Google AI Studio's cool interface
and powerful tools are shaking up the AI world. Everyone's talking about how it makes projects simpler and speeds up innovation. Why Google AI Studio? Because it's designed to make AI development more accessible
and efficient for everyone.

We will give you an overall
tool of the interface showing you how to navigate through it
and utilize its features with ease. Followed by this, will be a step-by-step
guide to solve your use cases. Let's get started.

In this video we will cover, introduction to Google AI Studio,
its interface tour and executing sample prompts. Google AI Studio is designed
to make AI development accessible, efficient, and
enjoyable for everyone. This powerful platform combines
an intuitive interface with advanced tools

allowing users to create, manage, and
deploy AI products seamlessly. Whether you are just starting
or an experienced developer, Google AI Studio offers
something for everyone. By simplifying complex
tasks and speeding up innovation, it empowers you to bring your AI ideas
to life with minimal hassle.

Now, let's explore Google AI Studio. You just have to go to
Google search engine and type Google AI Studio and click on the first link. You will land up on this homepage, where it states directly Sign in to Google AI Studio.

I'm clicking on this. Once I do this, it will open
the Google AI Studio interface. Since we are logging in, signing in
for the first time, it will ask for the consent which we have
to tick and press on, continue.

Once we do this, we get to see
these options and features. Let us understand one by one,
what all does it support, how does it work. First, let's talk about the model. If you could see it's automatically
selected Gemini 1.5 Flash.

If you click on this drop down, you will get to see 1.0 Pro, 1.5 Pro, 1.5 Pro Experimental 0801. You have Gemini 1.5 Flash
followed by the upcoming preview of the releases

2B, 2 9B, and 2 27B. Google is still working
on the release of version 2. Now, let's focus on understanding what is the difference between Gemini 1.5 Pro and Gemini 1.5 Flash.

As the name suggests,
flash means very fast. So if you have a high, huge content you do have to work with,
then you can select 1.5 Flash. If it is a standard content
which you want to work with, it can be Gemini 1.5 Pro.

This is the major difference
and very common examples or simple learning processes which
you want to understand, simple model tuning or working with
a basic ML model or data, you can go with 1.0 Pro. Generally, I prefer 1.5 Pro

or 1.5 Flash accordingly
as per the requirement. So let me change this to 1.5 Pro. Now, hopefully the model concept is clear. Let's come to tokens. So token count, the number of tokens taken in order to process your
statements will be counted here.

Generally, 1 million tokens are available to work with for free in Google AI Studio. You have other advanced settings. If you're working with any code executions or JSON files, you can just accordingly change this across.

The next option you can see is temperature. There is a range for this starting from zero up to two. What is this temperature? Is it normal heat or cold temperature? No. It is how frequently the tokens can be used, and in what frequency we will select it with the help of temperature.

So it is always good to keep it at one. It's neither zero nor two, it's in the middle. It's one. Hope the settings of this run model, temperature, and advanced settings are clear. Now, let's come towards our left side of the screen.

What do we have here? We can create a new prompt. We can create a new tuned model. We have our libraries. If you have worked before, you can have that complete details in the library. You could see an option to allow Drive access.

If your data or any information present in the Google Drive can be associated by giving permission, by clicking on this. Getting started, you have all the documentation, you have a prompt gallery, what kind of prompts you can give, how you have to give them, you have a Gemini cookbook.

How the Gemini model is working, what are the different elements it has? So you have documentation on this and you have a discourse forum. If you want to discuss about something, you can discuss in the forum, and a very important thing what you find it here is built with Vertex AI on Google Cloud.

Whatever the Google product is, it will go up to the cloud, end of the day. Be it Gmail, be it Drive, be it Slides, Colab, everything lands up in the cloud because it will work in that way.

