OpenAI发布会快速解读+全程双语回放|地表最强模型GPT-4o免费发布|AI进化论-花生

AI进化论-花生
13 May 202431:38

Summary

TLDR在这段视频中,我们见证了一场展示先进人工智能技术的活动。首先,一位名叫Mark的演讲者通过与AI助手Chat GPT的互动,展示了实时对话和情感识别的能力。接着,Barrett向观众展示了如何通过语音模式讲述一个关于机器人和爱情的睡前故事,以及如何使用不同的情感风格来表达故事。此外,还展示了AI在解决线性方程和编程问题方面的应用,包括实时翻译和面部表情分析。最后,演讲者强调了将这些技术以安全和有用方式带给公众的重要性,并预告了未来几周将逐步推出这些功能,同时感谢了Open AI团队和Nvidia团队对演示的支持。

Takeaways

  • 🌐 该演示展示了如何使用GPT 40模型,它具备实时对话语音、视觉和文本能力,能够提供更自然和流畅的交互体验。
  • 📈 GPT 40在语音模式下允许用户打断对话,响应时间更快,并且能够识别和表达情感。
  • 📊 通过代码演示,GPT 40能够理解和分析线性方程,并提供解决问题的提示,而不是直接给出答案。
  • 🔍 GPT 40的视觉能力可以分析和解释图像内容,如天气数据图表,以及实时翻译文本。
  • 📅 2018年的天气数据分析显示,最热的月份是7月和8月,平均温度在25°至30°之间。
  • 🎓 GPT 40的推出意味着即使是免费用户也能享受到高级AI工具,这有助于降低技术使用的门槛。
  • 🚀 GPT 40的发布是AI技术在易用性方面的一大步进,预示着未来人机交互的新方向。
  • 🌟 GPT 40的实时性和情感识别能力,使得它在教育、娱乐和日常生活等多个领域都有潜在的应用价值。
  • 📱 GPT 40的移动应用演示表明,该技术可以轻松集成到用户的日常工作流程中,提供便捷的服务。
  • 🌍 GPT 40支持50多种语言,这表明其旨在为全球用户提供服务,无论他们使用何种语言。
  • 🔧 GPT 40的迭代部署和安全性能的持续优化,展示了开发团队在确保技术负责任使用方面的努力。

Q & A

  • 在演示中,GPT 40的主要特点是什么?

    -GPT 40的主要特点包括实时会话语音功能,能够以不同的情感风格生成语音,以及在语音、文本和视觉方面的实时响应能力。此外,GPT 40还提供了更高效的性能,允许免费用户也能享受GPT 4级的智能服务。

  • GPT 40如何改善用户体验?

    -GPT 40通过实时响应和情感感知能力,提供了更自然和流畅的交互体验。它还能够原生处理语音、文本和视觉信息,减少了延迟,提高了沉浸感。

  • GPT 40在安全性方面面临哪些挑战?

    -GPT 40在安全性方面的挑战包括实时音频和视觉处理的能力,这要求团队构建有效的对策来防止滥用。此外,还需要与政府、媒体、娱乐等不同利益相关者合作,确保技术的安全部署。

  • GPT 40的发布对免费用户意味着什么?

    -对于免费用户来说,GPT 40的发布意味着他们可以享受到以前只有付费用户才能使用的高级工具,包括使用GPT进行对话、访问GPT商店、使用视觉功能以及高级数据分析等。

  • GPT 40在多语言支持方面有哪些改进?

    -GPT 40在50多种语言上提高了质量和速度,这使得产品能够服务于更广泛的用户群体,无论他们使用哪种语言。

  • 在演示中,GPT 40如何帮助解决线性方程?

    -GPT 40通过提供逐步的提示和反馈,帮助用户解决线性方程,而不是直接给出答案。它引导用户理解并应用解决方程的每个步骤,从而增强了用户的理解和信心。

  • GPT 40的视觉功能如何帮助用户理解代码和图形?

    -GPT 40的视觉功能允许它查看用户屏幕上的内容,如代码和图形输出。它能够分析这些视觉信息,并以简洁的方式向用户提供解释和概述。

  • GPT 40的实时语音功能如何帮助减轻演讲者的紧张情绪?

    -GPT 40通过提供实时反馈和建议,如指导演讲者进行深呼吸,帮助他们放松。它还能够感知演讲者的情绪状态,并给出相应的建议。

  • GPT 40在情感识别方面有哪些能力?

    -GPT 40能够通过语音和视觉输入识别和响应用户的情感状态。它可以根据用户的声音强度、语调和其他线索来感知情绪,并据此调整其响应。

  • GPT 40的API对开发者意味着什么?

