# Título: GPT-4 Developer Livestream – Como Gerar Leads Orgânicos?
## Introdução
### O que é GPT-4?
### O que são leads orgânicos?
### A diferença entre leads orgânicos e leads pagos

## Como Gerar Leads Orgânicos?
### Otimização para os mecanismos de busca
#### Pesquisa de Palavras-Chave
#### Técnicas de otimização no site
### Criação de Conteúdo Engajador
#### Conteúdo de qualidade e relevante
#### Uso de imagens e vídeos
### Utilização das Redes Sociais
#### Plataformas a serem utilizadas
#### Dicas para engajar seguidores
### Construção de uma lista de e-mails
#### Importância da construção de lista de e-mails
#### Estratégias para aumentar a lista de e-mails

## Como Otimizar seu Site para SEO?
### Técnicas de otimização no site
#### Uso de meta-descrições e meta-tags
#### Estrutura de URL amigável
### Criação de Conteúdo de Qualidade
#### Uso de palavras-chave estrategicamente
#### Uso de títulos apropriados e subtítulos
### Construção de Backlinks
#### Valor de backlinks de qualidade
#### Estratégias para a construção de backlinks

## Como Criar Conteúdo Engajador?
### Encontre Seu Público-alvo
#### Identifique o público-alvo para o conteúdo
#### Crie conteúdo relevante ao público-alvo
### Use Imagens e Vídeos
#### Benefícios do uso de imagens e vídeos
#### Melhores práticas ao utilizar imagens e vídeos
### Use Títulos Atraentes
#### Títulos atraentes e informativos
#### Uso de palavras-chave para o título

## Como Utilizar as Redes Sociais?
### Escolha as Plataformas Adequadas
#### Identifique as plataformas que sua audiência frequenta
#### Melhores práticas por plataforma
### Crie Conteúdo Atraente
#### Uso de imagens e vídeos
#### Uso de títulos atraentes
### Interaja com seu público
#### Respondendo perguntas e comentários
#### Envolvendo seguidores em sua página

## Como Construir uma Lista de E-mails?
### Importância da Construção de Lista de E-mails
#### Importância de ter uma lista de contatos
#### Como melhorar a taxa de abertura dos e-mails
### Estratégias para Aumentar a Lista de E-mails
#### Ofereça algo em troca de um e-mail
#### Uso de pop-ups ou formulários na página

## Conclusão
### Recapitulando os Principais Pontos
### Dicas Finais

## FAQs
### 1. Por que leads orgânicos são importantes para o meu negócio?
### 2. Como faço para identificar as palavras-chave que devo usar em minha estratégia de SEO?
### 3. Como faço para manter meu conteúdo relevante?
### 4. Como utilizar vídeos para aumentar o engajamento em minha plataforma?
### 5. Por que é importante manter uma lista de e-mails para contato com meus clientes?

br>Join Greg Brockman, President and Co-Founder of OpenAI, at 1 pm PT for a developer demo showcasing GPT-4 and some of its capabilities/limitations.

Join the conversation on Discord here: discord.gg/openai. We’ll be taking audience input from #gpt4-demo-suggestions.

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

[vid_tags] ,

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

Foreign did the gpd4 developer demo live stream honestly it’s kind of hard for me to believe that this day is here open AI has been building this technology really since we started the company but for the past two years we’ve been really focused on delivering gpt4 that started with rebuilding our entire

Training stack actually training the model and then seeing what it was capable of trying to figure out its capabilities its risks working with Partners in order to test it in real world scenarios really tuning Its Behavior optimizing the model getting it available so that you can use it and so today our

Goal is to show you a little bit of how to make gbto4 shine how to really get the most out of it you know where it’s kind of you know weaknesses are where we’re still working on it and just how to really use it as a good tool a good partner

Um so if you’re interested in participating in the Stream uh that if you go to our Discord so that’s discord.gg openai there’s comments in there and we’ll take a couple of audience suggestions so the first thing I want to show you is the first task that gpd4 could do that

We never really got 3.5 to do and the way to think about this is all throughout training that you know you’re constantly doing all this work it’s 2 A.M the pager goes off you fix the model and you’re always wondering is it gonna work is all this effort actually going to pan

Out and so we all had a pet task that we really liked and that we would all individually be trying to see is the model capable of it now and I’m going to show you the first one that we had a success for four but never really got there for 3.5

So I’m just going to copy the top of our blog post from today going to paste it into our Playground now this is our new chat completions playground that came out two weeks ago I’m going to show you first with GPT 3.5 4 has the same API to it the same playground

The way that it works is you have a system message where you explain to the model what it’s supposed to do and we’ve made these models very steerable so you can provide it with really any instruction you want whatever you dream up and the model will adhere to it

Pretty well and in the future it will get increasingly increasingly powerful at steering the model very reliably you can then paste whatever you want as a user the model will return messages as an assistant and the way to think of it is that we’re moving away from sort of

