you trying to do. The intent still has
28:56
you trying to do. The intent still has
28:56
you trying to do. The intent still has to go to the AI. If you don't know the
28:58
to go to the AI. If you don't know the
28:58
to go to the AI. If you don't know the intent, it's without the intent the
29:01
intent, it's without the intent the
29:01
intent, it's without the intent the prompting writing smart prompt will be
29:04
prompting writing smart prompt will be
29:04
prompting writing smart prompt will be really hard. So for example, I can just
29:08
really hard. So for example, I can just
29:08
really hard. So for example, I can just say, oh, create a movie. All right, but
29:12
say, oh, create a movie. All right, but
29:12
say, oh, create a movie. All right, but that's that's input. But I you know AI
29:14
that's that's input. But I you know AI
29:14
that's that's input. But I you know AI won't know LLM won't know what kind of
29:16
won't know LLM won't know what kind of
29:16
won't know LLM won't know what kind of movie you so it will ask more details.
29:18
movie you so it will ask more details.
29:18
movie you so it will ask more details. Hey is it action movie or is it a funny
29:21
Hey is it action movie or is it a funny
29:21
Hey is it action movie or is it a funny movie or is it a comedy or is it a
29:24
movie or is it a comedy or is it a
29:24
movie or is it a comedy or is it a whatever then you'll say action movie.
29:25
whatever then you'll say action movie.
29:25
whatever then you'll say action movie. Oh it's action movie. How do you want to
29:28
Oh it's action movie. How do you want to
29:28
Oh it's action movie. How do you want to use this as a warrior old age new action
29:31
use this as a warrior old age new action
29:31
use this as a warrior old age new action movie or you want like a Tom Cruz action
29:33
movie or you want like a Tom Cruz action
29:33
movie or you want like a Tom Cruz action movie. So this can AI can become more
29:36
movie. So this can AI can become more
29:36
movie. So this can AI can become more smart and smarter and rather than you
29:39
smart and smarter and rather than you
29:39
smart and smarter and rather than you just putting the prompt it can then
29:41
just putting the prompt it can then
29:41
just putting the prompt it can then start talking to you like somebody's
29:42
start talking to you like somebody's
29:42
start talking to you like somebody's interviewing you for example you know
29:45
interviewing you for example you know
29:45
interviewing you for example you know when you go to a director with a story
29:48
when you go to a director with a story
29:48
when you go to a director with a story director ask you all those questions
29:50
director ask you all those questions
29:50
director ask you all those questions right the story may not be complete you
29:51
right the story may not be complete you
29:52
right the story may not be complete you have to dig down so the AI this this AI
29:56
have to dig down so the AI this this AI
29:56
have to dig down so the AI this this AI could even become smarter and get all
29:58
could even become smarter and get all
29:58
could even become smarter and get all the input it needs for example soft
30:01
the input it needs for example soft
30:01
the input it needs for example soft right I'll give you perfect example when
30:03
right I'll give you perfect example when
30:03
right I'll give you perfect example when to when a client comes to me and says
30:06
to when a client comes to me and says
30:06
to when a client comes to me and says Mahes we want to build this new software
30:08
Mahes we want to build this new software
30:08
Mahes we want to build this new software application then they just say oh I want
30:10
application then they just say oh I want
30:10
application then they just say oh I want software application that's what they
30:12
software application that's what they
30:12
software application that's what they say sometime they say I want an app for
30:14
say sometime they say I want an app for
30:14
say sometime they say I want an app for my uh dog grooming business that's all
30:17
my uh dog grooming business that's all
30:17
my uh dog grooming business that's all they say app for my dog grooming
30:18
they say app for my dog grooming
30:18
they say app for my dog grooming business that's what they say so as a
30:20
business that's what they say so as a
30:20
business that's what they say so as a human what I do is I say okay let's ask
30:24
human what I do is I say okay let's ask
30:24
human what I do is I say okay let's ask all these questions based on that the
30:26
all these questions based on that the
30:26
all these questions based on that the requirements become bigger and bigger
30:28
requirements become bigger and bigger
30:28
requirements become bigger and bigger and clear so that's what it will become
30:30
and clear so that's what it will become
30:30
and clear so that's what it will become so answer is it won't be the AI models
30:32
so answer is it won't be the AI models
30:32
so answer is it won't be the AI models that will become smart enough but yeah
30:35
that will become smart enough but yeah
30:35
that will become smart enough but yeah it will come from the answers will come
30:36
it will come from the answers will come
30:36
it will come from the answers will come from questions more questions will come