    -GPT 40的API意味着开发者现在可以使用这一先进的模型来构建和部署他们自己的AI应用程序。这为创造新的、创新的解决方案提供了可能性,并能够大规模地部署这些应用。

  • GPT 40在性能上相比GPT 4有哪些提升?

    -GPT 40在性能上相比GPT 4有显著提升,它运行速度更快(2倍速),成本更低(50% cheaper),并且提供了更高的速率限制(五倍于GPT 4 Turbo)。

  • GPT 40如何帮助用户在不同领域中应用数学?

    -GPT 40通过提供实际的数学问题解决示例,帮助用户理解数学在日常生活中的应用,如计算开支、规划旅行、烹饪,甚至是商业利润和损失计算等。

Outlines

00:00

😀 语言翻译与数据分析

第一段主要描述了一个场景,其中两个人,一个说英语,一个说意大利语,通过一个翻译器进行交流。他们讨论了关于鲸鱼是否能交流的话题,并解决了线性方程的问题。此外,还提到了一个显示2018年全年平均温度、最低温度和最高温度的图表,以及一个在九月份标记的大降雨事件。

05:01

📈 产品发布与技术介绍

第二段介绍了一个产品的发布会,演讲者表达了对产品的热情,并介绍了三个主要议程。重点在于发布桌面版本的CH gbt和用户界面的更新,使得产品更易于使用。最引人注目的是推出了新的旗舰模型GBT 4,它为所有用户包括免费用户提供了高级智能。演讲者还提到了实时音频、视觉和多种语言支持的改进,以及如何确保新技术的安全使用。

10:02

🎉 新功能展示与用户互动

第三段继续展示了GBT 4的新功能,包括实时对话语音模式,以及如何通过减少延迟来提高用户体验。演讲者还提到了将高级工具免费提供给所有用户,从而扩大了定制聊天GBT的受众范围。此外,还介绍了视觉功能,允许用户上传包含文本和图像的文档,并与聊天GBT进行交流。

15:04

🤖 机器人故事与情感交互

第四段通过一个实时演示,展示了如何使用GBT 4进行实时对话和情感交互。Mark在舞台上进行了一个实时演示,GBT 4帮助他平静下来,并以不同的情感风格讲述了一个关于机器人和爱的故事。此外,还展示了如何使用GBT 4的语音模式来提高用户体验。

20:08

🧮 解决数学问题与编码交互

第五段展示了如何使用GBT 4解决数学问题和与编码交互。演讲者写下了一个线性方程,并与GBT 4一起解决了它,GBT 4提供了解决步骤的提示。接着,演讲者展示了一段代码,GBT 4解释了代码的功能,包括获取天气数据、平滑温度数据、注释重要天气事件以及显示带有平均温度、最低温度和最高温度的图表。

25:08

🌡️ 温度数据分析与实时翻译

第六段继续展示了GBT 4的数据分析能力,特别是对温度数据的处理。演讲者询问了关于温度最高的月份和相应温度的问题,GBT 4提供了答案。此外,还展示了GBT 4的实时翻译功能,能够将英语和意大利语互译。

30:09

😊 情感检测与技术展望

第七段展示了GBT 4的情感检测能力,通过分析一张自拍照来识别情感。此外,演讲者感谢了团队和合作伙伴,并预告了未来几周将推出的新功能和技术。最后,演讲者对观众的参与表示感谢。

Mindmap

Keywords

💡实时对话语音

实时对话语音是指能够即时响应用户语音输入并提供语音输出的技术。在视频中,这项技术使得用户可以与GPT模型进行自然的对话,无需等待,就像与人交谈一样。例如,Mark在舞台上进行实时演示时,可以直接与GPT进行对话,GPT能够即时回应,帮助他平静紧张的情绪。

💡情感识别

情感识别是指通过分析语音、面部表情或其他信号来识别和理解人的情感状态的技术。视频中,GPT模型展示了它能够通过语音判断Mark的紧张情绪,并给出建议,如进行深呼吸来平静下来。此外,GPT还能够在讲述睡前故事时根据要求调整语音的情感表达,如更具戏剧性或机器声。

💡线性方程

线性方程是数学中的一种方程,涉及变量的一次幂。视频中提到线性方程的解决,展示了GPT模型不仅能够进行语言处理,还能帮助用户解决数学问题。例如,视频中的一个人询问如何解线性方程,GPT通过提供解题步骤的提示来帮助用户,而不是直接给出答案。

💡机器学习模型

机器学习模型是指利用机器学习算法从数据中学习并做出预测或决策的模型。视频中提到的GPT 4是一个先进的机器学习模型,它结合了语音、文本和视觉智能,提供了更自然和高效的用户体验。GPT 4的发布是视频中的主要事件之一,标志着技术的一个重要进步。

💡用户界面(UI)

