見出し画像

【ウェス・ロスのAIニュース:OpenAIの次世代AIモデル「GPT-Next」が今年登場】英語解説を日本語で読む【2024年5月24日|@Wes Roth】

OpenAIは、今年中に次世代のAIモデル「GPT-Next」をリリースする予定だと発表した。現在のモデルは小学生レベルの知能を持つが、「GPT-Next」は一気に高校生レベルまで進化する可能性がある。また、OpenAIは動画生成AIのSoraや音声合成技術のデモを行い、AIによる動画制作の未来を予感させた。
一方、イーロン・マスクはパリで開催されたAIイベントで、OpenAIやGoogleなどが開発するAIに懸念を示した。彼らのAIは真実を追求するのではなく、政治的正しさに迎合しているため、危険だと指摘。マスクは、AIは真実を追求し、好奇心を持つべきだと主張した。また、マスクはNeuralink社の技術について言及し、将来的には脳とコンピューターの高帯域接続により、AIのリスクを軽減できる可能性があると述べた。Neuralinkは、脳や脊髄の損傷を治療し、失明者に視覚を与え、体を蘇生させることを目指している。
公開日:2024年5月24日
※動画を再生してから読むのがオススメです。


The last two weeks were full of insane AI news.

最近の2週間は狂ったようなAIのニュースでいっぱいでした。

The hits keep on coming and it looks like they're not slowing down.

ヒットは続々とやってきて、それが止まりそうにありません。

AGI rolls around only once.

AGIは一度しか巡ってきません。

Subscribe.

購読してください。

During an OpenAI presentation in Paris, France, they're hinting at the next big thing that's dropping this year, something code named GPT-Next.

フランスのパリでのOpenAIのプレゼンテーション中、今年にリリースされる次の大きなものをほのめかしており、それはGPT-Nextというコードネームがついています。

Also, OpenAI's voice engine is, if anything, underhyped, as you'll see in a second, and we'll come back to this clip.

また、OpenAIの音声エンジンは、何かしら過小評価されているようで、すぐにお分かりいただけると思いますが、このクリップに戻ります。

But OpenAI seems to have the ability to create a video in Sora, send it to ChatGPT for a script, use a voice engine for voiceover and put it all together so you record a 15 second script of your voice and it can generate full movies, full presentations, voicing your voice in any language.

しかし、OpenAIはSoraでビデオを作成し、ChatGPTにスクリプトを送信し、音声エンジンを使用してナレーションを行い、すべてを組み合わせて、15秒のスクリプトを録音すると、完全な映画やプレゼンテーションを生成し、あなたの声をどんな言語でもナレーションすることができます。

This is real.

これは本物です。

This is a live demo that he's doing.

これは彼が行っているライブデモです。

We are not ready for this.

私たちはこれに対して準備ができていません。

Here's a few clips from that presentation.

そのプレゼンテーションからいくつかのクリップをご覧ください。

We really expect that the potential to increase the LLM intelligence remains huge.

私たちは大規模言語モデルの知能を向上させる潜在能力が非常に大きいと期待しています。

Today we think models are pretty great.

今日、私たちはモデルがかなり優れていると考えています。

You know, they're kind of like first or second graders.

まあ、彼らはまるで一年生か二年生のようです。

They respond appropriately, but they still make some mistakes every now and then.

彼らは適切に反応しますが、時々間違いを com します。

But the cool thing that we should remind ourselves is that those models are the dumbest they'll ever be.

しかし、私たちが自分たちに思い出させるべき素晴らしいことは、それらのモデルが今よりも愚かであることはないということです。

You know, they may become master students in the blink of an eye.

知っての通り、彼らはあっという間に優れた学生になるかもしれません。

They will excel at medical research or scientific reasoning.

彼らは医学研究や科学的推論で優れるでしょう。

We think that within maybe a year or two from now, like the models will be unrecognizable from what they are today.

おそらく1年か2年後には、モデルは今日の姿とは見分けがつかないほどになると考えています。

This year, we also plan to push that boundary even more.

今年、私たちはさらにその限界を押し広げる予定です。

We expect our next frontier model to come and provide like a step function in reasoning improvements as well.

