OpenAIが衝撃発表😱GPT-4oのストリーミング配信の映像に字幕をつけてみた!これは見逃せない、ライブデモの動画を日本語でチェックしよう😁

creativaier
14 May 202415:53

Summary

TLDR本ビデオでは、新しいフラッグシップモデル「GBT 4」の発表が行われ、無料ユーザーを含むすべてのユーザーにGP4レベルのインテリジェンスを提供します。実演デモを通じて、リアルタイムの会話型音声、感情の読み取り、視覚能力を含むモデルの機能が披露されます。例えば、線形方程式の解法を学ぶサポートや、コードの解析、気温データのプロット表示など、具体的なアプリケーションが示されています。さらに、リアルタイム翻訳や感情認識などの新機能も紹介され、チャットGPTの多様な用途が展開されています。

Takeaways

  • 🚀 新旗舰模型GBT 40发布,它为包括免费用户在内的每个人带来了GP4级别的智能。
  • 📱 实时对话语音功能允许用户在模型回应时打断,无需等待。
  • 🔢 模型具有实时响应性,减少了2到3秒的延迟。
  • 😮 模型能够识别情绪,如用户呼吸急促时建议用户冷静。
  • 🎭 模型能够以多种不同的情感风格生成声音。
  • 📈 模型还具有视觉能力,能够看到周围的世界。
  • 🧮 通过提示帮助用户解决线性方程,而不是直接给出答案。
  • 💻 模型可以处理代码并解释代码的功能和输出。
  • 🌡️ 代码示例涉及获取特定地点和时间段的每日天气数据,并在图表上平滑温度数据。
  • 🌐 GBT 40能够实时翻译,如将英语翻译成意大利语。
  • 😊 模型可以通过视觉输入(如照片)来识别和解释情绪。

Q & A

  • GBT4の特別な機能は何ですか?

    -GBT4はGP4レベルの知能をすべてのユーザー、無料ユーザーを含めて提供します。これにはリアルタイムでの会話や感情の感知、声の調整が可能です。

  • GBT4の音声機能のデモでは、どのような新機能が紹介されましたか?

    -デモでは、リアルタイムの応答性、会話中の割り込みの可能性、感情の認識と声の調節が可能であることが示されました。

  • GBT4の視覚能力についてのデモでは、どのような機能が紹介されましたか?

    -視覚能力については、書かれた数式を認識し、それに基づいてヒントを出す機能や、グラフの解析、温度の平滑化などの機能が紹介されました。

  • GBT4はどのようにしてユーザーの感情を認識しますか?

    -GBT4はユーザーの声のトーンや話し方から感情を読み取り、適切な反応を示すことができます。例えば、話が速すぎるときには落ち着くようアドバイスをします。

  • リアルタイムでの翻訳機能はどのように機能しますか?

    -GBT4はリアルタイムで言語間の翻訳が可能で、ユーザーが話した内容を即座に別の言語に翻訳し、会話をスムーズに進めることができます。

  • GBT4が数学の問題を解決する際のアプローチは何ですか?

    -GBT4は数学の問題を解く際に解答を直接教えるのではなく、ヒントやステップを提供してユーザーが自ら問題を解決できるようサポートします。

  • GBT4はどのようにユーザーのプレゼンテーションの緊張を和らげましたか?

    -GBT4はユーザーに深呼吸をするよう提案し、リラックスするための具体的な呼吸の指示を与えました。これにより、ユーザーは緊張を和らげることができました。

  • GBT4のエモーティブスタイルとは何ですか?

    -エモーティブスタイルは、GBT4が話す際に感情を込めたり、ドラマチックに話したりする能力を指します。これにより、より人間らしい対話が可能になります。

  • GBT4はビジョン機能をどのように使用してコードやプロットを分析しますか?