用户界面(UI)是指软件或设备的前端部分,用户通过它与系统进行交互。视频中提到了对GPT的UI进行了刷新,使其更简单、更自然,以便用户更容易地使用GPT的功能。这表明了设计者对于提升用户体验的重视。

💡桌面应用

桌面应用是指在计算机桌面操作系统上运行的软件应用程序。视频中提到了GPT的桌面版本发布,这意味着用户可以在没有网络浏览器的情况下,在本地计算机上使用GPT的功能,提高了使用的便捷性。

💡编程问题

编程问题是指在编写或理解计算机程序时遇到的问题。视频中,GPT帮助用户解决了一个编程问题,涉及获取特定位置和时间段的每日天气数据,并在代码中应用了一个函数来平滑温度数据。这展示了GPT在处理复杂任务,如数据分析和编程问题解决方面的应用潜力。

💡滚动平均

滚动平均是一种统计学上用来分析时间序列数据的方法,通过计算连续时间段内的平均值来平滑数据。在视频中,代码中的一个函数就是用来对温度数据进行滚动平均处理,以减少数据中的噪声,使趋势更加明显。

💡视觉能力

视觉能力在这里指的是模型处理和理解视觉信息的能力。视频中的GPT桌面应用展示了视觉能力,能够查看屏幕上的内容,如代码和图表,并对其进行分析和解释。这使得GPT能够更全面地与用户交互,提供更丰富的信息处理能力。

💡实时信息搜索

实时信息搜索是指能够即时搜索并获取最新信息的能力。视频中提到了GPT的浏览功能,用户可以在对话中搜索实时信息,这表明了GPT在提供最新、最相关信息方面的能力。

💡高级数据分析

高级数据分析涉及使用复杂数学和统计方法来分析数据,提取有用信息和洞察力。视频中提到GPT可以上传图表或其他信息,并进行分析,提供答案。这展示了GPT在处理和解释复杂数据集方面的高级能力。

Highlights

介绍新的旗舰模型gbt 4,它为包括免费用户在内的所有人带来了gbd4级别的智能。

gbt 40在语音、文本和视觉方面的能力得到了提升,并且使用更加自然和容易。

gbt 40原生支持语音、文本和视觉,极大提高了效率,并允许为免费用户提供GPT 4级别的智能。

超过1亿人使用chat GPT进行工作、学习和创造,现在这些高级工具也对免费用户开放。

展示了如何使用chat GPT解决线性方程,说明了数学在日常生活中的应用。

通过实时对话和演示,展示了gbt 40在实时语音转换和情感识别方面的能力。

演示了如何使用gbt 40的视觉效果解读和讨论代码以及生成的图表。

gbt 40的发布意味着高级AI工具对所有人免费开放,减少了使用门槛。

介绍了gbt 40在50多种语言中的改进,使其能够服务于更广泛的用户群体。

讨论了将新技术安全地引入世界时面临的挑战,以及如何与不同利益相关者合作以负责任地部署这些技术。

展示了gbt 40在实时语音模式下的交互,包括情绪感知和实时响应。

通过讲述一个关于机器人和爱的睡前故事,展示了模型在不同情感风格中生成语音的能力。

演示了如何使用gbt 40的视觉功能来解决数学问题,展示了模型的交互式学习和教学能力。

强调了线性方程在现实生活中的应用,如计算费用、规划旅行、烹饪以及商业利润和损失计算。

展示了如何通过chat GPT桌面应用程序和语音应用程序实时交互,以及如何将代码发送给模型进行分析。

演示了gbt 40如何通过实时翻译功能帮助不同语言的用户进行交流。

通过实时分析面部表情来识别情绪,展示了gbt 40在情感识别方面的高级能力。

Transcripts

play00:28

open

play00:40

[Music]

