#GPT-4 Demo | GPT3 vs GPT4: O futuro da Inteligência Artificial

##Introdução

##O que é GPT-3?
###Como funciona o GPT-3?
###Principais recursos do GPT-3

##GPT-4: O que é?
###Funcionamento do GPT-4
###Diferenças entre o GPT-3 e GPT-4

##Como a IA afeta a indústria de conteúdo?
###Desafios da redação de conteúdo em IA

##Como criar um conteúdo engajador com IA?
###Dicas para escrever conteúdo envolvente com IA

##Como criar uma estratégia de SEO com IA?
###Benefícios do uso de IA para SEO
###Como usar IA para pesquisa de palavras-chave

##Redes sociais: como maximizar o engajamento com IA?
###Uso da IA na segmentação de público-alvo em redes sociais
###Como usar IA para criar conteúdo nas redes sociais

##Por que a construção de uma lista de e-mails é importante em sua estratégia de marketing?
###Vantagens da construção de uma lista de e-mails com IA
###Dicas para criar uma lista de e-mails com IA

##Leads orgânicos x Leads pagos: Qual a diferença?
###Comparação entre leads orgânicos e pagos
###Por que os leads orgânicos são importantes para seu negócio

##Como gerar leads orgânicos para seu negócio?
###Como a IA pode ajudar a gerar mais leads orgânicos
###Estratégias para gerar leads orgânicos com IA

##Conclusão

##FAQs
###O GPT-4 é melhor que o GPT-3?
###A IA pode ajudar a criar conteúdo original e envolvente?
###Como a IA pode ajudar nas estratégias de SEO?
###É possível utilizar IA para engajar o público-alvo nas redes sociais?
###Por que é necessário construir uma lista de e-mails?

br>OpenAI (the company behind ChatGPT) has released a new model called GPT-4. The current free version of ChatGPT is powered by GPT-3, with a premium version of ChatGPT you can select GPT-4 as a model. I am going to demo GPT-4 by showcasing some of its capabilities. We will also compare GPT-4 with GPT-3 and understand the differences.

Do you want to learn technology from me? Check for my affordable video courses.

Need help building software or data analytics/AI solutions? My company can help. Click on the Contact button on that website.

🎥 Codebasics Hindi channel:

#️⃣ Social Media #️⃣
🔗 Discord:
📸 Dhaval’s Personal Instagram:
📸 Codebasics Instagram:
🔊 Facebook:
📱 Twitter:
📝 Linkedin (Personal):
📝 Linkedin (Codebasics):
🔗 Patreon:

Este vídeo foi indexado através do Youtube link da fonte
gpt 4 ,

yt:cc=on,gpt-4,gpt4,gpt4 demo,gpt4 news,gpt 4 openai,gpt3 vs gpt4,gpt4 vs gpt3 ,

https://www.youtubepp.com/watch?v=IJCcHDLv7ss ,

Open AI the company behind chat GPT has released a latest model called gpt4 so if you are logged into chat GPT Right Now the default model is gpt3 I have a Premium plan for chat GPT it’s called GPT plus and due to that I am able to access gpd4 version okay so if

You have non-paid free plan you will see on the left hand side this kind of upgrade plan button which will appear somewhere here and when you upgrade you will have access to gpd4 what I’m going to do is show you the demo of gpt4 I will compare it with gpd3 so that you

Understand what kind of improvements they have made so right now I have a GPD 3.5 and I’m going to give some tasks to it such as show me the recipe of Samosa which is an Indian delicacy in five lines where every word starts with a okay and while it is doing

That let me open a new prompt here and I will do the same thing using gpt4 so I will select gpd4 here and see here 3.5 it give this particular recipe but see every word is not starting with a here’s the ajwain seed so as a oil it is

Not very very accurate but now let me do the same thing with gpt4 see this is a hard task if I ask you to write samosas recipe such that every word begins with a a it’s going to be difficult even for humans and look at how gpd4 is performing it is being

Innovative and it is just using words such that every later the first letter of every word is a see it did pretty decent job okay and I can say that now do the same thing but the first letter should be B see buy basic bread boiled baby potatoes

Bold Bishop suede Bland baby producer Bland balance bold bits see you can imagine for humans also this will be difficult and gpt4 model is doing pretty well here now the second task that I have given to GPD 3.5 is I have three people in my organization with different availability I want to

Know the common time uh if I want to schedule a meeting or between these three people and these people are here so Andrew John Hannah they they have this much time schedule where green slots shows the free time that they have and as a human when you look at this kind of