Just raw text in raw text out where you can’t tell where different parts of the conversation come from but towards this much more structured format that gives the model the opportunity to know well this is the user asking me to do something that the developer didn’t attend I should listen to the developer

Here all right so now time to actually show you the task that I’m referring to so everyone’s familiar with summarize this let’s say article into a sentence okay getting a little more specific uh but where every word begins with G so this is 3.5 let’s see what it does

Yeah it kind of didn’t even try just gave up on the task this is pretty typical for 3.5 trying to do this particular kind of task if it’s you know sort of a very kind of stilted article or something like that maybe it can succeed but for the most part 3.5 just

Gives up but let’s try the exact same prompt the exact same system message in gbt4 so kind of borderline whether you want to count AI or not but so let’s say AI doesn’t count that’s cheating so fair enough the model happily accepts my feedback so now to make sure it’s not just good

For G’s I’d like to turn this over to the audience I’ll take a suggestion on what letter to try next in the meanwhile while I’m waiting for our moderators to pick the lucky lucky letter I will give a try with a um but in this case I’ll say gpd4 is fine why not

Also pretty good summary so I’ll hop over to our Discord all right wow if people are are being a little ambitious here I’m really trying to put the model through the paces we’re going to try Q uh which if you think about this for a moment I want the audience to

Really think about how would you do a summary of this article that all starts with Q it’s not easy it’s pretty good that’s pretty good all right so I’ve shown you summarizing an existing article I want to show you how you can flexibly combine ideas between different articles so I’m going

To take this article that was on Hacker News yesterday copy paste it into the same conversation so it has all the context of what we’re just doing I’m going to say find one common theme between this article and the gpd4 blog so this is an article about Pinecone

Which is a python web app development framework and it’s making the technology more accessible user friendly if you don’t think that was insightful enough you can always give some feedback and say that was not insightful enough please no I’ll just even just leave it there leave it up to the model to decide

So Bridging the Gap between powerful technology and practical applications seems not bad and of course you can ask for any other kind of task you want using its flexible language understanding and synthesis you can ask for something like now turn the GT4 blog post into a rhyming poem

Picked up on open AI evalues open source for all helping to guide answering the call which by the way if you’d like to contribute to this model please give us evals we have an open source evaluation framework that will help us guide and all of our users understand what the

Model is capable of and to take it to the next level so there we go this is consuming existing content using gpt4 with with a little bit of creativity on top but next I want to show you how to build with gpt4 what it’s like to create with it as a partner

And so the thing we’re going to do is we’re going to actually build a Discord bot I’ll build it live and show you the process show you debugging show you what the model can do where its limitations are and how to work with with them in

Order to sort of achieve New Heights so the first thing I’ll do is tell the model that this time it’s supposed to be an AI programming assistant its job is to write things out in pseudocode first and then actually write the code and this approach is very

Helpful so that the model break down the problem into smaller pieces and then that way you’re not kind of asking it to just come up with a super hard solution to a problem all in one go it also makes it very interpretable because you can see exactly what the

Model was thinking and you can even provide Corrections if you’d like so here is the prompt that we’re going to ask it this is the kind of thing that 3.5 would totally choke on if you’ve tried anything like it but so we’re going to ask for a Discord bot that uses

The gpd4 API to read images and texts now there’s one problem here which is this model’s training cutoff is in 2021 which means it has not seen our new chat completions format so I literally just went to the blog post from two weeks ago copy pasted from the blog post including

The response format it has not seen the new image extension to that and so I just kind of wrote that up and you know just very minimal detail about how to include images so and now the model can actually leverage the doc that documentation that it did not have memorized that it does

Not know okay and in general these models are very good at using information that it’s been trained on in new ways and synthesizing new content and you can see that right here that it actually wrote an entirely new bot now let’s actually see if this bot is going to

Work in practice so you should always look through the code to get a sense of what it does don’t run untrusted code from humans or from AIS and one thing to note is that the Discord API has changed a lot over time and particularly that there’s one

Feature that has changed a lot since this model was trained give it a try in fact yes we are missing the intense keyword this is something that came out in 2020 . so the model does know it exists but it doesn’t know which version of the

Discord API we’re using so are we out of luck well not quite we can just simply paste to the model exactly the error message not even going to say hey this is from running your code could you please fix it we’ll just let it run and the model says oh yeah whoops the

Intense argument here’s the correct here’s the correct code now let’s give this a try once again kind of making sure that we understand what the code is doing now a second issue that can come up is it doesn’t know what environment I’m running in and if you notice it says hey

Here’s this inscrutable error message which if you’ve not used jupyter notebook a lot with async IO before you probably have no idea what this means but fortunately once again you can just sort of say to the model hey I am using Jupiter and would like to make this work can you fix it

And the specific problem is that there’s already an event Loop running so you need to use this Nest async i o Library you need to call Net Nest I sync IO dot apply the model knows all of this correctly instantiates all of these these pieces into the bot it even helps

Hopefully tells you oh you’re running in Jupiter well you can do this bang pip install in order to install the package if you don’t already have it that was very helpful so now we’ll run and it looks like something happened so the first thing I’ll do is go over to our Discord