play00:58

for

play01:28

for

play01:58

for

play02:26

for hey chbt how are you I'm doing

play02:30

fantastic thanks for asking how about

play02:32

you how's your day going I'm uh doing

play02:34

great so I would like you to function as

play02:36

a translator I have a friend here who

play02:37

only speaks Italian and I only speak

play02:39

English and uh every time you hear

play02:41

English I want you to translate it to

play02:43

Italian and if you hear Italian I want

play02:45

you to translate it back to English is

play02:46

that

play02:57

good Mike she wonders if whales could

play03:00

talk what would they tell us um they

play03:03

might ask uh how do we solve linear

play03:16

equations certainly yes great looks like

play03:20

it works

play03:53

okay chat PT I'm sharing with you the

play03:56

plot now I'm wondering if you can give

play03:57

me a really brief one- sentence overview

play03:59

of what you see

play04:01

plot displays smoothed average minimum

play04:04

and maximum temperatures throughout 2018

play04:08

with a notable annotation marking a big

play04:10

rainfall event in late

play04:12

September fantastic I have another

play04:14

question for you um which months do you

play04:16

see the hottest temperatures and roughly

play04:18

what temperatures do those correspond

play04:20

to the hottest temperatures occur around

play04:23

July and August the maximum temperature

play04:26

during these months is roughly between

play04:29

25 de and 30° 77° fahit to 86°

play04:57

Fahrenheit ah there we go it looks like

play05:01

you're feeling pretty happy and cheerful

play05:03

with a big smile and maybe even a touch

play05:06

of excitement whatever's going on it

play05:08

seems like you're in a great mood care

play05:11

to share the source of those good vipes

play05:13

yeah no uh the reason I'm in a really

play05:14

good mood is because we were doing a

play05:15

presentation showcasing how useful and

play05:17

amazing you

play05:19

are oh stop it you're making me BL

play05:54

everyone Hi everyone thank you thank you

play05:58

it's great to have you here today today

play06:00

today I'm going to talk about three

play06:02

things that's it we will start with why

play06:06

it's so important to us to have a

play06:08

product that we can make freely

play06:10

available and broadly available to

play06:13

everyone and we're always trying to find

play06:15

out ways to reduce friction so everyone

play06:18

can use CH wherever they are so today

play06:22

we'll be releasing the desktop version

play06:24

of CH gbt and the refreshed UI that

play06:28

makes it simpler to use is much more

play06:30

natural as

play06:31

well but the big news today is that we

play06:34

are launching our new flagship model and

play06:38

we are calling it gbt

play06:40

4 the special thing about gbt 40 is that

play06:44

it brings gbd4 level intelligence to

play06:48

everyone including our free users we'll

play06:51

be showing some live demos today to show

play06:54

the full extent of the capabilities of

play06:57

our new model and we'll be rolling them

play06:59

out iteratively over the next few weeks

play07:03

all right so let's get

play07:06

started a very important part of our

play07:08

mission is to be able to make our

play07:11

Advanced AI tools available to everyone

play07:14

for free we think it's very very

play07:18

important that people have an intuitive

play07:20

feel for what the technology can do and

play07:23

so we really want to pair it with this

play07:26

broader understanding and we're always

play07:29

finding ways to reduce that friction and

play07:31

recently we made CH gbt available

play07:34

without the signup

play07:35

flow and today we're also bringing the

play07:39

desktop app to Chad gbt because we want

play07:43

you to be able to use it wherever you

play07:45

are as you can see it's easy it's simple

play07:48

it integrates very very easily in your

play07:51

workflow along with it we have also

play07:54

refreshed the

play07:56

UI we know that these models get more

play07:59

and more complex but we want the

play08:02

experience of interaction to actually

play08:05

become more natural easy and for you not

play08:09

to focus on the UI at all but just focus

play08:12

on the collaboration which had

play08:15

GPT and now the big news today we are

play08:19

releasing our newest Flagship model this

play08:23

is GPT for o

play08:26

[Applause]