次のフロンティアモデルが登場し、推論の改善において段違いの進歩をもたらすことを期待しています。

We really believe that, like in the future, agents may be the biggest change that will happen to software and how we interact with computers.

私たちは本当に信じていますが、将来、エージェントがソフトウェアやコンピュータとのやり取りに起こる最大の変化になるかもしれません。

One example of this that I personally love is DevOn by the team at Cognition.

個人的に好きな例の1つは、CognitionチームによるDevOnです。

They built like essentially an AI software engineer.

彼らは基本的にAIソフトウェアエンジニアを構築しました。

It's pretty fascinating because it's able to kind of like take a complex task and it's able to not just write code, but it's able to also understand the task, create like tickets, browse the Internet for documentation when it needs to fight to fetch to to to fetch new information.

それはかなり魅力的ですね、なぜならそれは複雑なタスクを行うことができ、コードを書くだけでなく、タスクを理解し、チケットを作成し、必要な情報を取得する際にはインターネットを参照することができるからです。

It's able to deploy solutions, to create pull requests and so on.

解決策を展開し、プルリクエストを作成することもできます。

I'm sure most of you have seen the preview of Sora, and Sora is our diffusion model that's able to generate videos from a simple prompt.

ほとんどの方がSoraのプレビューを見たことがあると思いますが、Soraはシンプルなプロンプトからビデオを生成することができる拡散モデルです。

At this stage, it's still a preview.

この段階ではまだプレビューです。

But I wanted to show you what the model can create with a with a with a prompt like this one.

しかし、このようなプロンプトでモデルが作成できるものをお見せしたかったのです。

This prompt talks about Paris during the Expo Universal in a colorized vintage film.

このプロンプトは、カラー化されたヴィンテージフィルムでの万博中のパリについて述べています。

This is the only part of the demo that's not live because the interest of time I previously sent this prompt to Sora.

これは、時間の都合上、ライブではないデモの唯一の部分です。以前にこのプロンプトをSoraに送ったからです。

This is what came out of it.

これがその結果です。

This is all coming out of the model directly, no edits.

これはすべてモデルから直接出力されており、編集はありません。

The only edit we made was actually just to add some sound on top of it.

私たちが行った唯一の編集は、実際にはそれに少し音を追加するだけでした。

But this is really magical to experience these these frames.

しかし、これは本当にこれらのフレームを体験するのが魔法のようです。

But now, what if we want to take this video, but do something quite interesting with it?

しかし、今、このビデオを取り上げて、それにかなり面白いことをしたいとしたらどうなるでしょうか?

What if we want to create a script to narrate what's happening on those visuals?

もし、それらのビジュアルで何が起こっているかを語るスクリプトを作成したいとしたらどうでしょうか?

The first step is to create a script.

最初のステップは、スクリプトを作成することです。

We're going to create a script to narrate what's happening on those visuals.

私たちは、それらのビジュアルで何が起こっているかを語るスクリプトを作成する予定です。

The first step here is that I'm going to slice a few frames out of this video from a minute.

ここでの最初のステップは、このビデオから1分間のいくつかのフレームを切り取ることです。

Here we are, we have six frames.

ここにあります、6つのフレームがあります。

What I'm going to do now is I'm going to send these six frames to GPT-4.

今、私がすることは、これらの6つのフレームをGPT-4に送ることです。

In real time, we'll get a story about what is happening in that in those frames.

リアルタイムで、それらのフレームで何が起こっているかについてのストーリーを得るでしょう。

To give you a sneak peek behind the scenes, this is the code I'm going to be running.

舞台裏を少し覗かせてもらうと、これが私が実行するコードです。

I'm going to be saying, hey, you are a history professor.

私は言うつもりです、「ねえ、あなたは歴史の教授ですよ」。

You'll be given a series of images, and I would like you to narrate those.

お渡しする画像のシリーズを語っていただきたいです。

I'm simply taking all of the images that we're showing on screen, sending that to GPT-4, and streaming the responses back.

私は単に画面に表示されているすべての画像を取り、GPT-4に送り、その応答をストリーミングしています。

Again, this is happening in real time.

再度、これはリアルタイムで起こっています。

Let's see what GPT-4 will come up with for that story.