30:38
from questions more questions will come
30:38
from questions more questions will come from AI models they are already there
30:41
from AI models they are already there
30:41
from AI models they are already there you just have to put this middle layer
30:44
you just have to put this middle layer
30:44
you just have to put this middle layer smart prompting to get all those
30:47
smart prompting to get all those
30:47
smart prompting to get all those in-depth answers from there right and
30:49
in-depth answers from there right and
30:49
in-depth answers from there right and that's also that concept of this is
30:52
that's also that concept of this is
30:52
that's also that concept of this is called wive coding right you heard of
30:54
called wive coding right you heard of
30:54
called wive coding right you heard of the term w coding w prompting it's more
30:57
the term w coding w prompting it's more
30:57
the term w coding w prompting it's more like being a fluid and creative and
31:00
like being a fluid and creative and
31:00
like being a fluid and creative and interactive with building these prompts
31:04
Great question. All right, let's take a
31:06
Great question. All right, let's take a
31:06
Great question. All right, let's take a few more questions, guys. Hey, happy
31:08
few more questions, guys. Hey, happy
31:08
few more questions, guys. Hey, happy birthday. Looks like Babu has a birthday
31:10
birthday. Looks like Babu has a birthday
31:10
birthday. Looks like Babu has a birthday today. Happy birthday, Babu. We have
31:12
today. Happy birthday, Babu. We have
31:12
today. Happy birthday, Babu. We have Look like some people are using chat GPT
31:14
Look like some people are using chat GPT
31:14
Look like some people are using chat GPT cloud. Is there any guys you want to
31:17
cloud. Is there any guys you want to
31:17
cloud. Is there any guys you want to share that anybody
31:21
loves any of these new tools which are
31:24
loves any of these new tools which are
31:24
loves any of these new tools which are not popular, please share here. So, Amit
31:27
not popular, please share here. So, Amit
31:27
not popular, please share here. So, Amit says, Amit is for that try Google AI
31:29
says, Amit is for that try Google AI
31:30
says, Amit is for that try Google AI studio. They have a prompt gallery that
31:31
studio. They have a prompt gallery that
31:32
studio. They have a prompt gallery that provides you with pre-esigned prompts.
31:34
provides you with pre-esigned prompts.
31:34
provides you with pre-esigned prompts. There you go. Right. So they have lot of
31:36
There you go. Right. So they have lot of
31:36
There you go. Right. So they have lot of lot of pre-esigned prompts are. So we
31:38
lot of pre-esigned prompts are. So we
31:38
lot of pre-esigned prompts are. So we have something called sharp GPT. Similar
31:41
have something called sharp GPT. Similar
31:41
have something called sharp GPT. Similar to what just AIT said we have something
31:43
to what just AIT said we have something
31:43
to what just AIT said we have something called sharp GPT and it is it has
31:46
called sharp GPT and it is it has
31:46
called sharp GPT and it is it has predefined prompts already for for C#
31:49
predefined prompts already for for C#
31:50
predefined prompts already for for C# corner users about the interview
31:51
corner users about the interview
31:51
corner users about the interview preparations about the uh writing code
31:54
preparations about the uh writing code
31:54
preparations about the uh writing code about learning about career growth
31:57
about learning about career growth
31:57
about learning about career growth career advice. So you don't have to
31:59
career advice. So you don't have to
31:59
career advice. So you don't have to write the prompt because writing prompt
32:01
write the prompt because writing prompt
32:01
write the prompt because writing prompt if you're you know obviously not
32:03
if you're you know obviously not
32:03
if you're you know obviously not everybody's creative or everybody
32:05
everybody's creative or everybody
32:05
everybody's creative or everybody doesn't have these English and most of
32:06
doesn't have these English and most of
32:06
doesn't have these English and most of these models are done in English right
32:08
these models are done in English right
32:08
these models are done in English right what if my English vocabulary is not
32:11
what if my English vocabulary is not
32:11
what if my English vocabulary is not that strong or creativity not there so
32:13
that strong or creativity not there so
32:13
that strong or creativity not there so those yeah those apps are already being
32:15
those yeah those apps are already being
32:16
those yeah those apps are already being developed as well so great question
32:19
developed as well so great question
32:19
developed as well so great question all right looks like any any other
32:21
all right looks like any any other
32:21
all right looks like any any other questions we have guys everybody
32:23
questions we have guys everybody
32:23
questions we have guys everybody hopefully you learned something unless
32:25
hopefully you learned something unless
32:25
hopefully you learned something unless you have any more questions ask them.
32:27
you have any more questions ask them.