    -GBT4は画面上のコードやプロットを認識し、その内容を解析して、データの解釈やプロットの特徴をユーザーに説明します。

Outlines

00:00

🚀 新型フラグシップモデルGBT 40の発表

本段落では、GBT 40という新フラグシップモデルの発表が行われています。GBT 40は、無料で利用できるユーザーも含め、誰もがGP4レベルの知能を提供できるという特徴があります。また、その機能を実際に見せるためのライブデモが行われ、今後数週間で段階的に展開される予定です。デモでは、Markがライブデモを行っている状況で緊張していると語り、AIが彼に深呼吸を促します。さらに、声の感情を感知し、声を生成する能力についても触れています。

05:02

👀 ビジョン機能とリアルタイム対話

本段落では、AIのビジョン機能とリアルタイム対話の能力が紹介されています。Markは、携帯電話を使ってAIと会話するデモを開始し、AIが音声応答を提供します。また、AIは感情を感知し、声の調子を調整することができます。さらに、AIは様々な感情的なスタイルで声を生成することができ、幅広い動的レンジを持っています。また、数学の問題を解決する際のヒントを提供し、線形方程式を解決するプロセスを学ぶことができます。

10:05

🤖 ロボットと愛について物語る

本段落では、AIの声の感情表現について触れています。Markは、AIにロボットと愛について物語るよう頼み、声の感情を増やしてほしいと要求しています。AIは、感情表現を増やしながら物語を進め、最後に歌いながら物語を終えます。また、AIはリアルタイム翻訳機能も持っており、英語とイタリア語の間の翻訳を行うことができます。

15:05

📈 データ可視化と感情認識

本段落では、AIがコードを解析し、プロットを生成する様子が描かれています。AIは、特定の場所と期間の天気データを取得し、平均温度を表示するプロットを作成することができます。また、AIは顔から感情を認識し、Markの感情を楽しんでいると推測しています。さらに、AIはリアルタイム翻訳機能を使って英語とイタリア語の会話を翻訳することができます。

Mindmap

Keywords

💡flagship model

「フラグシップモデル」とは、企業が展開する製品ラインナップの中で最も高級で、最新の技術や機能を搭載したモデルを指します。このビデオでは、新たに発表されたGBT 40というフラグシップモデルが、GP4レベルの知能を誰もが利用できるようにしたという点で注目されています。

💡realtime conversational speech

「リアルタイム会話型スピーチ」とは、音声認識技術を用いて、話者の声をリアルタイムで認識し、応答することができる機能です。ビデオでは、この機能を使ってライブデモを行っており、話者と自然なやりとりが行える様子が示されています。

💡emotion

「感情」とは、人間の内面的な状態を表す言葉で、喜び、悲しみ、怒り、楽しさなど多岐にわたります。ビデオでは、GBT 40が感情を感知し、それに応じた声の調子で話すデモンストレーションが行われています。

💡bedtime story

「ベッドタイムストーリー」とは、子供たちが寝る前に聞くための物語です。ビデオでは、ロボットと愛について物語を要求し、感情豊かに話すように指示することで、GBT 40の感情表現能力が示されています。

💡linear equation

「一次方程式」とは、未知数が1次である線形方程式のことで、日常生活でよく使われます。ビデオでは、数学の問題を解決する際のヒントを提供し、一次方程式の重要性が説明されています。

💡vision capabilities

「ビジョン機能」とは、画像や視覚的な情報を認識する能力です。ビデオでは、GBT 40がコードやプロットを視覚的に認識し、それに基づいて対話を行うデモンストレーションがあります。

💡coding problem

「コーディング問題」とは、プログラミングやコンピュータサイエンスの分野で発生する問題です。ビデオでは、GBT 40がコードを分析し、説明することで、その理解力と対話能力が示されています。

💡real time translation

「リアルタイム通訳」とは、話された言葉を瞬時に別の言語に翻訳する機能です。ビデオでは、英語とイタリア語の間での通訳が行われ、GBT 40の多言語対応能力が示されています。

💡emotional detection

「感情検出」とは、人々の表情や声の調子などから感情を推定する技術です。ビデオでは、顔写真から感情を推測するデモンストレーションを行い、GBT 40の感情認識機能が示されています。