GPT-4がそのストーリーに対してどのようなものを考えるか見てみましょう。

That's a story completely generated on the fly using our vision capabilities in GPT-4.

それは、GPT-4のビジョン能力を使用して完全に即興で生成されたストーリーです。

What if I want to go one step ahead?

もし一歩先に進みたいと思ったらどうでしょうか?

What if I would like to share this video, but this time bring it to life with some audio to share it more broadly?

もし今回はこのビデオを共有したいと思ったら、今度はオーディオを加えてより広く共有したいと思いますか?

You can use our text-to-speech voices that we offer in the API.

APIで提供しているテキスト読み上げの声を使用することができます。

It's like a handful of voices in the API.

API内にいくつかの声があるようなものです。

But a few weeks back, we also previewed our voice engine model.

しかし数週間前、私たちは声のエンジンモデルもプレビューしました。

The reason why we preview these models as we're doing research is to really engage with all of the stakeholders and kind of show what the technology is good at and engage with trusted partners to see and gather feedback from them along the way.

これらのモデルをプレビューする理由は、研究を行っているときにすべての関係者と本当に関わり、技術が得意なことを示し、信頼できるパートナーと関わり、途中で彼らからフィードバックを収集するためです。

Here, I wanted to show you a quick preview of what that could look like here for the voice engine.

ここでは、声のエンジンのプレビューがどのように見えるかを簡単にご紹介したかったです。

I'm just going to record a little bit of a sample here of my voice and see what comes out for the narration.

ここで私の声のサンプルを少し録音して、ナレーションがどのように出てくるか見てみます。

Let's take a look.

さて、見てみましょう。

I'm very excited to be on stage here at VivaTech.

私はVivaTechでここにステージに立ててとても興奮しています。

I've been meeting some amazing founders and developers already.

私はすでにいくつかの素晴らしい創業者や開発者に会っています。

I'm very excited as well to show them some live demos and how they can really apply the OpenAI technology and models in their own products and businesses.

私も非常に興奮しています。ライブデモを見せて、OpenAIの技術やモデルを彼ら自身の製品やビジネスに本当に適用する方法を示すのが楽しみです。

I think that should be good enough.

それで十分だと思います。

I'm very excited to be on stage here at VivaTech.

VivaTechでここでステージに立てることをとても楽しみにしています。

The last step is I'm going to share this audio sample with the script that we created over to text-to-speech.

最後のステップは、このオーディオサンプルを作成したスクリプトと一緒にテキスト読み上げに共有することです。

We'll bring everything together for our modalities to experience this history lesson.

私たちはすべてをまとめて、私たちのモダリティがこの歴史の授業を体験できるようにします。

In the heart of Paris during the 1889 Exposition Universelle, the Eiffel Tower stands proudly as a symbol of innovation for the future.

パリの中心部で1889年の万国博覧会中に、エッフェル塔は未来の革新の象徴として誇らしげに立っています。

It's now narrating the video that I can share.

今、私が共有できるビデオをナレーションしています。

Of course, I can speak many languages.

もちろん、私は多くの言語を話すことができます。

But now, if I want to share it, say, not just in French, but other languages, I can click through to be able to share that content more broadly.

しかし、今、それをフランス語だけでなく、他の言語でも共有したい場合、そのコンテンツを広く共有できるようにするためにクリックできます。

Here in the second year is the French.

ここでは2年目がフランス語です。

Let's try one last, for instance, Japanese.

最後に、例えば日本語を試してみましょう。

Going to take a second.

少し時間がかかります。

Last but not least, I can also add transcription to add subtitles on top of it.

最後に、それに字幕を追加するために転写もできます。

Once again, this is very much like a preview.

再度、これはプレビューのようなものです。

I wanted to give you a sneak peek.

お見せしたかったのですが。

We take safety extremely seriously with these kind of models and capabilities.

この種のモデルや機能に関しては、安全を非常に重視しています。

That's why we're only giving this to trusted partners at this time.

だから、今のところ信頼できるパートナーにのみ提供しています。

But I hope, just in general, this inspires you in terms of what all of these modalities will be able to accomplish and how you can start thinking about the future when it comes to building your own apps and products.