Diagram you can immediately figure out that 12 to 12 30 is the common time slot if you want to schedule a meeting between these three people this is a common problem in any organization where they want to schedule a between meeting between multiple people and they want to

Figure out a common time slot okay so if you have any calendar type of function this is a very very common useful feature and here for my condition the time between 12 to 12 30 is the only half an hour slot that is available okay

Now when I gave this that to uh gpd3 model it gives some confusing answers you know start the meeting at four this works for these two people and he can join it’s some of these things are not accurate okay if you give the same prompt to gpt4 let’s see what happens

While it is typing the answer you can read the question and this question I have translated into this Excel file okay in terms of this time slots and obviously 12 to 12 30 is on is the only time slot and it actually gave the correct answer for the option number one

It is perfect but option number two is wrong 230 to 3 3 pm it says Andre and Jonah are available but see John is not available from 2 30 to 3 P.M two is wrong Joanne is not aware label from 2 30 to 3 P.M and when you give this kind

Of feedback it will adapt it will realize that the only option you have is 12 to 12 30 PM so overall GPT 4 has a better reasoning capabilities the next question I have given is I want you to explain me the plot of Avengers the endgame movie you might have watched

That movie in a rhyming poem of four sentences I asked it to write a poem Only in four sentences but see it it took so many sentences but if you give the same thing here it will perform better so overall GPT 4 is performing better it has a lot more parameters than

Gpt3 open AI has not revealed what how many parameters are there and they have not given any of the internal details uh but overall you can see it is better see uh rhyming poem a battle for ages Infinity is worth wages Avengers Assemble till Victory is one Whatever

Seek wages cages done one so the it’s the rhyming poem of four sentences same thing with gpd3 the answer was not that great as a next step I am going to ask a difficult puzzle I remember uh this puzzle from uh puzzles to puzzle you book by sakunzla Devi When I Was

Preparing for interviews uh many years back I used to practice puzzles from this book and it is giving some answers see gpd3 is giving the answer which is 10 minutes okay and if I give the same question to gpd4 it will give some answer by the way the correct answer for

This question is 26 minutes okay you can Google it the correct answer is 26 minutes it’s not very very easy puzzle uh as a human also you might find it uh challenging but any smart person will will solve it so it’s not like impossible but you can see that

Gpt3 definitely it gave a wrong answer 10 minutes until 6 o’clock it’s a wrong answer and here uh negative reason gpd4 gave a negative result so folks you understand this is not not any Silver Bullet it can give you wrong results and sometimes it gives you result with so

Much confidence so that you can’t blindly trust this AI okay these AI models will always have their limitations now here is the website for gpt4 and there are some example prompts where you can also say okay give me a plot of Cinderella where every words first

Letter begins from A to Z so you can say A B C D E F G H this is pretty insane let me give the exact same question here and see how it does see it did not perform well so I don’t know maybe gpt4 website shows this but in reality

This doesn’t work uh but yes all of these things I I just showed the example of a meeting time that works well it also performed well in various tests there is biology Olympiad that’s an exam and GPT 4 came in 99th percentile which is pretty good performance whereas chat

GPT GPT 3 Model previous model perform it got 31st percentile uniform bar exam gbd4 got in 19th percentile 19th percentile is like top top you know top performer so obviously it is having a dazzling performance it can also work with images folks right now in GPT you can’t try

Images but it can work on visual images so let’s say you give this kind of image and say what can I make with these ingredients it will say there are many options you can make so there is eggs that is I think some kind of flour see pancakes french toast omelette you

Can make all these recipes so this model understands an image very well and it can actually tell you what you can make out of it so let’s say if you have a fridge you can take a picture of your fridge and it will recognize the ingredients or you know vegetables milk

Whatever you have in that fridge and it can give you an option what kind of recipe you can make I can say I am and I’m a vegetarian I love Indian food here’s the picture of my fridge these are the ingredients show me the options of recipes and maybe I can send that

List to my wife because my wife she everyday asked me what she she should make for the dinner and that’s the most difficult question to answer so maybe next time I will take GPT 4’s help and then it can handle longer context so when you talk about natural language

Processing you know like how much a window of words you can process here it can process up to 25 000 words of text which is much longer context that you have and it can make the NLP performance are super amazing I’m also going to link an article from TechCrunch where they have highlighted

Main capabilities number one is it can see and understand images gpt3 cannot understand images it can be harder to trick whereas gpd3 you can trick it sometimes gpd4 has a longer memory it is more multilingual so it has support in different languages all right that was a quick demo whenever they release the

Support for visual inputs I will make another video If you like this video folks give it a thumbs up and share it with your friends so that you are up to date with the latest and greatest things happening in the world of AI