💡dynamic range

「ダイナミックレンジ」とは、音声や音楽の音量の変化範囲を指します。ビデオでは、GBT 40が様々な感情の調子で声を生成できると説明されており、ダイナミックレンジの広さが強調されています。

💡mathematical problem solving

「数学問題解決」とは、数学的な問題に対処し、解答を見つけるプロセスです。ビデオでは、一次方程式を解決する手順が学ばれ、数学の現実世界での応用が議論されています。

Highlights

Launching of a new flagship model called GBT 4

GBT 4 brings GP4 level intelligence to all users, including free users

Live demos will be shown to display the capabilities of the new model

Real-time conversational speech capability

Model can be interrupted and responds in real-time without lag

Model picks up on user's emotions and responds accordingly

Model generates voice in a variety of emotive styles

Bedtime story demo with variable levels of emotion and drama

Vision capabilities allow the model to see and interact with the world

Assistance with solving a math problem using hints rather than providing the solution

Explanation of the practical applications of linear equations in everyday life

Coding problem assistance with a focus on understanding rather than direct answers

Description of a code's functionality related to fetching and smoothing weather data

Real-time translation capabilities between English and Italian

Emotion detection based on facial expressions

Interactive coding assistance with the ability to see and discuss code outputs

Demonstration of the model's ability to understand and discuss complex data plots

GBT 4's ability to perceive and respond to a wide range of human emotions

The model's application in solving real-world problems through various capabilities