What is this Vertex AI and why is it inside Google AI Studio? First and foremost, the basic difference is that both will work for ML models themselves. But Google AI Studio is free of cost. It can let you work with 1 million tokens for free,

wherein Vertex AI it will always give you an enterprise edition. That means you have to pay for what you work for. Accordingly, there are different features as well, which is more convenient for the core developers or the people who are into the domain of machine learning

or artificial intelligence, can use this Vertex AI by paying the license amount. So if you are a professional, if you are solely working on this, you can select Vertex. Once you click, where will it land? It will land to a console of a Google Cloud

and it will ask you to agree and continue. Before doing that it states, '$150 credit now to try Google Cloud Vertex AI with Gemini models for 90 days.' So, you have certain conditions and a pay-as-you-go model is followed here.

You can agree and continue for now, just to give an overview of how a free trial of Vertex AI looks like. So here we are in the interface of Vertex AI. We agree and continue. We don't want the welcome tour. So I'm canceling this.

This is how the interface looks like. Again, you have a dashboard,
you have pipelines, which you can maintain. You have a workbench, you can
collaborate with many APIs as possible, you can try with Gemini as well. You can connect Vertex AI
with a Gemini model.

You have different kinds
of language, vision, speech prompt gallery, and tuning models available. Also, the basic extensions
and advanced levels of datasets can also be used, experiments can
be done, training is available, you have metadata connected to it,
and also you have a marketplace.

So, this is how Vertex AI Studio will look like. Now, let's switch back
to our Google AI Studio. Now, let's start with our first prompt. You have to click on the create new prompt and if you want to change the name,
you can just go on, edit,

say Demo 1 and click on save. Now, the interface is
ready to take your prompts. But there is one option
here, System Instructions. Why is this used? This is just to set
the tone of the prompt, the responses we get and the
style of instructions we give to the model.

We are just trying to
set a context before we start communication with the model. I'm trying to tell the system, the response we receive should be poetic. So, it is just a sample. I'll just give 'hi' as a prompt and run.

Once I do this, it has taken two tokens in order to work. Just to say a reply for 'hi', it is giving in a poetic manner. It is giving a complete, a whisper soft, a greeting brief 'hi', you say, a single leaf.

So it is trying to give
a poetic response to it. So, if I say, 'How are you?', then it will try to give us
an answer in a poetic manner itself. You could see the token count here. It's moving on and on.

The time it processes the number of tokens
it will use, those are kind of building blocks in order to process the statements
we give to the system. So you can see for just one, 'How are you?',
giving us four lines of answer?

Because I set the context that the answer
should be in a poetic manner. The theme should be put. So this is gender prompting, which you can do by setting this system
instructions with Google AI Studio. Now let's move on to a new build model.

So what is this all about? Before that? See, it does giving us 'Do we want to save this? Yes.' If you want to click on okay, it will save us 'demo one'. Once you select the Newton model,
this is where you land.

You have an option to create a structure from the prompt
which we gave was a normal prompt. You have a second form of prompt
that is structured prompt. So I click on 'structure prompt' again,
if you want to change the name for this,

you can really put the name. I'm using the name 'demo' to. I'll save it. You could see
there are those sections here. So let's section this 'input' then
'output'. Later you have 'best will prompt'. You have the first three in the system,
then you have 'best the system'.

This is the concept
you are doing in a structured manner. You're not just giving a random
prompt command. Now let's set a context
or style of instruction. The first thing, the structure prompting. Let's set an optional tone
or system information with

you want to save it across? I'm getting the overall context. I'm working as a news reporter. I need the normal news to be displayed in a more excited manner or as a breaking news. So when I do this, I'm setting up context.

I'm just saving. For now,
it has been saved. Now let's start giving that input and specified output to the model
which is trying to train the model. So this time Olympics is held in Paris, right? Another input.

The player from Turkey won a silver medal in shooting. So I give this next it will try to suggest to certain things
it will just leave us a this. Now let's type the response as the 'Did you know the Olympics is held in Paris?'

exciting right so this is a normal sentence
which I'm using then well a player from Turkey won a silver medal in shooting. Isn't it exciting? So I've just typed
certain inputs and outputs that is training a model
out of 500 samples, obviously to samples.