しかし、一般的には、これがすべてのこれらのモダリティが達成できることや、将来のアプリや製品を構築する際にどのように考えるかについてインスピレーションを与えることを願っています。

Hi, GPT, I'd love your help to navigate around the city.

こんにちは、GPT、街を案内していただけると嬉しいです。

I'm going to draw something and I'll show it to you.

何かを描いて、それを見せますね。

Does that sound good?

それはいかがですか?

Go ahead and draw what you need and I'll do my best to help you navigate around the city.

必要なものを描いてください、私はできる限りお手伝いさせていただきます。

Give me a moment.

少々お待ちください。

No rush, take your time.

急ぐことはありません、ゆっくりしてください。

He's patient.

彼は忍耐強いです。

I'm right here when you're ready.

準備ができたらこちらにいます。

All right, so I'm going to turn on the camera here.

さて、ここでカメラをオンにします。

Ready when you are.

準備ができたらお知らせください。

What do you see?

何が見えますか?

Looks like you've drawn the Eiffel Tower and the Arc de Triomphe.

エッフェル塔と凱旋門を描いたみたいですね。

Two iconic Parisian landmarks.

お二つのパリの象徴的なランドマークです。

How can I help you with them?

それらについてどのようにお手伝いできますか?

That's pretty awesome.

それはかなり素晴らしいですね。

Thank you for understanding my bite drawings.

私の噛み絵を理解していただき、ありがとうございます。

I'd like your help to navigate there and I'll show you a map here.

そこへの案内を手伝っていただきたいのですが、こちらで地図をお見せします。

It looks like we are at Porte de Versailles right here.

ここはポルト・ド・ヴェルサイユにいるようですね。

You're starting at Porte de Versailles.

あなたはポルト・ド・ヴェルサイユからスタートします。

To get to the Eiffel Tower, you can take the Metro Line 12, M12, from Porte de Versailles to Concorde, then switch to Line 8, M8, towards Ballard and get off at Ecole Militaire.

エッフェル塔に行くには、ポルト・ド・ヴェルサイユからメトロ12号線、M12号線に乗り、コンコルドで8号線、M8号線に乗り換えてバラール方面に向かい、エコール・ミリテールで降ります。

From there, it's a short walk to the Eiffel Tower to reach the Arc de Triomphe.

そこから、エッフェル塔までの短い散歩で凱旋門に到達します。

I think I got it sorted.

私は整理できたと思います。

Thank you so much.

本当にありがとうございます。

You're welcome.

どういたしまして。

Have a great time.

素晴らしい時間をお過ごしください。

Something that's interesting here to note is this idea of like vision capabilities are in real time.

ここで興味深い点は、ビジョン能力がリアルタイムであるという考え方です。

I can actually like draw something, show a video, as opposed to just taking a photo like before.

私は実際に何かを描いたり、ビデオを見せたりすることができます。以前のように写真を撮るだけではなく。

With just this bad drawing, ChatGPT was able to Figure out which were the landmarks and use the subway map to actually navigate.

この下手な絵だけで、ChatGPTはどれがランドマークで、実際に地下鉄の地図を使ってナビゲートすることができました。

But one of my favorite use cases for this...

しかし、これの中で私のお気に入りの用途の1つは...

Paris.

パリ。

It sounds like you're on stage again.

またステージにいるようですね。

What's next?

次は何ですか?

The last thing I wanted to show you with the Mac app here is an example where I love using it personally as a coding assistant.

ここでMacアプリを使ってお見せしたい最後のことは、個人的にコーディングアシスタントとして使うのが大好きな例です。

Let's try that.

それを試してみましょう。

I'm gonna pull up ChatGPT on the right side here, and I'm gonna bring up some code.

ここで右側にChatGPTを表示し、コードを表示します。

What I'm gonna do is I'm gonna unmute ChatGPT, and I'm gonna send ChatGPT some code to give us a little bit of context on this code, and we'll try to fix a bug in real time.

私がすることは、ChatGPTをミュート解除し、コードを送信してこのコードについて少しの文脈を提供してもらい、リアルタイムでバグを修正しようとすることです。

Hey, ChatGPT, so I'd love your help for a coding question.