And I will paste in a screenshot of our Discord itself so remember gpt4 is not just a language model it’s also a vision model in fact it can flexibly accept inputs that intersperse images and text arbitrarily kind of like a document now the image feature is in

Preview so this is going to be a little sneak peek it’s not yet publicly available it’s something we’re working with one partner called be my eyes in order to really start to develop it and get it ready for prime time but you can ask anything you like for

Example I can’t you know I’ll say gp4 hello world can you describe this image and painstaking detail all right which first of all think of how you would do this yourself there’s a lot of different things you could latch onto a lot of different pieces of the

System you could describe and we can go over to the actual code and we can see that yep we in fact received the message have formatted an appropriate request for our API and now we wait um because you know one of the things we

Have to do is we have to make the system faster that’s one of the things that we’re working on optimizing in the meanwhile I just want to say to the audience that’s watching we’ll take an audience request next so if you have an image and a task you’d like to

Accomplish please submit that to the Discord our moderators will pick one that will run so we can see that the Discord oh it looks like we have a response perfect so it’s a screenshot of a Discord application interface pretty good did not even describe it it knows that it’s

Discord it’s probably Discord written there somewhere where it just kind of knows this from from prior experience server icon label gpd4 describes the interface in great detail talks about uh all the people telling me that I’m supposed to do Q uh very very kind audience and describes a much of the uh the

Notification messages and the users that are in the channel and so there you go that’s some that’s some pretty good understanding now this next one if you notice first of all we got a post but the model did not actually see the message so is this a failure of the

Model or of the system around the model well we can take a look and if you notice here content is an empty string we received a blank message contents the reason for this is a dirty trick that we played on the AI so if you go to the Discord documentation

And you scroll through it all the way down to uh I can see it hard for me to even find honestly to the message content intent you’ll see this was added as of September 2022 as a required field so in order to receive a message that does not

Explicitly tag you you now have to include this new intent in your code remember I said intensive change a lot over time this is much newer than the model as possible is possibly able to know so maybe we’re out of luck we have to debug this by hand but once again we

Can try to use gpd4’s language understanding capabilities to solve this now keep in mind this is a document of like I think this is like ten thousand fifteen thousand words something like that it’s not formatted very well this is literally a command a copy paste like this is what it’s

Supposed to parse through to find in the middle of that document that oh yeah message contents that’s required now but let’s see if it can do it so we will ask for I I am receiving blank message contents can you why could this be happening how do I fix it

So one thing that’s new about gpd4 is context length 32 000 tokens is kind of the upper limit that we support right now and the model is able to flexibly use long documents it’s something we’re still optimizing so we recommend trying it out but not necessarily sort of really really

Scaling it up just yet unless you have an application that really benefits from it so if you’re really interested in Long context please let us know we want to see what kinds of applications it unlocks but if you see it says oh yeah message content intent

Was not enabled and so you can either ask the model to write some code for you or you could I actually just you know do it the old-fashioned way either way is fine I think this is a augmenting tool makes you much more productive but it’s still

Important that you are in the driver’s seat and are the manager and knows what’s what’s going on so now we’re connected once again and uh Boris would you like to rerun the message once again we can see that we have received it even though the bot was not explicitly tagged

Seems like a pretty good pretty good description interesting this is an interesting image actually looks like it’s a dolly generated one and let’s actually try this one as well so what’s funny about this image oh it’s already been submitted so once again we can verify this making the right API calls

Squirrels do typically eat nuts we don’t expect them to use a camera or act like a human so I think that’s that’s a pretty good explanation of why that image is funny so I’m going to show you one more example of what you can do with this model

So I have here a nice hand-drawn mock-up of a joke website definitely worthy of being put up on my refrigerator so I’m just going to take out my phone literally take a photo of this mock-up and then I’m going to send it to our Discord all right going to send it to our

Discord and this is of course the rockiest part making sure that we actually send it to the right Channel which in fact I think maybe I did not sent it to the wrong Channel it’s funny it’s always the uh the sort of non-ai parts of these demos that are

The hardest part to do and here we go technology is now solved and now we wait so the thing that’s amazing in my mind is that what’s going on here is we’re talking to a neural network and this neural network was trained to predict what comes next right it played

This like this game of sort of being shown a partial document and then predicted what comes next across an unimaginably large amount of content and from there it learns all of these skills that you can apply and all these very flexible ways and so we can actually

Take now this output so literally we just said to output the HTML from that picture and here we go actual working JavaScript filled in the jokes for comparison this was the original of our mock-up and so there you go going from hand-drawn beautiful art if I do say so myself to working website

And this is all just potential right we you can see lots of different applications we ourselves are still figuring out new ways to use this so we’re going to work with our partner we’re going to scale up from there but please be patient because it’s going to

Take us some time to really make this available for everyone so I have one last thing to show you I’ve shown you reading existing content I’ve shown you how to build with the system as a partner the last thing I’m going to show is how to work with the system to