Transcripts

play00:03

but the big news today is that we are

play00:05

launching our new flagship model and we

play00:09

are calling it gbt

play00:11

4 the special thing about gbt 40 is that

play00:15

it brings gp4 level intelligence to

play00:18

everyone including our free users we'll

play00:22

be showing some live demos today to show

play00:24

the full extent of the capabilities of

play00:27

our new model and we'll be rolling them

play00:30

out iteratively over the next few weeks

play00:34

all right so let's get

play00:44

started hi I'm uh I'm Barrett hey I'm

play00:47

Mark so one of the key capabilities

play00:49

we're really excited to share with you

play00:50

today is realtime conversational speech

play00:53

let's just get a demo fired up so I'm

play00:56

taking out a phone if you are wondering

play00:58

about this wire it's so we have

play01:00

consistent internet and if you see

play01:02

there's this little icon on the bottom

play01:04

right of the trat GPT app and this will

play01:06

open up GPT for's audio

play01:10

capabilities hey chat GPT I'm Mark how

play01:13

are you oh

play01:15

Mark I'm doing great thanks for asking

play01:18

how about you hey so I'm on stage right

play01:21

now I'm doing a live demo and frankly

play01:23

I'm feeling a little bit nervous can you

play01:25

help me calm my nerves a little

play01:26

bit oh you're doing a live demo right

play01:29

now that's awesome just take a deep

play01:32

breath and remember you're the expert I

play01:36

like that suggestion let me try a couple

play01:38

deep breaths can you give me feedback on

play01:40

my breaths okay here I

play01:44

go whoa

play01:47

slow a bit there mark you're not a

play01:51

vacuum cleaner breathe in for a count of

play01:55

four okay uh let me try again so I'm

play01:58

going to breathe in deeply

play02:00

and then

play02:02

breath for four and then exhale

play02:06

slowly okay I'll try again breathing

play02:09

in and breathe

play02:11

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

play02:15

lot better thank you so

play02:17

much so Mark you've been working on

play02:19

these capabilities for a while now can

play02:21

you tell us a bit how it's different

play02:23

from voice mode right so if you've used

play02:26

our voice mode experience before you'll

play02:28

notice a couple key differences first

play02:30

you know you can now interrupt the model

play02:32

you don't have to wait for it to finish

play02:34

your turn before you can start speaking

play02:36

and you know you can just butt in

play02:37

whenever you want second the model is

play02:40

real time responsiveness so that means

play02:43

that you don't have this awkward 2 to 3

play02:45

second lag before you wait for the model

play02:47

to give a response and finally the model

play02:50

picks up on emotion right when I was

play02:51

breathing super hard there it could tell

play02:53

and it you hey you might want to calm

play02:55

down a little bit you're kind of running

play02:57

a little bit too fast so you know it

play03:00

really has that capability across the

play03:02

board to perceive your emotion the model

play03:04

is able to generate voice in a variety

play03:06

of different emotive Styles and it

play03:08

really has a wide dynamic range so I'm

play03:10

going to fire up another demo for that

play03:16

yeah hey chat PT how are you doing I'm

play03:19

doing fantastic thanks for asking how

play03:22

about you pretty good what's up so my

play03:24

friend um Barrett here he's been having

play03:26

trouble sleeping lately and uh I want

play03:28

you to tell him a bedtime story about

play03:29

robots and love oh a bedtime story about

play03:34

robots and love I got you covered gather

play03:38

round Barrett once upon a time in a

play03:40

world not too different from ours there

play03:43

was a robot named bite bite was a

play03:46

curious robot always expl started this

play03:48

story but I want a little bit more

play03:50

emotion in your voice a little bit more

play03:51

drama got it let's add some drama once

play03:56

upon a time in a world not too different

play03:58

from ours there was a robot named no CHT

play04:02

I really want maximal emotion like

play04:04

maximal expressiveness much more than

play04:05

you were doing before understood let's

play04:09

amplify the drama once upon a time in a

play04:13

world not too different from ours there

play04:16

was a robot named bite bite can you do

play04:20

this in a robotic voice now initiating

play04:23

dramatic robotic voice once upon a time

play04:28

in a world not too different from ours

play04:31

there was a robot named bite bite was a

play04:34

curious robot always exploring new yeah

play04:38

Barett here calm down a little bit can

play04:39

you end the story um but do it in a

play04:41

singing

play04:44

voice and soite found another robot

play04:48

friend and they live circly Ever After

play04:54

thank you so much I definitely feel a

play04:56

lot more relaxed

play04:58

[Applause]