32:27
you have any more questions ask them. Now we have few more minutes to go.
32:29
Now we have few more minutes to go.
32:29
Now we have few more minutes to go. Looks like a lot of people join. I
32:31
Looks like a lot of people join. I
32:31
Looks like a lot of people join. I appreciate you joining and thank you so
32:33
appreciate you joining and thank you so
32:33
appreciate you joining and thank you so much for participating and chiming in.
32:36
much for participating and chiming in.
32:36
much for participating and chiming in. By the way, if you want to come on these
32:38
By the way, if you want to come on these
32:38
By the way, if you want to come on these shows, you want to talk about some of
32:40
shows, you want to talk about some of
32:40
shows, you want to talk about some of these topics, let us know also. We may
32:43
these topics, let us know also. We may
32:43
these topics, let us know also. We may start bringing some guests here, right?
32:46
start bringing some guests here, right?
32:46
start bringing some guests here, right? Uh who can share their own um in
32:50
Uh who can share their own um in
32:50
Uh who can share their own um in innovations or anything experiences
32:53
innovations or anything experiences
32:53
innovations or anything experiences here. Uh yeah. So Rajes here is Rajes
32:56
here. Uh yeah. So Rajes here is Rajes
32:56
here. Uh yeah. So Rajes here is Rajes welcome good to see you after a long
32:58
welcome good to see you after a long
32:58
welcome good to see you after a long time as in chat GPT when option for web
33:03
time as in chat GPT when option for web
33:03
time as in chat GPT when option for web search is selected how does LLM work
33:07
search is selected how does LLM work
33:07
search is selected how does LLM work um web searches is just a different kind
33:10
um web searches is just a different kind
33:10
um web searches is just a different kind of model right so chat GPT because think
33:14
of model right so chat GPT because think
33:14
of model right so chat GPT because think about think of this way chat GPT uses
33:17
about think of this way chat GPT uses
33:17
about think of this way chat GPT uses this LLM which cost lot of money because
33:21
this LLM which cost lot of money because
33:21
this LLM which cost lot of money because when you think of large language Model
33:23
when you think of large language Model
33:23
when you think of large language Model think of like a massive database. So
33:26
think of like a massive database. So
33:26
think of like a massive database. So when you search if you have a massive
33:28
when you search if you have a massive
33:28
when you search if you have a massive database of say six all six billion
33:31
database of say six all six billion
33:31
database of say six all six billion people on the planet and you search for
33:33
people on the planet and you search for
33:33
people on the planet and you search for a name now the database is going to go
33:36
a name now the database is going to go
33:36
a name now the database is going to go through the entire database until it
33:39
through the entire database until it
33:39
through the entire database until it finds those that name right you search
33:41
finds those that name right you search
33:41
finds those that name right you search for Rajesh Nandwani so there's a
33:43
for Rajesh Nandwani so there's a
33:43
for Rajesh Nandwani so there's a database of entire planet with all the
33:46
database of entire planet with all the
33:46
database of entire planet with all the names called Raj and you say let's
33:48
names called Raj and you say let's
33:48
names called Raj and you say let's search Rajes Nandwani now you're going
33:50
search Rajes Nandwani now you're going
33:50
search Rajes Nandwani now you're going and searching Rajes for in all these six
33:52
and searching Rajes for in all these six
33:52
and searching Rajes for in all these six seven billion people right so that's lot
33:55
seven billion people right so that's lot
33:55
seven billion people right so that's lot lot of processing going on behind the
33:57
lot of processing going on behind the
33:57
lot of processing going on behind the scene. But what if you say oh there is a
34:00
scene. But what if you say oh there is a
34:00
scene. But what if you say oh there is a Rajes in in Noa
34:02
Rajes in in Noa
34:02
Rajes in in Noa and there's a Rajes in Noa who's a
34:05
and there's a Rajes in Noa who's a
34:05
and there's a Rajes in Noa who's a software engineer or software architect
34:07
software engineer or software architect
34:07
software engineer or software architect and he lives in this. Now there's a
34:10
and he lives in this. Now there's a
34:10
and he lives in this. Now there's a small model sitting on the side which is
34:13
small model sitting on the side which is
34:13
small model sitting on the side which is just for the NOA. Now the query will go
34:16
just for the NOA. Now the query will go
34:16
just for the NOA. Now the query will go to the NOA and that's much faster less
34:18
to the NOA and that's much faster less
34:18
to the NOA and that's much faster less processing. So it's it's more different
34:20
processing. So it's it's more different
34:20
processing. So it's it's more different not more not different than having lot
34:23
not more not different than having lot
34:23
not more not different than having lot of smaller tables focused on different
34:25