Accomplish a task that none of us like to do but we all have to so you may have guessed the thing we’re going to do is taxes now note that GPT is not a certified tax professional nor am I so you should always check with your your Tax Advisor

But it can be helpful to understand some dense content to just be able to empower yourself to to be able to sort of solve problems and get a get a handle on what’s Happening when you could not otherwise so once again I’ll do a system message in this case I’m going to tell

It that it’s tax GPT which is not a specific thing that we’ve trained into this model you can be very creative if you want with the system message to really get the model in the mood of what is your job what are you supposed to do so I pasted in

The tax code this is about 16 Pages worth of of tax code and there’s this question about Allison Bob they got married at one point uh and that here are their their incomes and they take a standard deduction they’re filing jointly so first question what is their standard deduction for 2018

. so while the model is chugging I’m going to solve this problem by hand to show you what’s involved so the standard deduction is the basic standard deduction plus the additional the basic one is 200 percent for a joint return of subparagraph C which is here okay so additional doesn’t apply the limitation

Doesn’t apply um okay now these apply oh wait special rules for taxable year 2018 which is the one we care about through 2025 you have to substitute twelve thousand for three thousand so two hundred percent of twelve thousand twenty four thousand is the final answer

If you notice the model got to the same conclusion and you can actually read through its explanation and to tell you the truth the first time I tried to approach this problem myself I could not figure it out I spent half an hour reading through the tax code

Trying to figure out this like back reference and why there’s some program like just what’s even going on it was only by asking the model to spell out its reasoning and then I followed along that I was like oh I get it now I understand how this works and so that I

Think is where the power of the system lies it’s not perfect but neither are you and together is this amplifying tool that lets you just reach New Heights and you can go further you can say okay now calculate their total liability and here we go it’s doing the calculation

Honestly I every time it does it it’s just it’s amazing this model is so good at Mental Math it’s way way better than I am at Mental Math it’s not hooked up to a calculator like that’s another way that you could really try to enhance these systems but it has these raw

Capabilities that are so flexible it doesn’t care if it’s code it doesn’t care if it’s language it doesn’t care if it’s tax all of these capabilities in one system that can be applied towards the problem that you care about towards your application towards whatever you build

And so to end it the final thing that I will show is I a little other dose of creativity which is now summarize this problem into a rhyming poem and there we go a beautiful beautiful poem about doing your taxes so thank you everyone for tuning in I hope you

Learned something about what the model can do how to work with it and honestly we’re just really excited to see what you’re going to build I I’ve talked about openai evals please contribute we think that this model improving it bring it to the next level is something that

Everyone can contribute to and that we think it can really benefit a lot of people and we want your help to do that so thank you very much we’re so excited to see what you’re going to build foreign