,00:00 open AI the company behind chat GPT has
00:03 released a latest model called gpt4 so
00:07 if you are logged into chat GPT Right
00:09 Now the default model is gpt3
00:12 I have a Premium plan for chat GPT it’s
00:16 called GPT plus and due to that I am
00:19 able to access gpd4 version okay so if
00:23 you have non-paid free plan you will see
00:26 on the left hand side this kind of
00:28 upgrade plan button which will appear
00:30 somewhere here and when you upgrade you
00:33 will have access to gpd4 what I’m going
00:36 to do is show you the demo of gpt4 I
00:40 will compare it with gpd3 so that you
00:42 understand what kind of improvements
00:44 they have made so right now I have a GPD
00:48 3.5 and I’m going to
00:51 give some tasks to it such as show me
00:54 the recipe of Samosa which is an Indian
00:56 delicacy in five lines where every word
00:59 starts with a okay and while it is doing
01:03 that let me open a new prompt here and I
01:07 will do the same thing using gpt4 so I
01:10 will select gpd4 here and see here
01:15 3.5 it give this particular recipe but
01:20 see every word is not starting with a
01:23 here’s the ajwain seed so as a oil it is
01:28 not very very accurate but now let me do
01:31 the same thing with gpt4
01:34 see this is a hard task if I ask you to
01:36 write samosas recipe such that every
01:39 word begins with a a it’s going to be
01:42 difficult even for humans and look at
01:44 how gpd4 is performing it is being
01:47 Innovative and it is just using words
01:50 such that every later the first letter
01:54 of every word is a
01:58 see it did pretty decent job okay and I
02:02 can say that now
02:05 do the same thing but
02:09 the first letter should be B
02:16 see buy basic bread boiled baby potatoes
02:19 bold Bishop suede Bland baby producer
02:22 Bland balance bold bits
02:25 see you can imagine for humans also this
02:27 will be difficult and gpt4 model is
02:31 doing pretty well here
02:32 now the second task that I have given to
02:35 GPD 3.5 is
02:38 I have three people in my organization
02:40 with different availability I want to
02:42 know the common time uh if I want to
02:46 schedule a meeting or between these
02:48 three people and these people are here
02:52 so Andrew John Hannah they they have
02:56 this much time schedule where green
02:57 slots shows the free time that they have
03:01 and as a human when you look at this
03:03 kind of
03:05 diagram you can immediately figure out
03:07 that 12 to 12 30 is the common time slot
03:11 if you want to schedule a meeting
03:12 between these three people this is a
03:15 common problem in any organization where
03:17 they want to schedule a between meeting
03:19 between multiple people and they want to
03:21 figure out a common time slot okay so if
03:23 you have any calendar type of function
03:24 this is a very very common useful
03:26 feature and here for my condition the
03:30 time between 12 to 12 30 is the only
03:32 half an hour slot that is available okay
03:34 now when I gave this that to uh gpd3
03:39 model it gives some confusing answers
03:42 you know start the meeting at four this
03:44 works for these two people and he can
03:47 join it’s some of these things are not
03:49 accurate okay
03:50 if you give the same prompt to gpt4
03:55 let’s see what happens
03:56 while it is typing the answer you can
03:58 read the question and this question I
04:01 have translated into this Excel file
04:04 okay in terms of this time slots and
04:05 obviously 12 to 12 30 is on is the only
04:08 time slot and it actually gave the
04:10 correct answer for the option number one
04:13 it is perfect but option number two is
04:15 wrong 230 to 3 3 pm it says Andre and
04:19 Jonah are available but see John is not
04:22 available from 2 30 to 3 P.M two is
04:25 wrong Joanne is not aware label from 2
04:31 30 to 3 P.M and when you give this kind
04:35 of feedback it will adapt it will
04:37 realize
04:38 that the only option you have is 12 to
04:41 12 30 PM so overall GPT 4 has a better
04:45 reasoning capabilities
04:47 the next question I have given is I want
04:50 you to explain me the plot of Avengers
04:52 the endgame movie you might have watched
04:54 that movie in a rhyming poem of four
04:57 sentences I asked it to write a poem
04:59 Only in four sentences but see it it
05:02 took so many sentences but if you give
05:04 the same thing here it will perform
05:08 better so overall GPT 4 is performing
05:11 better it has a lot more parameters than
05:14 gpt3 open AI has not revealed what how
05:19 many parameters are there and they have
05:20 not given any of the internal details uh
05:24 but overall you can see it is better see
05:27 uh rhyming poem a battle for ages