And so you can give a third one as the 'India lost medals in badminton, right?' You can just give us simple input and 'Oh my God, India missed chance to bag medals in badminton.' It is just exaggerating or giving an excited output for us.

Normal sentences. This is how I'm trying to beat. Then we will test the inputs. I see India won gold in wrestling. Now let's try to run to get the output. I'm clicking on run. It gives them exaggeration here.

Boom! India takes gold in wrestling. This is huge. What a performance! We are trying to exaggerate. First we try to train this model and then it is not trained across. It is also giving... As we mentioned, we need the output.

An excited man. So this is the structure prompting. Now what do we do? We try to delete this, right? The outputs will try to delete and we will just give the same input. India won one goal in wrestling and run this 100%.

It will follow the system instruction which we have given at the top, not the model, because we don't have any samples here in order to compare. So it is a big paragraph breaking the rules. India claims gold in wrestling, so it just tries to give good information.

Often use. So this is how the structure prompting looks. So now we have understood the normal prompt and the structure prompt. Let's click on new, the old model. Before that, if you want to save, you can click on okay

and you can produce a model. You can import. Either it should be RC, as we find on a Google sheet file and they generally recommend how to find it. Examples in the cheat are available. If you want the modeling guide, you can have a look with this document.

Now if you click on the select data source, you have structure Prompt, which is the motto which I created recently that has also been available here. And you have news headlines and summarization, which are the samples. So I prefer taking news headlines.

Once you do that, you can see the prefix column available, you can select the prefix column, input and output. This is how it will be and just it is a sample and then you can change sample your own model. Once you do this,

you can give a description if you want to and we are not going to touch this. Advanced settings. How many tuning epochs should be? What is the learning rate multiplier in epochs and stuff? Let it be as it has been set and you have two options here.

That is one point or Pro not not one and 1.5 epochs, not not one tuning. So generally we were working with the help of Pro. So I'm using this Pro again. If you want a tune, you can click on the build button.

Once you click on the build button, you could find a sample build model available in your library. Once you click on the sample model towards your left, you can see the running model results. It takes some time in order to load this and you have to wait for 5 to 6 minutes. Why?

The reason being I selected it took certain time in order to load this box. See, the graph is going in the right way, but the value is 1.2. Actually, it's supposed to be at least 4 to 5.

That is what we consider a good value. But this is how model tuning looks like in Germany. As a studio, you can also see less than one minute left in order to finish this process and it has all the information available, the box size, the epochs, 

the learning rate, the examples it has been doing for. So the graph, as the last graph, has also been put up here. This is the overall view of model tuning which you can do in Google Studio. Coming back to the main interface, 

you have an option here, which is plus. If you click on plus, you have various availability of options in order to choose. But that is my right, upload to try record audio sample you may just and sample video. So let's check out with the help of sample video. 

There are predefined videos here. I'm trying to select one of the video and add a prompt: What is this video all about? So I click on Run. Once we do this, it will try to analyze the multimedia, that is the video and give us the details. 

What is this 10 minutes of video dealing with? It takes certain time. If you could see the tokens of this taking in order to process. This is huge because of the 10 minutes of video. So the response is in two lines. 

It is starting to give a gist of what does this video all about. So you can also upload the videos from your local system and understand what is happening in that video. It has the capability to analyze the video as well, 

and you can change the model here in order to immediately get the responses. See Sample video. I just gave them the same video or another
10-minute video. How does this particular viral load and how fast will it work? What is in the video exactly?

I'm trying to give this option. I'm clicking on run. I'm just changing the model and checking
how this works? While we were executing the second video,
we turned this model in 1.5 flash. Let's see what is the difference here?

In the first video,
it took around 30 seconds in order to work here.
We got 40 seconds and it depends upon what is the content with you,
What is the file size of the video? Everything. The more content you have, it's
good to use flash, right?

So just a simple comparison
between Pro and flash. You can do the same thing
for sample images. If you click on
it means you have total here. So if I ask what is this, it will try to give us the information.