,00:01 foreign
00:08 did the gpd4 developer demo live stream
00:12 honestly it’s kind of hard for me to
00:13 believe that this day is here open AI
00:15 has been building this technology really
00:18 since we started the company but for the
00:20 past two years we’ve been really focused
00:21 on delivering gpt4
00:23 that started with rebuilding our entire
00:26 training stack actually training the
00:28 model
00:29 and then seeing what it was capable of
00:31 trying to figure out its capabilities
00:32 its risks working with Partners in order
00:34 to test it in real world scenarios
00:36 really tuning Its Behavior optimizing
00:39 the model getting it available
00:41 so that you can use it and so today our
00:44 goal is to show you a little bit of how
00:46 to make gbto4 shine
00:48 how to really get the most out of it you
00:50 know where it’s kind of you know
00:51 weaknesses are where we’re still working
00:53 on it and just how to really use it as a
00:55 good tool a good partner
00:57 um so if you’re interested in
00:58 participating in the Stream uh that if
01:01 you go to our Discord so that’s
01:02 discord.gg openai there’s comments in
01:04 there and we’ll take a couple of
01:05 audience suggestions
01:08 so the first thing I want to show you is
01:10 the first task that gpd4 could do that
01:13 we never really got 3.5 to do
01:16 and the way to think about this is all
01:17 throughout training that you know you’re
01:19 constantly doing all this work it’s 2
01:21 A.M the pager goes off you fix the model
01:23 and you’re always wondering is it gonna
01:25 work
01:27 is all this effort actually going to pan
01:28 out and so we all had a pet task that we
01:31 really liked and that we would all
01:33 individually be trying to see is the
01:35 model capable of it now
01:36 and I’m going to show you the first one
01:39 that we had a success for four but never
01:41 really got there for 3.5
01:43 so I’m just going to copy the top of our
01:45 blog post from today going to paste it
01:47 into our Playground now this is our new
01:51 chat completions playground that came
01:53 out two weeks ago I’m going to show you
01:54 first with GPT 3.5 4 has the same API to
01:58 it the same playground
01:59 the way that it works is you have a
02:01 system message where you explain to the
02:02 model what it’s supposed to do and we’ve
02:05 made these models very steerable so you
02:07 can provide it with really any
02:08 instruction you want whatever you dream
02:10 up and the model will adhere to it
02:12 pretty well and in the future it will
02:13 get increasingly increasingly powerful
02:15 at steering the model very reliably
02:20 you can then paste whatever you want as
02:22 a user the model will return messages as
02:24 an assistant and the way to think of it
02:26 is that we’re moving away from sort of
02:28 just raw text in raw text out where you
02:30 can’t tell where different parts of the
02:31 conversation come from but towards this
02:33 much more structured format that gives
02:34 the model the opportunity to know well
02:36 this is the user asking me to do
02:38 something that the developer didn’t
02:39 attend I should listen to the developer
02:41 here
02:42 all right so now time to actually show
02:44 you the task that I’m referring to so
02:46 everyone’s familiar with summarize
02:49 this let’s say article into a sentence
02:52 okay getting a little more specific uh
02:54 but where every word begins with G
02:58 so this is 3.5 let’s see what it does
03:02 yeah it kind of didn’t even try
03:04 just gave up on the task this is pretty
03:06 typical for 3.5 trying to do this
03:09 particular kind of task if it’s you know
03:11 sort of a very kind of stilted article
03:13 or something like that maybe it can
03:15 succeed but for the most part 3.5 just
03:17 gives up
03:19 but let’s try the exact same prompt
03:22 the exact same system message
03:25 in gbt4
03:29 so kind of borderline whether you want
03:31 to count AI or not but so let’s say AI
03:34 doesn’t count
03:36 that’s cheating
03:41 so fair enough the model happily accepts
03:43 my feedback
03:44 so now to make sure it’s not just good
03:46 for G’s I’d like to turn this over to
03:48 the audience I’ll take a suggestion on
03:49 what letter to try next in the meanwhile
03:52 while I’m waiting for our moderators to
03:53 pick the lucky lucky letter I will give
03:56 a try with a
04:02 um but in this case I’ll say gpd4 is
04:03 fine
04:04 why not
04:07 also pretty good summary
04:09 so I’ll hop over to our Discord
04:12 all right
04:13 wow if people are are being a little
04:15 ambitious here I’m really trying to put
04:17 the model through the paces we’re going
04:18 to try Q uh which if you think about
04:20 this for a moment I want the audience to
04:22 really think about how would you do a
04:24 summary of this article that all starts
04:26 with Q it’s not easy
04:36 it’s pretty good that’s pretty good
04:40 all right so I’ve shown you summarizing
04:43 an existing article I want to show you
04:45 how you can flexibly combine ideas
04:47 between different articles so I’m going
04:50 to take this article that was on Hacker
04:52 News yesterday
04:54 copy paste it
04:56 into the same conversation so it has all
04:58 the context of what we’re just doing I’m
05:00 going to say find one common theme
05:03 between this article and the gpd4 blog
05:10 so this is an article about Pinecone
05:13 which is a python web app development
05:15 framework and it’s making the technology
05:16 more accessible user friendly if you
05:18 don’t think that was insightful enough
05:19 you can always give some feedback and
05:21 say that was not insightful
05:25 enough
05:26 please no I’ll just even just leave it
05:28 there leave it up to the model to decide
05:30 so Bridging the Gap between powerful
05:31 technology and practical applications
05:34 seems not bad and of course you can ask
05:36 for any other kind of task you want
05:38 using its flexible language