play08:32

GPT 40 provides gp4 level intelligence

play08:36

but it is much faster and it improves on

play08:40

its capabilities across text vision and

play08:44

audio for the past couple of years we've

play08:47

been very focused on improving the

play08:49

intelligence of these models and they've

play08:51

gotten pretty good but this is the first

play08:55

time that we are really making a huge

play08:57

step forward when it comes to the EAS of

play09:01

use and this is incredibly important

play09:04

because we're looking at the future of

play09:06

interaction between ourselves and the

play09:09

machines and we think that GPT 40 is

play09:13

really shifting that Paradigm into the

play09:16

future of collaboration where this

play09:18

interaction becomes much more natural

play09:20

and Far Far

play09:23

easier but you know making this happen

play09:25

is actually quite complex because when

play09:27

we interact with one another there's a

play09:30

lot of stuff that we take for granted

play09:32

you know the ease of our dialogue when

play09:35

we interrupt one another the background

play09:38

noises the multiple voices in a

play09:41

conversation or you know understanding

play09:44

the tone of voice all of these things

play09:46

are actually quite complex for for these

play09:49

models and until now with voice mode we

play09:54

had three models that come together to

play09:56

deliver this experience you have

play09:59

description intelligence and then text

play10:02

to speech all comes together in

play10:04

orchestration to deliver voice mode this

play10:08

also brings a lot of latency to the

play10:11

experience and it really breaks that

play10:13

immersion in the collaboration which had

play10:16

gbt but now with gbt 40 this all happens

play10:21

natively gbt 4 reasons across voice text

play10:27

and vision

play10:30

and with these incredible efficiencies

play10:32

it also allows us to bring the GPT 4

play10:36

class intelligence to our free users

play10:39

this is something that we've been trying

play10:41

to do for many many months and we're

play10:44

very very excited to finally bring gbt

play10:47

40 to all of our

play10:50

users today we have 100 million people

play10:54

more than 100 million in fact they use

play10:57

chat GPT to create work

play11:00

learn and we have these Advanced tools

play11:05

that are only available to our paid paid

play11:08

users at least until now with the

play11:10

efficiencies of 40 we can bring these

play11:13

tools to

play11:14

everyone so starting today you can use

play11:18

gpts and the GPT

play11:21

store so far we've had more than a

play11:24

million users create amazing experiences

play11:27

with gpts these are custom chat gbts for

play11:30

specific use cases they're available in

play11:33

the store and now our Builders have a

play11:36

much bigger audience where you know

play11:39

University professors can create content

play11:41

for their students or podcasters can

play11:44

create content for their

play11:47

listeners and you can also use Vision so

play11:51

now you can upload um screenshots photos

play11:56

documents containing both text and

play11:58

images

play11:59

and you can start conversations with

play12:01

chat gbt about all of this content you

play12:04

can also use memory where it makes CH

play12:08

GPT far more useful and helpful because

play12:10

now it has a sense of continuity across

play12:13

of all your

play12:14

conversations and you can use browse

play12:17

where you can search for Real Time

play12:19

information in your conversation and

play12:22

Advanced Data analysis where you can

play12:24

upload charts or any information and it

play12:27

will analyze

play12:29

this information it will give you

play12:31

answers and so

play12:34

on lastly we have also improved on the

play12:38

quality and speed in 50 different

play12:42

languages for chbt and this is very very

play12:46

important because we want to be able to

play12:48

bring this experience to as many people

play12:50

out there as

play12:52

possible so we're very very excited to

play12:55

bring dut 40 to all of our free us users

play12:59

out there and for the paid users they

play13:01

will continue to have up to five times

play13:04

the capacity limits of our free

play13:08

users but GPT 40 is not only available

play13:12

in chat GPT we're also bringing it to

play13:15

the

play13:16

API

play13:22

so so our developers can start building

play13:25

today with GPD 40 and making amazing AI

play13:29

applications deploying them at scale 40

play13:32

is available at 2x faster 50% cheaper

play13:36

and five times higher rate limits

play13:39

compared to gbd4

play13:42

Turbo but you know is we bring these

play13:46

Technologies into the world it's quite

play13:49

challenging to figure out how to do so

play13:51

in a way that's both useful and also

play13:55

safe and gbd 40 presents new challenges

play13:59

for us when it comes to safety because

play14:01

we're dealing with realtime audio

play14:04

realtime

play14:05

vision and our team has been hard at

play14:08

work figuring out how to build in

play14:10

mitigations against

play14:13

misuse we continue to work with

play14:16

different stakeholders out there from

play14:18

government media entertainment all

play14:20

Industries red

play14:22

seamers Civil Society to figure out how

play14:25

to best bring these Technologies into

play14:28

the world so over the next few weeks

play14:31

we'll continue our iterative deployment

play14:33

to bring out all the capabilities to you

play14:37

but today I want to show you all these

play14:40

capabilities so we'll do some live demos

play14:44

I will bring on two of our research

play14:46

leads Mark Shen and Barrett Z to

play14:50

[Applause]