Let's see how much time this will take
when it doesn't flash. It took around
28 seconds in order to save. This is a picture of a startup. So this is how you can load your images,
your videos, your multimedia,

and understand
you have to first upload to the drive. Or you can directly take the information
from the drive. So this is how the Prompting works
in Google Studio. So you have to choose it
from the available options that is from your drive
or the sample we use on record.

The audio as it is. So we have understood a few
of the information with regards to Google. As to your first thing we saw,
how does this interface look like? What is Google Ad Studio all about? Then we walked through a few prompts,

which is a general prompting
and a structured prompting. Later, we also created a sample model and also
we saw the graph of the same after tuning. But thank you for watching this company
tour of Google Studio. We hope you found it
informative and helpful.

If you have any questions or need further
assistance, feel free to leave a comment. Don't forget to like, subscribe
and hit the notification bell for more tutorials and guides. Happy learning
and see you in the next video.

Google AI Studio | How to use | Complete Guide

Great Learning

Google AI 스튜디오는 사용자 친화적인 인터페이스와 강력한 도구로 AI 개발을 지원합니다.

이 동영상은 인터페이스 둘러보기와 사용 사례에 대한 단계별 가이드를 제공합니다.

이 플랫폼은 AI 개발에 대한 접근성, 효율성, 즐거움을 제공하는 것을 목표로 합니다.

초보자와 숙련된 개발자 모두 복잡한 작업을 간소화하고 혁신의 속도를 높일 수 있습니다.

Google AI 스튜디오 인터페이스 탐색 및 기능 이해하기

Google 검색 엔진을 통해 Google AI 스튜디오에 액세스합니다.

모델 옵션에 대한 설명입니다: Gemini 1.5 Flash, 1.0 Pro, 1.5 Pro 및 출시 예정 버전.

콘텐츠 크기에 따른 Gemini 1.5 Pro와 Gemini 1.5 Flash의 차이점.

토큰 수, 온도 설정 및 고급 설정에 대한 설명입니다.

인터페이스 옵션 개요: 프롬프트 만들기, 조정된 모델, 라이브러리, 드라이브 액세스, 문서 및 포럼.

구글 AI 스튜디오와 버텍스 AI의 비교

Google AI Studio는 1백만 토큰으로 무료이며, Vertex AI는 유료 엔터프라이즈 에디션입니다.

Vertex AI는 전문 머신러닝 개발자를 위한 더 많은 기능을 제공합니다.

Google Cloud 콘솔 및 무료 평가판을 통해 Vertex AI에 액세스합니다.

Vertex AI 인터페이스 개요: 대시보드, 파이프라인, 워크벤치, API 및 모델 연결.

Google AI 스튜디오에서 프롬프트 만들기 및 사용

새 프롬프트를 만들고 응답 스타일을 정의하는 시스템 지침을 설정합니다.

시스템 지침을 사용하여 프롬프트에 대한 시적 응답의 예입니다.

프롬프트 저장 및 구조화된 프롬프트 소개.

구조화된 프롬프트에 대한 컨텍스트 및 스타일 설정(예: 뉴스 보도).

구조화된 프롬프트(입력 및 출력 예제)로 모델을 훈련합니다.

모델 구축을 위한 데이터 소스 선택하기(예: 구조화된 프롬프트, 뉴스 헤드라인).

모델 튜닝을 위한 고급 설정(에포크, 학습 속도)을 설정합니다.

모델 구축 및 프로세스 모니터링

모델 튜닝 결과 및 그래프를 검토합니다.

멀티미디어 작업(비디오 및 이미지 분석)

프롬프트를 사용하여 샘플 동영상을 분석하고 처리 시간을 관찰합니다.

서로 다른 모델 간의 처리 속도 비교(Pro와 플래시).

프롬프트를 사용하여 샘플 이미지를 분석합니다.

멀티미디어 데이터 업로드 및 분석을 위한 옵션 요약입니다.