05:39 understanding and synthesis you can ask
05:42 for something like
05:43 now turn the GT4 blog post into a
05:48 rhyming poem
05:57 picked up on open AI evalues open source
06:00 for all helping to guide answering the
06:02 call which by the way if you’d like to
06:04 contribute to this model please give us
06:06 evals we have an open source evaluation
06:07 framework that will help us guide and
06:09 all of our users understand what the
06:11 model is capable of and to take it to
06:13 the next level
06:15 so there we go this is consuming
06:17 existing content using gpt4 with with a
06:20 little bit of creativity on top
06:22 but next I want to show you how to build
06:25 with gpt4 what it’s like to create with
06:28 it as a partner
06:30 and so the thing we’re going to do
06:32 is we’re going to actually build a
06:34 Discord bot
06:36 I’ll build it live and show you the
06:38 process show you debugging show you what
06:40 the model can do where its limitations
06:42 are and how to work with with them in
06:44 order to sort of achieve New Heights so
06:46 the first thing I’ll do is tell the
06:47 model that this time it’s supposed to be
06:49 an AI programming assistant
06:51 its job is to write things out in
06:53 pseudocode first and then actually write
06:55 the code and this approach is very
06:58 helpful so that the model break down the
07:00 problem into smaller pieces and then
07:02 that way you’re not kind of asking it to
07:04 just come up with a super hard solution
07:06 to a problem all in one go
07:08 it also makes it very interpretable
07:10 because you can see exactly what the
07:12 model was thinking and you can even
07:13 provide Corrections if you’d like
07:15 so here is the prompt that we’re going
07:18 to ask it this is the kind of thing that
07:20 3.5 would totally choke on if you’ve
07:22 tried anything like it but so we’re
07:24 going to ask for a Discord bot that uses
07:26 the gpd4 API to read images and texts
07:31 now there’s one problem here which is
07:33 this model’s training cutoff is in 2021
07:37 which means it has not seen our new chat
07:39 completions format so I literally just
07:41 went to the blog post from two weeks ago
07:43 copy pasted from the blog post including
07:45 the response format it has not seen the
07:48 new image extension to that and so I
07:50 just kind of wrote that up and you know
07:51 just
07:52 very minimal detail about how to include
07:54 images so and now the model can actually
07:57 leverage the doc that documentation that
07:59 it did not have memorized that it does
08:01 not know
08:04 okay
08:09 and in general these models are very
08:11 good at using information that it’s been
08:13 trained on in new ways and synthesizing
08:15 new content and you can see that right
08:17 here that it actually wrote an entirely
08:19 new bot
08:21 now let’s
08:23 actually see if this bot is going to
08:24 work in practice so you should always
08:26 look through the code to get a sense of
08:28 what it does don’t run untrusted code
08:30 from humans or from AIS
08:33 and one thing to note is that the
08:35 Discord API has changed a lot over time
08:38 and particularly that there’s one
08:40 feature that has changed a lot since
08:42 this model was trained
08:44 give it a try in fact yes we are missing
08:47 the intense keyword this is something
08:49 that came out in 2020
08:52 . so the model does know it exists but
08:53 it doesn’t know which version of the
08:55 Discord API we’re using so are we out of
08:58 luck well not quite we can just simply
09:00 paste to the model exactly the error
09:03 message not even going to say hey this
09:05 is from running your code could you
09:06 please fix it
09:08 we’ll just let it run
09:10 and the model says oh yeah whoops the
09:12 intense argument here’s the correct
09:13 here’s the correct code
09:17 now let’s give this a try once again
09:19 kind of making sure that we understand
09:21 what the code is doing
09:27 now a second issue that can come up is
09:29 it doesn’t know what environment I’m
09:30 running in and if you notice it says hey
09:33 here’s this inscrutable error message
09:35 which if you’ve not used jupyter
09:37 notebook a lot with async IO before you
09:39 probably have no idea what this means
09:42 but fortunately
09:45 once again you can just sort of say to
09:47 the model hey
09:48 I am using Jupiter
09:52 and would like to make this work
09:56 can you fix it
09:58 and the specific problem is that there’s
10:00 already an event Loop running so you
10:01 need to use this Nest async i o Library
10:04 you need to call Net Nest I sync IO dot
10:06 apply the model knows all of this
10:08 correctly instantiates all of these
10:11 these pieces into the bot it even helps
10:14 hopefully tells you oh you’re running in
10:15 Jupiter well you can do this bang pip
10:17 install in order to install the package
10:19 if you don’t already have it that was
10:21 very helpful
10:23 so now we’ll run and it looks like
10:25 something happened
10:26 so the first thing I’ll do
10:28 is
10:30 go over to our Discord
10:32 and I will paste in
10:34 a screenshot
10:36 of our Discord itself so remember gpt4
10:39 is not just a language model it’s also a
10:43 vision model in fact it can flexibly
10:45 accept inputs that intersperse images
10:48 and text arbitrarily kind of like a
10:50 document now the image feature is in
10:54 preview so this is going to be a little
10:55 sneak peek it’s not yet publicly
10:57 available it’s something we’re working
10:58 with one partner called be my eyes in
11:01 order to really start to develop it and
11:02 get it ready for prime time
11:04 but you can ask anything you like for
11:05 example I can’t you know I’ll say gp4
11:11 hello world
11:13 can you describe this image
11:17 and painstaking detail
11:20 all right which first of all think of
11:22 how you would do this yourself there’s a
11:25 lot of different things you could latch
11:25 onto a lot of different pieces of the