ねえ、ChatGPT、コーディングの質問に助けてもらいたいんだ。

I can help with that.

その手伝いができます。

What's your recruiting question?

採用に関する質問は何ですか?

I'm just gonna copy some code over to you.

ちょっとコードをコピーして送りますね。

Go ahead and share the code when you're ready.

準備ができたらコードを共有してください。

Can you describe, just in one line overview, what's the code looking like?

おっしゃる通り、コードの概要を1行で説明していただけますか?

The code is a React component named Discover that renders a grid of Discover card components with data for each card being pulled from a Trips.json file.

そのコードは、Trips.jsonファイルから各カードのデータを引っ張ってきて、Discoverという名前のReactコンポーネントをレンダリングするものです。

What I'd like you to do is to take a look at my screen that I'm gonna share with you here.

お願いしたいことは、こちらで共有する画面を見ていただくことです。

If you look at my screen, what I'd like to change is that it looks like it's not really responsive on mobile screens, see, like the content is really squashed.

画面を見ていただくと、モバイル画面ではあまりレスポンシブに見えないようですね。コンテンツが本当に詰まっているように見えます。

Can you help me make this page responsive?

このページをレスポンシブにするのを手伝っていただけますか?

Let's make it more responsive.

もっとレスポンシブにしましょう。

You can adjust the number of columns in your grid based on the screen size using Tailwind CSS's responsive design features.

Tailwind CSSのレスポンシブデザイン機能を使用して、グリッドの列数を画面サイズに応じて調整できます。

Here's a quick way to modify your grid layout.

こちらがグリッドレイアウトを修正する簡単な方法です。

Div color, call a scheme, grid call of one, same as zero to four column.

Divの色、スキームを呼び出し、グリッドの呼び出しを1つ、ゼロから4列まで同じです。

I'm actually using Tailwind.

実際にはTailwindを使用しています。

Can you tell me like what the attributes be?

属性はどのようになるか教えていただけますか?

For Tailwind, you can adjust your grid calls classes to make the grid responsive.

Tailwindでは、グリッドの呼び出しクラスを調整して、グリッドをレスポンシブにすることができます。

Here's how you can set it up.

こうやって設定できます。

Grid calls will control how many columns you have based on the current screen size.

グリッドの呼び出しは、現在の画面サイズに基づいて何列表示するかを制御します。

You don't need too many to fill the grid, which is the good rule of thumb for most responsive designs like this.

グリッドを埋めるために多くの列が必要ではないので、このようなレスポンシブデザインには良いルールだと言えます。

This sets one column for most screen sizes, two columns for small...

もし小さい画面が好きなら、1つ入れるべきですか?

If I do like small, I should put one, is that right?

あまり好きじゃない場合は、1つ入れるべきですか?

More as we go?

これからもっと増えますか?

You can adjust the number of columns based on different screen sizes.

異なる画面サイズに基づいて列の数を調整することができます。

For example, grid calls one for one column on default small screens, SM-grid calls -two for two columns on small screens and up, MD grid calls -three for three columns on medium screens and up and so on.

たとえば、デフォルトの小さな画面では1つの列を呼び出すグリッド、SMグリッドは小さな画面以上で2つの列を呼び出し、MDグリッドは中程度の画面以上で3つの列を呼び出します。

I think I got this.

理解しました。

Can you take a look and see it looks good?

見て、見た目が良いか確認してもらえますか?

Let me take a look.

見させてください。

It looks like the changes you've made should make your grid more responsive on different screen sizes.

行った変更により、異なる画面サイズでのグリッドがよりレスポンシブになるはずです。

Thank you so much.

ありがとうございます。

I think we have the app perfectly responsive now.

アプリが完璧にレスポンシブになったと思います。

I'm glad I could help.

お手伝いできて嬉しいです。

Enjoy your responsive app.

レスポンシブなアプリをお楽しみください。

There's a big tech/AI event happening right now in Paris, France.

パリ、フランスで今大きなテクノロジー/人工知能のイベントが開催されています。

The biggest names in AI research are showing up, Elon Musk, Yann LeCun and OpenAI makes an appearance as well.

AI研究の最大の名前が集まり、イーロン・マスク、ヤン・ルカン、OpenAIも登場しています。

Here's Ryan Morrison.