05:30 Infinity is worth wages Avengers
05:32 Assemble till Victory is one Whatever
05:35 seek wages cages done one so the it’s
05:38 the rhyming poem of four sentences same
05:40 thing with gpd3 the answer was not that
05:43 great
05:45 as a next step I am going to ask a
05:47 difficult puzzle I remember uh this
05:49 puzzle from uh puzzles to puzzle you
05:52 book by sakunzla Devi When I Was
05:54 preparing for interviews uh many years
05:57 back I used to practice puzzles from
05:59 this book and it is giving some answers
06:03 see gpd3 is giving the answer which is
06:05 10 minutes okay and if I give the same
06:08 question to gpd4 it will give some
06:11 answer by the way the correct answer for
06:13 this question is 26 minutes okay you can
06:16 Google it the correct answer is 26
06:19 minutes it’s not very very easy puzzle
06:21 uh as a human also you might find it uh
06:25 challenging but any smart person will
06:27 will solve it so it’s not like
06:29 impossible but you can see that
06:32 gpt3 definitely it gave a wrong answer
06:35 10 minutes until 6 o’clock it’s a wrong
06:37 answer and here uh negative reason
06:41 gpd4 gave a negative result so folks you
06:44 understand this is not not any Silver
06:46 Bullet it can give you wrong results and
06:49 sometimes it gives you result with so
06:50 much confidence so that you can’t
06:53 blindly trust this AI okay these AI
06:56 models will always have their
06:58 limitations
06:59 now here is the website for gpt4 and
07:02 there are some example prompts where you
07:04 can also say okay give me a plot of
07:06 Cinderella where every words first
07:09 letter begins from A to Z so you can say
07:12 A B C D E F G H this is pretty insane
07:18 let me give the exact same question here
07:20 and see how it does see it did not
07:22 perform well so I don’t know maybe
07:26 gpt4 website shows this but in reality
07:29 this doesn’t work uh but yes all of
07:32 these things I I just showed the example
07:34 of a meeting time that works well it
07:37 also performed well in various tests
07:39 there is biology Olympiad that’s an exam
07:42 and GPT 4 came in 99th percentile which
07:46 is pretty good performance whereas chat
07:48 GPT GPT 3 Model previous model perform
07:51 it got 31st percentile
07:55 uniform bar exam gbd4 got in 19th
07:59 percentile 19th percentile is like top
08:01 top you know top performer so obviously
08:06 it is having a dazzling performance it
08:09 can also work with images
08:11 folks right now in GPT you can’t try
08:14 images but it can work on visual images
08:18 so let’s say you give this kind of image
08:20 and say what can I make with these
08:22 ingredients it will say there are many
08:24 options you can make so there is
08:26 eggs that is I think some kind of flour
08:29 see pancakes french toast omelette you
08:33 can make all these recipes so this model
08:36 understands an image very well and it
08:39 can actually tell you what you can make
08:41 out of it so let’s say if you have a
08:43 fridge you can take a picture of your
08:44 fridge and it will recognize the
08:47 ingredients or you know vegetables milk
08:49 whatever you have in that fridge and it
08:51 can give you an option what kind of
08:53 recipe you can make I can say I am and
08:56 I’m a vegetarian I love Indian food
08:58 here’s the picture of my fridge these
09:00 are the ingredients show me the options
09:02 of recipes and maybe I can send that
09:04 list to my wife because my wife she
09:07 everyday asked me what she she should
09:09 make for the dinner and that’s the most
09:11 difficult question to answer so maybe
09:13 next time I will take GPT 4’s help and
09:17 then it can handle longer context so
09:20 when you talk about natural language
09:22 processing you know like how much a
09:25 window of words you can process here it
09:28 can process up to 25
09:30 000 words of text which is much longer
09:33 context that you have and it can make
09:37 the NLP performance are super amazing
09:41 I’m also going to link an article from
09:43 TechCrunch where they have highlighted
09:45 main capabilities number one is it can
09:47 see and understand images gpt3 cannot
09:50 understand images it can be harder to
09:54 trick whereas gpd3 you can trick it
09:57 sometimes gpd4 has a longer memory it is
10:00 more multilingual so it has support in
10:03 different languages all right that was a
10:06 quick demo whenever they release the
10:08 support for visual inputs I will make
10:11 another video If you like this video
10:13 folks give it a thumbs up and share it
10:15 with your friends so that you are up to
10:16 date with the latest and greatest things
10:18 happening in the world of AI
, , , #GPT4 #Demo #GPT3 #GPT4 , [agora]