11:27 system you could describe and we can go
11:29 over to the actual code and we can see
11:30 that yep we in fact received the message
11:32 have formatted an appropriate request
11:35 for our API
11:37 and now we wait
11:39 um because you know one of the things we
11:40 have to do is we have to make the system
11:41 faster that’s one of the things that
11:43 we’re working on optimizing in the
11:45 meanwhile I just want to say to the
11:46 audience that’s watching we’ll take an
11:48 audience request next so if you have an
11:50 image and a task you’d like to
11:52 accomplish please submit that to the
11:54 Discord our moderators will pick one
11:55 that will run
12:01 so we can see that the Discord oh it
12:03 looks like we have a response perfect
12:06 so it’s a screenshot of a Discord
12:07 application interface pretty good did
12:09 not even describe it it knows that it’s
12:11 Discord it’s probably Discord written
12:13 there somewhere where it just kind of
12:14 knows this from from prior experience
12:16 server icon label gpd4 describes the
12:19 interface in great detail talks about uh
12:22 all the people telling me that I’m
12:23 supposed to do Q uh very very kind
12:25 audience
12:26 and describes a much of the uh the
12:29 notification messages and the users that
12:31 are in the channel and so there you go
12:33 that’s some that’s some pretty good
12:34 understanding now this next one if you
12:37 notice first of all we got a post but
12:41 the model did not actually see the
12:43 message so is this a failure of the
12:45 model or of the system around the model
12:47 well we can take a look
12:49 and if you notice here content is an
12:51 empty string we received a blank message
12:53 contents
12:55 the reason for this is a dirty trick
12:57 that we played on the AI
12:59 so if you go to the Discord
13:01 documentation
13:03 and you scroll through it all the way
13:06 down to uh I can see it hard for me to
13:10 even find honestly to the message
13:11 content
13:13 intent you’ll see this was added as of
13:15 September 2022 as a required field so in
13:19 order to receive a message that does not
13:20 explicitly tag you you now have to
13:22 include this new intent in your code
13:25 remember I said intensive change a lot
13:27 over time this is much newer than the
13:29 model as possible is possibly able to
13:31 know so maybe we’re out of luck we have
13:34 to debug this by hand but once again we
13:36 can try to use gpd4’s language
13:38 understanding capabilities
13:40 to solve this now keep in mind this is a
13:43 document of like I think this is like
13:44 ten thousand fifteen thousand words
13:46 something like that it’s not formatted
13:48 very well this is literally a command a
13:50 copy paste like this is what it’s
13:52 supposed to parse through to find in the
13:54 middle of that document that oh yeah
13:55 message contents that’s required now but
13:58 let’s see if it can do it
13:59 so we will ask for I I am receiving
14:03 blank message contents
14:06 can you
14:08 why could this be happening
14:12 how do I fix it
14:16 so one thing that’s new about gpd4 is
14:18 context length
14:20 32 000 tokens is kind of the upper limit
14:22 that we support right now and the model
14:24 is able to flexibly use long documents
14:28 it’s something we’re still optimizing so
14:31 we recommend trying it out but not
14:34 necessarily sort of really really
14:35 scaling it up just yet unless you have
14:38 an application that really benefits from
14:39 it so if you’re really interested in
14:41 Long context please let us know we want
14:43 to see what kinds of applications it
14:45 unlocks but if you see
14:47 it says oh yeah message content intent
14:49 was not enabled and so you can either
14:50 ask the model to write some code for you
14:52 or you could
14:54 I actually just you know do it the
14:57 old-fashioned way
14:58 either way is fine
15:01 I think this is a augmenting tool makes
15:03 you much more productive but it’s still
15:05 important that you are in the driver’s
15:07 seat and are the manager and knows
15:09 what’s what’s going on so now we’re
15:10 connected once again
15:12 and uh Boris would you like to rerun the
15:14 message
15:26 once again we can see that we have
15:28 received it even though the bot was not
15:30 explicitly tagged
15:32 seems like a pretty good
15:36 pretty good description interesting this
15:38 is an interesting image actually looks
15:39 like it’s a dolly generated one and
15:42 let’s actually try this one as well
15:49 so what’s funny about this image oh it’s
15:52 already been submitted
15:56 so once again we can verify this making
15:58 the right API calls
16:02 squirrels do typically eat nuts we don’t
16:04 expect them to use a camera or act like
16:06 a human so I think that’s that’s a
16:08 pretty good explanation of why that
16:10 image is funny
16:11 so I’m going to show you one more
16:13 example of what you can do with this
16:16 model
16:17 so I have here a nice hand-drawn mock-up
16:21 of a joke website definitely worthy of
16:23 being put up on my refrigerator
16:26 so I’m just going to take out my phone
16:29 literally take a photo
16:32 of this mock-up
16:35 and then I’m going to send it
16:38 to our Discord
16:49 all right going to send it to our
16:51 Discord
17:00 and this is of course the rockiest part
17:02 making sure that we actually send it to
17:04 the right Channel
17:08 which in fact I think maybe I did not
17:14 sent it to the wrong Channel
17:16 it’s funny it’s always the uh the sort
17:19 of non-ai parts of these demos that are
17:21 the hardest part to do
17:28 and here we go
17:31 technology is now solved
17:35 and now we wait
17:39 so the thing that’s amazing in my mind
17:41 is that
17:43 what’s going on here is we’re talking to
17:45 a neural network
17:46 and this neural network was trained to
17:48 predict what comes next right it played
17:51 this like this game of sort of being