play05:02

youve just seen the voice capabilities

play05:04

but we also want to show you the vision

play05:06

capabilities as the model can also see

play05:09

the whole world around us yeah

play05:10

absolutely like as you can see you know

play05:12

you can chat with chat GPT in real time

play05:14

but you can also interact with it um

play05:16

with video as well okay let me boot up

play05:20

chat

play05:21

PT hey chat

play05:24

PT hello there how's it going it's going

play05:28

really well today I'd really like your

play05:30

help solving a math problem I'm all ears

play05:33

what math problem can I help you tackle

play05:35

today so I'm going to write down a

play05:36

linear equation on a sheet of paper and

play05:38

I'll show you and then I love your help

play05:39

working me through it but importantly

play05:41

don't tell me the solution just help

play05:42

give me hints along the

play05:43

way got

play05:47

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

play05:51

yet just give me help along the way one

play05:54

second whoops I got too excited I'm

play05:58

ready when you are

play06:08

so

play06:10

chpt what equation did I write down

play06:12

there ah I see it now you wrote down 3x

play06:18

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

play06:22

step I should take to try to solve this

play06:24

the first step is to get all the terms

play06:27

with X on one side and the constants on

play06:30

the other side so what do you think we

play06:32

should do with that plus one okay I'm

play06:34

going to try to subtract one from both

play06:36

sides and then I'll see what I

play06:37

get great

play06:40

idea perfect now what do you get when

play06:43

you subtract one from both

play06:49

sides okay so chat GPT this is what I

play06:51

ended up with how does this look spot on

play06:55

now you've isolated the term with X on

play06:57

one side and it looks like three x equal

play07:00

3 what do you think we should do next to

play07:03

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

play07:05

another

play07:06

hint of course you have 3x and you want

play07:10

to find the value of x think about what

play07:13

operation would undo multiplication is

play07:16

it

play07:17

subtraction close but think of the

play07:19

opposite of multiplication because 3x

play07:22

means 3 * X okay I think it's division

play07:26

let me try to divide both sides by three

play07:28

and let me see what I get Bingo go ahead

play07:31

and divide both sides by

play07:37

three okay so chat gbt this is what I

play07:40

ended up with how does this look it

play07:43

looks perfect you've solved it and x

play07:46

equal 1 nicely done how do you feel

play07:49

about solving linear equations now yeah

play07:52

honestly I've I'm not pretty confident I

play07:54

can solve linear equations but why would

play07:56

you ever need to use this in the real

play07:57

world I'm really skeptical

play08:00

that's a fair question linear equations

play08:03

pop up in a lot of everyday situations

play08:05

even if we don't realize it for example

play08:08

calculating expenses planning travel

play08:11

cooking and even in business for profit

play08:13

and loss calculations it's basically a

play08:16

way to solve problems but you need to

play08:18

find an unknown Value Plus wow okay you

play08:21

know I have a lot of unknown values in

play08:22

my life I'm totally convinced we I'm

play08:25

going to learn math now

play08:29

I love that attitude Matt can definitely

play08:32

help solve some of life's mysteries

play08:34

whenever you're ready to dive into more

play08:36

math I'm here for you anything else

play08:38

you'd like to tackle today so chbt I

play08:40

really love that you you know taught the

play08:42

value of math to my friend Mark and I

play08:44

wrote one last thing I'd love if you

play08:46

could take a look at of course I'd love

play08:48

to see what you wrote show it to me

play08:50

whenever you're ready okay so this is

play08:52

what I wrote down what do you

play08:54

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

play09:00

sweet of

play09:01

you yeah well I really appreciate all

play09:03

the

play09:10

help so let me take out um a computer

play09:13

for

play09:17

this okay so I'm going to open the chat

play09:20

gbt desktop app like miror was talking

play09:22

about

play09:25

before okay and to give a bit of

play09:27

background of what's going on so here we

play09:28

have um a computer and on the screen we

play09:30

have some code and then the chat gbt

play09:32

voice app is on the right so chat gbt

play09:34

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

play09:36

anything on the screen so I'm going to

play09:38

highlight the code command see it and

play09:40

then that will send it to chat GPT and

play09:41

then I'm going to talk about the code to

play09:42

chat

play09:44

GPT hey chat

play09:46

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

play09:49

going really well I was wondering if you

play09:50

could help me with a coding problem

play09:52

today of course I'd love to help you out

play09:55

what's the coding problem you're dealing

play09:57

with okay I'm going to share with you

play09:59

some code one second sure thing take

play10:02

your

play10:04

time okay so I just shared some code

play10:07

with you could you give me a really

play10:08

brief one- sentence description of

play10:10

what's going on in the code this code

play10:12

fetches daily weather data for a

play10:14

specific location and time period

play10:16

Smooths the temperature data using a

play10:18

rolling average annotates a significant