of smaller tables focused on different
34:26
of smaller tables focused on different things different entities different
34:27
things different entities different
34:27
things different entities different areas right so with search search you
34:30
areas right so with search search you
34:30
areas right so with search search you don't need to go and and search the
34:33
don't need to go and and search the
34:33
don't need to go and and search the entire LLM the run the whole LLM so they
34:37
entire LLM the run the whole LLM so they
34:37
entire LLM the run the whole LLM so they have a different search database search
34:38
have a different search database search
34:38
have a different search database search model and you just go in there and
34:39
model and you just go in there and
34:40
model and you just go in there and search is simple it's just search right
34:41
search is simple it's just search right
34:41
search is simple it's just search right you don't need to run the whole large
34:43
you don't need to run the whole large
34:43
you don't need to run the whole large language model for that which is less
34:45
language model for that which is less
34:46
language model for that which is less costly and runs much faster otherwise
34:48
costly and runs much faster otherwise
34:48
costly and runs much faster otherwise Guys, it's crazy too. So, there is a
34:51
Guys, it's crazy too. So, there is a
34:51
Guys, it's crazy too. So, there is a research just came out that if you say
34:54
research just came out that if you say
34:54
research just came out that if you say hi,
34:55
hi,
34:55
hi, thank you to these models, it cost them
34:58
thank you to these models, it cost them
34:58
thank you to these models, it cost them millions of dollars by just saying that
35:00
millions of dollars by just saying that
35:00
millions of dollars by just saying that because they still have to take that
35:03
because they still have to take that
35:03
because they still have to take that thank you or high in the prompt and
35:07
thank you or high in the prompt and
35:07
thank you or high in the prompt and process it through these these whole
35:10
process it through these these whole
35:10
process it through these these whole this you know computers and there are
35:12
this you know computers and there are
35:12
this you know computers and there are lot of computers running that
35:13
lot of computers running that
35:13
lot of computers running that processing. So it's it's it's better to
35:16
processing. So it's it's it's better to
35:16
processing. So it's it's it's better to be more straightforward with these these
35:19
be more straightforward with these these
35:19
be more straightforward with these these you know um jai tools sometimes rather
35:23
you know um jai tools sometimes rather
35:23
you know um jai tools sometimes rather than being more formal because it's then
35:25
than being more formal because it's then
35:25
than being more formal because it's then every everything you type there it's
35:27
every everything you type there it's
35:27
every everything you type there it's cost money behind the scene. Uh great
35:30
cost money behind the scene. Uh great
35:30
cost money behind the scene. Uh great questions guys. Thank you everybody. I
35:32
questions guys. Thank you everybody. I
35:32
questions guys. Thank you everybody. I think this was a good show. A lot of
35:34
think this was a good show. A lot of
35:34
think this was a good show. A lot of people are still hanging around there.
35:36
people are still hanging around there.
35:36
people are still hanging around there. As I said uh if you like it feel free to
35:39
As I said uh if you like it feel free to
35:39
As I said uh if you like it feel free to post your comments. You have any
35:41
post your comments. You have any
35:41
post your comments. You have any feedback what we need to improve. We're
35:43
feedback what we need to improve. We're
35:43
feedback what we need to improve. We're always looking for you know great um
35:47
always looking for you know great um
35:47
always looking for you know great um feedback how can we improve this timing
35:51
feedback how can we improve this timing
35:51
feedback how can we improve this timing sound topics you want to come as a guest
35:55
sound topics you want to come as a guest
35:55
sound topics you want to come as a guest feel free to share all your all your
35:57
feel free to share all your all your
35:57
feel free to share all your all your feedback and then next week we will talk
36:00
feedback and then next week we will talk
36:00
feedback and then next week we will talk about prompt engineering um if you want
36:03
about prompt engineering um if you want
36:03
about prompt engineering um if you want to if you just joined us later go watch
36:05
to if you just joined us later go watch
36:05
to if you just joined us later go watch it this full full show it's will be
36:07
it this full full show it's will be
36:07
it this full full show it's will be available recorded uh after a few
36:09
available recorded uh after a few
36:09
available recorded uh after a few minutes on all social media and also Dar
36:13
minutes on all social media and also Dar
36:13
minutes on all social media and also Dar Corner YouTube YouTube channel. Thank
36:16
Corner YouTube YouTube channel. Thank
36:16
Corner YouTube YouTube channel. Thank you so much. I will see you next week.
36:26
[Music]
Computers & Electronics
Internet & Telecom