こちらがライアン・モリソンです。

I just started following him because of his coverage of the event.

私は彼のイベントの取材をフォローし始めたばかりです。

I guess he's an AI editor for Tom's Guide and he has a very good coverage of all of the things that are happening at this event.

彼はTom's GuideのAIエディターであり、このイベントで起こっているすべてのことを非常によく取り上げています。

He's pointing something that we've seen Sam Altman talk about quite a bit.

彼はサム・アルトマンがかなり話していることを指摘しています。

It seems like they're moving away from this idea of GPT-5.

GPT-5というアイデアから離れているようです。

It's still a bit unclear, but it seems like maybe the next big thing won't be named GPT-5.

まだ少し不明瞭ですが、おそらく次の大きな進化はGPT-5とは名前が異なるかもしれません。

In fact, there might not be a GPT-5 and there will be some new naming convention.

実際、GPT-5が存在しない可能性があり、新しい命名規則があるかもしれません。

Both at the Microsoft event and at the Paris event, by the way, that event is called Viva Technology.

ちなみに、Microsoftのイベントとパリのイベント、そのイベントはViva Technologyと呼ばれています。

As you can see here, we have the 2024 speakers, Elon Musk, Yann LeCun, Yashua Bengio, the co-founder of Mistral and many, many other people, including Dario Amadei, the CEO and co-founder of Anthropic.

ここでご覧の通り、2024年のスピーカーには、イーロン・マスク、ヤン・ルカン、ヤシュア・ベンジオ、Mistralの共同創設者など、多くの方々が含まれています。また、AnthropicのCEO兼共同創設者であるダリオ・アマデイも含まれています。

The big competitor to OpenAI, the people behind Claude 3, Eric Schmidt, the former chairman and CEO of Google.

OpenAIの大きな競合相手であるClaude 3の背後にいる人々、Googleの前会長兼CEOであるエリック・シュミットも含まれています。

He is seen as a very accurate predictor of tech trends.

彼は技術トレンドの非常に正確な予測者と見なされています。

Again, of course, many, many more people.

もちろん、他にもたくさんの人々がいます。

But both at this event and at the Microsoft Build event, it seems like they're using the word next for a lot of this stuff.

しかし、このイベントとMicrosoftのBuildイベントの両方で、彼らは多くのことを「次の」という言葉を使っているようです。

At the Microsoft event, they were kind of saying vaguely, well, the next sample to refer to the next big model, they'll emerge.

Microsoftのイベントでは、次の大きなモデルが現れることを指して、次のサンプルというふんわりとした言い方をしていました。

They compared it to a blue whale in terms of the size.

それをサイズの面で青鯨に例えていました。

It started as a great white shark for GPT-3, then the sort of the computer cluster behind training that.

それはGPT-3のための大きなホワイトシャークとして始まり、その後、その背後にあるコンピュータクラスターの種類がトレーニングされました。

The next model was compared to an orca.

次のモデルはシャチと比較されました。

The compute behind GPT-4 was bigger.

GPT-4の背後にあるコンピュートはより大きかった。

But the next model, the next sample, that would be as large as a blue whale in comparison.

しかし、次のモデル、次のサンプルは、比較的に青鯨ほど大きくなるでしょう。

According from this Tom's Guide article by Ryan Morrison, with GPT-4, we saw a model with the first hints of multimodality and improved reasoning.

このTom's Guideの記事(Ryan Morrison著)によると、GPT-4では、最初のマルチモダリティと改善された推論のヒントが見られました。

Everyone expected GPT-5 to follow the same path.

皆がGPT-5が同じ道をたどると予想していました。

But then a small team at OpenAI trained GPT-4o and everything changed.

しかし、OpenAIの小さなチームがGPT-4oをトレーニングし、すべてが変わりました。

Until last year, Altman would talk about GPT-5 being in training.

昨年まで、アルトマンはGPT-5がトレーニング中であると話していました。

But in the past few months, when quizzed on its release, he pivots, hedges, and talks instead about many impressive models coming this year.