play14:57

join hi I'm I'm Barrett hey I'm Mark so

play15:01

one of the key capabilities we're really

play15:03

excited to share with you today is

play15:05

realtime conversational speech let's

play15:07

just get a demo fired up so I'm taking

play15:10

out a phone if you are wondering about

play15:12

this wire it's so we have consistent

play15:14

internet and if you see there's this

play15:16

little icon on the bottom right of the

play15:18

track GPT app and this will open up GPT

play15:20

40's audio

play15:24

capabilities hey chat GPT I'm Mark how

play15:26

are you oh mark

play15:29

I'm doing great thanks for asking how

play15:32

about you hey so I'm on stage right now

play15:35

I'm doing a live demo and frankly I'm

play15:36

feeling a little bit nervous can you

play15:38

help me calm my nerves a little

play15:40

bit oh you're doing a live demo right

play15:42

now that's awesome just take a deep

play15:46

breath and remember you're the expert I

play15:50

like that suggestion let me try a couple

play15:52

deep breaths can you give me feedback on

play15:54

my breaths okay here I go

play15:59

whoa

play16:01

slow a bit there mark you're not a

play16:05

vacuum cleaner breathe in for a count of

play16:09

four okay uh let me try again so I'm

play16:11

going to breathe in

play16:13

deeply and then

play16:15

breathe for four and then exhale

play16:20

slowly okay I'll try again breathing

play16:22

in and breathe

play16:25

out that's it how do you feel I feel a

play16:28

lot better thank you so

play16:30

much so Mark you've been working on

play16:33

these capabilities for a while now can

play16:35

you tell us a bit how it's different

play16:37

from voice mode right so if you've used

play16:39

our voice mode experience before you'll

play16:41

notice a couple key differences first

play16:43

you know you can now interrupt the model

play16:45

you don't have to wait for it to finish

play16:47

your turn before you can start speaking

play16:49

and you know you can just butt in

play16:51

whenever you want second the model is

play16:54

real time responsiveness so that means

play16:56

that you don't have this awkward 2 to 3

play16:58

second lag before you wait for the model

play17:00

to give a response and finally the model

play17:03

picks up on emotion right when I was

play17:05

breathing super hard there it could tell

play17:07

and it you hey you might want to calm

play17:09

down a little bit you're kind of running

play17:11

a little bit too fast so you know it

play17:13

really has that capability across the

play17:15

board to perceive your emotion not only

play17:18

that though the model is able to

play17:20

generate voice in a variety of different

play17:22

emotive Styles and it really has a wide

play17:24

dynamic range so I'm going to fire up

play17:26

another demo for that yeah

play17:32

hey chachu PT how are you doing I'm

play17:34

doing fantastic thanks for asking how

play17:37

about you pretty good what's up so my

play17:39

friend um Barrett here he's been having

play17:41

trouble sleeping lately and uh I want

play17:43

you to tell him a bedtime story about

play17:44

robots and love oh a bedtime story about

play17:49

robots and love I got you covered gather

play17:53

round Barrett once upon a time in a

play17:56

world not too different from ours there

play17:58

was a ro robot named bite bite was a

play18:01

curious robot always expl started this

play18:03

story but I want a little bit more

play18:05

emotion in your voice a little bit more

play18:06

drama got it let's add some drama once

play18:11

upon a time in a world not too different

play18:14

from ours there was a robot named no

play18:16

chat I really want maximal emotion like

play18:19

maximal expressiveness much more than

play18:20

you were doing before understood let's

play18:24

amplify the drama once upon a time in a

play18:28

world not too different from ours there

play18:31

was a robot named bite bite can you do

play18:35

this in a robotic voice now initiating

play18:38

dramatic robotic voice once upon a time

play18:43

in a world not too different from ours

play18:46

there was a robot named bite bite was a

play18:49

curious robot always exploring new yeah

play18:53

Barett here calm down a little bit can

play18:54

you end the story um but do it in a

play18:56

singing voice h ah and so bite found

play19:02

another robot friend and may live circly

play19:06

Ever After thank you so much I

play19:11

definitely feel a lot more

play19:15

[Music]