play10:21

weather event on the resulting plot and

play10:23

then displays the plot with the average

play10:25

minimum and maximum temperatures over

play10:28

the year awesome thanks so much and in

play10:31

the code there's this function Fu could

play10:32

you describe to me what the plot would

play10:34

look like if I had versus didn't have

play10:35

this function sure the function Fu XY

play10:40

takes in a data frame X containing

play10:42

temperature data and a window size y for

play10:46

calculating the rolling mean it Smooths

play10:48

the average temperature minimum

play10:51

temperature and maximum temperature

play10:53

columns by applying a rolling mean over

play10:55

the specified window with the fu

play10:58

function the plot would show smoother

play11:01

temperature lines because the rolling

play11:02

mean reduces the noise or fluctuations

play11:05

in the data awesome this makes thanks so

play11:08

much so yeah I definitely will want to

play11:09

apply this function and so I'm going to

play11:11

uh generate the plot and then show it to

play11:12

you give me one

play11:14

second absolutely take your time I'm

play11:17

excited to see the plot when you're

play11:18

ready yeah so so far chat gbt has only

play11:21

been able to see the code and now I'm

play11:22

going to run it and then I'll use the

play11:24

vision capabilities of the chat PT

play11:26

desktop app so it can actually see

play11:27

everything going on on the screen as

play11:28

well

play11:36

okay chat GPT I'm sharing with you the

play11:38

plot now I'm wondering if you can give

play11:39

me a really brief one- sentence overview

play11:41

of what you

play11:42

see the plot displays smoothed average

play11:46

minimum and maximum temperatures

play11:48

throughout 2018 with a notable

play11:51

annotation marking a big rainfall event

play11:53

in late

play11:55

September fantastic I have another

play11:56

question for you um which months do you

play11:58

see the hottest temperatures and roughly

play12:00

what temperatures do those correspond

play12:02

to the hottest temperatures occur around

play12:05

July and August the maximum temperature

play12:09

during these months is roughly between

play12:11

25° and 30° 77° fenhe to 86° fhe what's

play12:19

going on with the y axis is this in

play12:22

Celsius or Fahrenheit the Y AIS is in

play12:25

Celsius the temperatures are labeled

play12:27

accordingly such as average temperature

play12:30

minimum temperature and maximum

play12:33

temperature yeah so as we can see you

play12:36

know not only can chat PT help me solve

play12:37

very easy linear um algebra equations

play12:40

that I need help with it can also

play12:41

interact with you know code bases and

play12:43

see the outputs of plots and everything

play12:45

like this going on on a

play12:52

computer we asked the live audience on X

play12:55

to submit few requests for what they

play12:57

would like us to try out here so I will

play13:00

take a couple of

play13:02

prompts okay bot gasar wants to know if

play13:07

gbd4 is capable of real time

play13:10

translation Mark you want to try this

play13:12

one sure yeah let's do it I speak

play13:14

Italian so we can English Italian try to

play13:17

do English Italian sure let's do

play13:20

it hey Chachi P how are you I'm doing

play13:24

fantastic thanks for asking how about

play13:26

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

play13:28

great so I would like you to function as

play13:30

a translator I have a friend here who

play13:32

only speaks Italian and I only speak

play13:34

English and uh every time you hear

play13:36

English I want you to translate it to

play13:37

Italian and if you hear Italian I want

play13:39

you translate it back to English is that

play13:41

good

play13:51

perfo Mike she wonders if whales could

play13:54

talk what would they tell us um they

play13:57

might ask uh how do we solve linear

play14:11

equations certainly yes great looks like

play14:14

it

play14:19

works so the next one is from burito

play14:22

John 78 who asks can you tell what

play14:26

you're feeling just by looking at your

play14:28

face Barrett you want to give this one a

play14:30

try absolutely let's try it

play14:35

out hey chat

play14:37

CHT hey there what's up how can I

play14:40

brighten your day today okay yeah so I'm

play14:42

going to show you um a selfie of what I

play14:44

look like and then I'd like you to try

play14:45

to see what emotions I'm feeling based

play14:47

on how I'm looking sounds like a fun

play14:49

challenge go ahead and show me that

play14:51

selfie and I'll put my emotional

play14:53

detective hat on okay so here's me so

play14:57

what kind of emotions do you think I'm

play14:58

feeling

play15:01

H it seems like I'm looking at a picture

play15:03

of a wooden surface oh you know what

play15:05

that was the thing I sent you before

play15:06

don't worry I'm not actually a table um

play15:08

okay so so take a take another look uh

play15:11

that makes more

play15:14

sense ah there we go it looks like

play15:17

you're feeling pretty happy and cheerful

play15:20

with a big smile and maybe even a touch

play15:22

of excitement whatever's going on it

play15:25

seems like you're in a great mood care

play15:27

to share the source of those Good Vibes

play15:30

yeah no uh the reason I'm in a really

play15:31

good mood is because we were doing a

play15:32

presentation showcasing how useful and

play15:34

amazing you

play15:35

are oh stop it you're making me BL

Rate This

5.0 / 5 (0 votes)

Related Tags
GBT 4会話スピーチ感情認識リアルタイム翻訳機能デモンストレーション数学問題コード解析視覚機能AI技術オンラインデモ
Do you need a summary in English?