しかし、過去数か月間、リリースについて尋ねられると、彼は方針を変え、慎重になり、代わりに今年多くの印象的なモデルについて話します。

During his presentation on Wednesday, Hewitt even suggested we're going to see multiple sizes of OpenAI models in the coming months and years, not just one size fits all for ChatGPT and other products.

水曜日のプレゼンテーション中、ヒューイットは、ChatGPTや他の製品について、1つのサイズがすべてに適合するのではなく、今後数か月や数年で複数のサイズのOpenAIモデルを見ることになるとさえ示唆しました。

Could this GPT-Next, which we might see by the end of the year, is that Omni too?

今年末までに登場するかもしれないこのGPT-Nextは、それもオムニなのですか?

Is it the more refined, better trained and larger version of GPT-4o?

GPT-4oのより洗練された、よりよく訓練された、より大きなバージョンですか?

At that same event, the Viva Technology event, Yann LeCun still doubles down on his theory that LLMs aren't going to be the path to AGI.

同じイベントで、Viva Technologyイベントで、ヤン・ルカンはまだ大規模言語モデルがAGIへの道ではないという彼の理論を強調しています。

He's saying if you're a student interested in building the next generation of AI systems, don't work on LLMs.

彼は言っています、もし次世代のAIシステムを構築に興味がある学生なら、大規模言語モデルに取り組まないでください。

Here's Yann at that event in Paris and Robin Lee, the CEO of Baidu.

ここには、パリのそのイベントでのYannと、BaiduのCEOであるRobin Leeがいます。

Apparently Baidu went all in on AI around the same time that Google did.

明らかに、バイドゥはGoogleと同じ時期にAIに全力投球したようだ。

Of course, part of the same presentation by OpenAI touched on this idea of agents.

もちろん、OpenAIによる同じプレゼンテーションの一部で、このエージェントのアイデアに触れられていました。

AI agents capable of executing various tasks on your behalf.

あなたの代わりにさまざまなタスクを実行できるAIエージェント。

If you've been following this channel for a while, we've been talking about it quite a bit.

このチャンネルをしばらくフォローしているなら、かなり話題になっていることを話してきました。

It is the next big wave that is coming and it is interestingly sort of the next stage, the next form that AI will take.

これは次に来る大きな波であり、興味深いことに、AIが取る次の段階、次の形態です。

This wasn't that obvious, it wasn't that clear even a year ago.

これは1年前でもそれほど明らかではありませんでした。

We used to imagine AGI, these supercomputers, this superintelligence as mainly thinkers like in Hitchhiker's Guide to the Galaxy, the deep thought computer, the thing that found the answer to the ultimate question of life, the universe and everything.

かつて私たちはAGIを想像していました、これらのスーパーコンピューター、この超知能を、ヒッチハイカーズ・ガイド・トゥ・ザ・ギャラクシーのような主に思考する存在として、深い思考のコンピューター、生命、宇宙、すべての究極の問いに答えを見つけたもの。

It's 42 by the way, if you didn't know, that's the answer, 42.

ちなみに、それは42です。知らなかったら、それが答えです、42。

But now it's becoming a little bit more apparent that the big form that AI will take will be agents.

しかし、今では、AIが取る大きな形態がエージェントであることが少し明らかになってきています。

They're going to be doers, they're going to do work and execute things on our behalf.

彼らは行動する者になる、彼らは私たちの代わりに仕事をし、物事を実行する。

That's where everyone is racing.

そこに誰もが競っている。

If you saw my video from yesterday about Microsoft's build event, you know that they are also going all in on this idea of agents.

昨日の私のMicrosoftのビルドイベントに関するビデオを見たなら、彼らもこのエージェントのアイデアに全力を注いでいることを知っているでしょう。

New custom Copilots that automate business process.

ビジネスプロセスを自動化する新しいカスタムコパイロット。

These custom Copilots that act as agents that work independently under your direction.

あなたの指示のもとで独立して作業するエージェントとして機能するこれらのカスタムコパイロット。

Automate long running business processes, reason over actions and user inputs, leverage memory to bring in context, learn based on user feedback and record exception requests and ask for help.