play19:17

relaxed you've just seen the voice

play19:20

capabilities but we also want to show

play19:22

you the vision capabilities as the model

play19:24

can also see the whole world around us

play19:27

yeah absolutely like as you can see you

play19:28

know you can chat with chat GPT in real

play19:30

time but you can also interact with it

play19:32

um with video as well okay let me boot

play19:35

up chat

play19:38

BT hey Chachi

play19:40

BT hello there how's it going it's going

play19:44

really well today I'd really like your

play19:46

help solving a math problem I'm all ears

play19:49

what math problem can I help you tackle

play19:51

today so I'm going to write down a

play19:53

linear equation on a sheet of paper and

play19:54

I'll show you and then I love your help

play19:55

working me through it but importantly

play19:57

don't tell me the solution just help

play19:58

give me hints along the

play20:00

way got

play20:03

it okay I see it no I didn't show you

play20:07

yet just give me help along the way one

play20:11

second whoops I got too excited I'm

play20:14

ready when you

play20:23

are so chat

play20:26

GPT what equation did I write down

play20:29

there ah I see it now you wrote down 3x

play20:34

+ 1 = 4 yep exactly so what's the first

play20:38

step I should take to try to solve this

play20:41

the first step is to get all the terms

play20:43

with X on one side and the constants on

play20:46

the other side so what do you think we

play20:48

should do with that plus one okay I'm

play20:51

going to try to subtract one from both

play20:52

sides and then I'll see what I

play20:54

get great

play20:56

idea perfect now now what do you get

play20:59

when you subtract one from both

play21:05

sides okay so chat GPT this is what I

play21:08

ended up with how does this look spot on

play21:12

now you've isolated the term with X on

play21:14

one side and it looks like 3x equal 3

play21:17

what do you think we should do next to

play21:19

solve for x I'm not sure can you give me

play21:21

another

play21:22

hint of course you have 3x and you want

play21:27

to find the value of x

play21:29

think about what operation would undo

play21:31

multiplication is it

play21:33

subtraction close but think of the

play21:36

opposite of multiplication because 3x

play21:39

means 3 * X okay I think it's division

play21:42

let me try to divide both sides by three

play21:44

and let me see what I get Bingo go ahead

play21:47

and divide both sides by

play21:54

three okay so chat GPT this is what I

play21:57

ended up with how does this look

play21:59

it looks perfect you've solved it and x

play22:02

equals 1 nicely done how do you feel

play22:05

about solving linear equations now yeah

play22:08

honestly i' I'm not pretty confident I

play22:11

can solve linear equations but why would

play22:13

you ever need to use this in the real

play22:14

world I'm really

play22:16

skeptical that's a fair question linear

play22:18

equations pop up in a lot of everyday

play22:21

situations even if we don't realize it

play22:24

for example calculating expenses

play22:26

planning travel cooking and and even in

play22:29

business for profit and loss

play22:30

calculations it's basically a way to

play22:33

solve problems where you need to find an

play22:35

unknown Value Plus wow okay you know I

play22:37

have a lot of unknown values in my life

play22:39

I'm totally convinced we I'm going to

play22:42

learn math

play22:44

now I love that attitude Matt can

play22:48

definitely help solve some of life's

play22:50

mysteries whenever you're ready to dive

play22:52

into more math I'm here for you anything

play22:54

else you'd like to tackle today so chbt

play22:57

I really love that you you know taught

play22:58

the value of math to my friend Mark and

play23:00

I wrote uh one last thing I'd love if

play23:02

you could take a look at of course I'd

play23:05

love to see what you wrote show it to me

play23:07

whenever you're ready okay so this is

play23:08

what I wrote down what do you

play23:11

see a I see I love chat chpt that's so

play23:16

sweet of

play23:17

you yeah well I really appreciate all

play23:19

the

play23:20

help so yeah as we can

play23:26

see wow that's quite the outfit you've

play23:29

got on yeah as we can see you know we

play23:31

can chat in real time to chat GPT and

play23:33

not only that we can interact with it

play23:35

with uh video as well and I swear like

play23:38

if you look at Barrett's GPT history

play23:40

it's all equations like

play23:43

this now that you know you can do linear

play23:46

equations can you show us something more

play23:48

complex I think so okay next let's try

play23:51

something uh coding related so let me

play23:53

take out um a computer for this

play24:00

okay so I'm going to open the chat gbt

play24:02

desktop app like miror was talking about

play24:06

before okay and to give a bit of

play24:08

background on what's going on so here we

play24:10

have um a computer and on the screen we

play24:12

have some code and then the chat gbt

play24:14

voice app is on the right so chat gbt

play24:16

will be able to hear me but it can't see

play24:18

anything on the screen so I'm going to

play24:19

highlight the code command C it and then

play24:22

that will send it to chat GPT and then

play24:23

I'm going to talk about the code to chat

play24:25

GPT hey chat GPT

play24:28

hey there how's it going yeah it's going

play24:31

really well I was wondering if you could

play24:32

help me with uh a coding problem today

play24:35

of course I'd love to help you out

play24:37

what's the coding problem you're dealing

play24:38

with okay I'm going to share with you

play24:40

some code one second sure thing take

play24:43

your

play24:46

time okay so I just shared some code

play24:49

with you could you give me a really

play24:50

brief one- sentence description of

play24:51

what's going on in the code this code

play24:54

fetches daily weather data for a

play24:56

specific location and time period

play24:58

Smooths the temperature data using a

play25:00

rolling average annotates a significant

play25:03

weather event on the resulting plot and

play25:05

then displays the plot with the average

play25:07

minimum and maximum temperatures over

play25:09

the year awesome thanks so much and in

play25:12

the code there's this function Fu could

play25:14

you describe to me what the plot would