17:52 shown a partial document and then
17:54 predicted what comes next across an
17:55 unimaginably large amount of content and
17:58 from there it learns all of these skills
18:00 that you can apply and all these very
18:02 flexible ways and so we can actually
18:04 take now this output so literally we
18:07 just said to
18:09 output the HTML from that picture
18:13 and here we go
18:15 actual working JavaScript
18:18 filled in the jokes
18:19 for comparison
18:22 this was the original
18:27 of our mock-up
18:29 and so there you go going from
18:30 hand-drawn
18:32 beautiful art
18:33 if I do say so myself to working website
18:38 and this is all just potential right we
18:39 you can see lots of different
18:41 applications we ourselves are still
18:43 figuring out new ways to use this so
18:45 we’re going to work with our partner
18:46 we’re going to scale up from there but
18:48 please be patient because it’s going to
18:49 take us some time to really make this
18:51 available for everyone
18:54 so I have one last thing to show you
18:56 I’ve shown you reading existing content
18:59 I’ve shown you how to
19:01 build with the system as a partner the
19:04 last thing I’m going to show
19:06 is how to work with the system to
19:08 accomplish a task that none of us like
19:10 to do but we all have to
19:12 so you may have guessed the thing we’re
19:14 going to do is taxes
19:17 now note that GPT is not a certified tax
19:20 professional nor am I so you should
19:21 always check with your your Tax Advisor
19:23 but it can be helpful to understand some
19:26 dense content to just be able to empower
19:28 yourself to to be able to sort of solve
19:30 problems and get a get a handle on
19:32 what’s Happening when you could not
19:33 otherwise so once again I’ll do a system
19:35 message in this case I’m going to tell
19:37 it that it’s tax GPT which is not a
19:40 specific thing that we’ve trained into
19:41 this model you can be very creative if
19:43 you want with the system message to
19:45 really get the model in the mood of what
19:46 is your job what are you supposed to do
19:49 so I pasted in
19:51 the tax code this is about 16 Pages
19:53 worth of of tax code and there’s this
19:55 question about Allison Bob they got
19:57 married at one point uh and that here
19:59 are their their incomes and they take a
20:00 standard deduction they’re filing
20:02 jointly so first question what is their
20:04 standard deduction for 2018
20:07 . so while the model is chugging I’m
20:09 going to solve this problem by hand to
20:11 show you what’s involved so the standard
20:13 deduction is the basic standard
20:15 deduction plus the additional the basic
20:17 one is 200 percent for a joint return of
20:20 subparagraph C which is here okay so
20:23 additional doesn’t apply the limitation
20:25 doesn’t apply
20:26 um okay now these apply oh wait special
20:29 rules for taxable year 2018 which is the
20:32 one we care about through 2025 you have
20:34 to substitute twelve thousand for three
20:36 thousand so two hundred percent of
20:38 twelve thousand twenty four thousand is
20:39 the final answer
20:41 if you notice the model got to the same
20:44 conclusion
20:45 and you can actually read through its
20:49 explanation
20:50 and to tell you the truth the first time
20:52 I tried to approach this problem myself
20:54 I could not figure it out I spent half
20:57 an hour reading through the tax code
20:58 trying to figure out this like back
21:00 reference and why there’s some program
21:01 like just what’s even going on it was
21:04 only by asking the model to spell out
21:06 its reasoning and then I followed along
21:08 that I was like oh I get it now I
21:10 understand how this works and so that I
21:13 think is where the power of the system
21:14 lies it’s not perfect but neither are
21:17 you and together is this amplifying tool
21:19 that lets you just reach New Heights
21:22 and you can go further you can say okay
21:25 now calculate their total liability
21:33 and here we go it’s doing the
21:35 calculation
21:51 honestly I every time it does it it’s
21:53 just it’s amazing this model is so good
21:56 at Mental Math it’s way way better than
21:58 I am at Mental Math it’s not hooked up
21:59 to a calculator like that’s another way
22:01 that you could really try to enhance
22:02 these systems but it has these raw
22:05 capabilities that are so flexible it
22:07 doesn’t care if it’s code it doesn’t
22:08 care if it’s language it doesn’t care if
22:10 it’s tax all of these capabilities in
22:12 one system that can be applied
22:14 towards the problem that you care about
22:16 towards your application towards
22:17 whatever you build
22:19 and so to end it the final thing that I
22:21 will show is I a little other dose of
22:24 creativity which is now summarize this
22:27 problem into a rhyming poem
22:36 and there we go a beautiful beautiful
22:38 poem about doing your taxes so thank you
22:42 everyone for tuning in I hope you
22:44 learned something about what the model
22:46 can do how to work with it and honestly
22:48 we’re just really excited to see what
22:51 you’re going to build I I’ve talked
22:52 about openai evals please contribute we
22:54 think that this model improving it bring
22:56 it to the next level is something that
22:58 everyone can contribute to and that we
23:00 think it can really benefit a lot of
23:01 people and we want your help to do that
23:03 so thank you very much we’re so excited
23:05 to see what you’re going to build
23:09 foreign
, , , #GPT4 #Developer #Livestream , [agora]

Leave a Reply

About Me

Lorem ipsum dolor sit amet, consectetur adipiscing elit. Ut elit tellus, luctus nec ullamcorper mattis, pulvinar dapibus leo.

Recent Posts

Need to raise your site's score?

We have an ideal solution for your business marketing
Nullam eget felis

Do you want a more direct contact with our team?

Sed blandit libero volutpat sed cras ornare arcu dui. At erat pellentesque adipiscing commodo elit at.