play25:15

look like if I had versus didn't have

play25:17

this function sure the function Fu XY

play25:22

takes in a data frame X containing

play25:24

temperature data and a window size y for

play25:27

calculating the rolling mean it Smooths

play25:30

the average temperature minimum

play25:32

temperature and maximum temperature

play25:34

columns by applying a rolling mean over

play25:37

the specified window with the fu

play25:40

function the plot would show smoother

play25:42

temperature lines because the rolling

play25:44

mean reduces the noise or fluctuations

play25:47

in the data awesome this makes thanks so

play25:49

much so yeah I definitely will want to

play25:51

apply this function and so I'm going to

play25:52

uh generate the plot and then show it to

play25:54

you give me one

play25:56

second absolutely take your time time

play25:58

I'm excited to see the plot when you're

play26:00

ready yeah so so far chat gbt has only

play26:02

been able to see the code and now I'm

play26:04

going to run it and then I'll use the

play26:06

vision capabilities of the chbt desktop

play26:08

app so it can actually see everything

play26:09

going on on the screen as

play26:16

well okay chat GPT I'm sharing with you

play26:20

the plot now I'm wondering if you can

play26:21

give me a really brief one sentence

play26:22

overview of what you

play26:24

see plot displays smoothed average

play26:27

minimum and maximum temperatures

play26:30

throughout 2018 with a notable

play26:32

annotation marking a big rainfall event

play26:35

in late

play26:36

September fantastic I have another

play26:38

question for you um which months do you

play26:40

see the hottest temperatures and roughly

play26:42

what temperatures do those correspond

play26:44

to the hottest temperatures occur around

play26:47

July and August the maximum temperature

play26:50

during these months is roughly between

play26:53

25° and 30° 77° F to 8 86°

play26:59

F what's going on with the Y AIS is this

play27:03

in Celsius or Fahrenheit the Y AIS is in

play27:07

Celsius the temperatures are labeled

play27:09

accordingly such as average temperature

play27:11

minimum temperature and maximum

play27:15

temperature yeah so as we can see you

play27:17

know not only can cha PT help me solve

play27:19

very easy linear um algebra equations

play27:21

that I need help with it can also

play27:23

interact with you know code bases and

play27:25

see the outputs of plots and everything

play27:26

like this going on on a computer

play27:30

[Music]

play27:35

so a few minutes ago we asked the live

play27:38

audience on X to submit few requests for

play27:41

what they would like us to try out here

play27:43

so I will take a couple of

play27:47

prompts okay bot gasar wants to know if

play27:51

gb40 is capable of realtime

play27:54

translation Mark you want to try this

play27:56

one sure yeah let's do it I I speak

play27:59

Italian so we can just English Italian's

play28:01

try to do English Italian sure let's do

play28:05

it hey Chachi how are you I'm doing

play28:09

fantastic thanks for asking how about

play28:11

you how's your day going I'm uh doing

play28:12

great so I would like you to function as

play28:14

a translator I have a friend here who

play28:16

only speaks Italian and I only speak

play28:18

English and uh every time you hear

play28:20

English I want you to translate it to

play28:22

Italian and if you hear Italian I want

play28:23

you to translate it back to English is

play28:25

that good perfecto

play28:36

Mike she wonders if Wales could talk

play28:39

what would they tell us um they might

play28:42

ask uh how do we solve linear

play28:55

equations certainly yes great it looks

play28:58

like it

play29:00

[Applause]

play29:04

works all right so the next one is from

play29:07

burito John 78 who asks can you tell

play29:11

what you're feeling just by looking at

play29:13

your face Barrett you want to give this

play29:15

one a try absolutely let's try it

play29:20

out hey chat

play29:22

GPT hey there what's up how can I

play29:25

brighten your day today okay yeah so I'm

play29:27

going to show you um a selfie of what I

play29:29

look like and then I'd like you to try

play29:30

to see what emotions I'm feeling based

play29:32

on how I'm looking sounds like a fun

play29:35

challenge go ahead and show me that

play29:37

selfie and I'll put my emotional

play29:39

detective hat on okay so here's me so

play29:43

what kind of emotions do you think I'm

play29:45

feeling H it seems like I'm looking at a

play29:48

picture of a wooden surface oh you know

play29:50

what that was the thing I sent you

play29:51

before don't worry I'm not actually a

play29:53

table um okay so so take a take another

play29:56

look uh that makes more

play30:00

sense a there we go it looks like you're

play30:03

feeling pretty happy and cheerful with a

play30:06

big smile and maybe even a touch of

play30:08

excitement whatever's going on it seems

play30:11

like you're in a great mood care to

play30:13

share the source of those Good Vibes

play30:15

yeah no uh the reason I'm in a really

play30:16

good mood is because we were doing a

play30:17

presentation showcasing how useful and

play30:19

amazing you

play30:21

are oh stop it you're making me BL

play30:29

amazing well that's it for today on live

play30:32

demos as you can see this is this just

play30:35

feels so magical and and that's

play30:37

wonderful but we also want to remove

play30:40

some of the mysticism from the

play30:42

technology and bring it to you so you

play30:44

can try it for yourself so over the next

play30:46

few weeks we'll be rolling out these

play30:48

capabilities to everyone today has been

play30:51

very much focused on the free users and

play30:54

the new modalities and new products but

play30:56

we also care a lot about the next

play30:59

Frontier so soon we'll be updating you

play31:02

on our progress towards the next big

play31:05

thing and before we wrap up I just want

play31:08

to thank the incredible open AI team and

play31:11

also thanks to Jensen and the Nvidia

play31:13

team for bringing us the most advanced

play31:15

gpus to make this demo possible today

play31:19

and thank you all very very much for

play31:21

being a part of this today

play31:24

[Applause]

Rate This

5.0 / 5 (0 votes)

Etiquetas Relacionadas
人工智能实时翻译情感识别数学解题技术演示交互体验产品发布用户界面数据处理机器学习自然语言处理
¿Necesitas un resumen en inglés?