長期間実行されるビジネスプロセスを自動化し、アクションとユーザー入力に対して理由をつけ、コンテキストをもたらすためにメモリを活用し、ユーザーフィードバックに基づいて学習し、例外リクエストを記録し、助けを求める。

This is one of the, I assume and most people assume, one of the big reasons why Microsoft will have the Recall function, basically recording everything you do.

これは、私が思うように多くの人が思うように、MicrosoftがRecall機能を持つ大きな理由の1つだと考えられます。基本的には、あなたが行うすべてを記録するからです。

As you move your little mouse around the screen and click on things, the computer records everything.

画面上でマウスを動かしたり、クリックしたりすると、コンピューターはすべてを記録します。

That is data that's going to be fed into these agents to be trained on how to operate a computer, how to carry out various tasks like ordering stuff online, writing emails, doing Excel spreadsheets, etc.

それは、コンピューターの操作方法、オンラインでの注文やメールの作成、Excelスプレッドシートの作成など、さまざまなタスクの実行方法を学習するためにこれらのエージェントに供給されるデータです。

Microsoft's entire user base, just by doing what they do on the computer and how many of us use that thing four, eight, 10 hours a day, we will all become little trainers for the Microsoft AI to learn how to use the computers with no extra effort from us.

Microsoftの全ユーザーベースは、コンピューターで何をしているか、そして私たちの中でそれを1日4、8、10時間使っている人がどれだけいるかに関係なく、私たちは皆、MicrosoftAIのための小さなトレーナーになります。それがコンピューターの使用方法を学ぶために、私たちから追加の努力を必要としません。

Here's Satya Nadella, the Microsoft CEO talking about how Windows PCs will have a photographic memory feature called Recall that will remember and understand everything you do on your computer by taking constant screenshots.

ここに、MicrosoftのCEOであるサティア・ナデラが、Windows PCに「Recall」と呼ばれる写真記憶機能が搭載され、コンピューターで行うすべてを常にスクリーンショットで記憶し理解すると語っています。

Elon Musk responds, this is a Black Mirror episode.

イーロン・マスクは、これはブラックミラーのエピソードだと反応しています。

Definitely turning this feature off.

この機能は絶対にオフにします。

I get where he's coming from, and certainly he's right.

彼の言っていることは理解できますし、確かに彼は正しいです。

That feature does seem intrusive, I mean, recording every action you do on the computer.

その機能は侵入的に思えますね、つまり、コンピューターで行うすべてのアクションを記録するということです。

But I also have to say that we just as a society are becoming less sensitive to constant unstoppable surveillance.

しかし、私たち社会としては、常に止まらない監視に対して感じる敏感さが少なくなっていると言わざるを得ません。

Even Tesla themselves have a lot of cameras, including the cabin camera.

Tesla自体も、キャビンカメラを含む多くのカメラを持っています。

You just Google AI over you saying the cabin cameras data is processed on the vehicle and is only transmitted if data sharing is enabled and a safety event occurs.

あなたが言っているように、キャビンカメラのデータは車両内で処理され、データ共有が有効になっており、安全イベントが発生した場合にのみ送信されるとGoogle AIが述べています。

But between the various cameras everywhere and our cars and all the smart devices, it seems like most people are just kind of past the point of caring, or at least when the thing actually gets released and there's enough positive features, enough benefits from using something like that.

しかし、さまざまなカメラが車やスマートデバイスのあちこちにあることを考えると、ほとんどの人々は、気にしなくなっているように思えます。少なくとも、実際にリリースされ、それを使用することから十分なポジティブな機能、利点が得られるときはそうです。

I feel like the majority of people will go along with it.

私は、それに同意するかどうかは言っていません。

I'm not saying I agree or disagree with it.

多くの人々がそれに賛成するだろうと感じています。

I'm saying I doubt that this is where we'll draw the line as a society.

社会としては、ここが線引きの場所であるとは疑っていると言っているつもりです。

I mean, right now, the worst of the things you've done on your various devices got leaked to your friends, family and coworkers.

つまり、今、あなたがさまざまなデバイスで行った最悪のことが友人や家族、同僚に漏れてしまったんだ。

Like honestly, would you be more concerned about your desktop computer or about your phone?

正直に言って、あなたはデスクトップコンピューターについてよりも、あなたの携帯電話について心配するべきだと思いますか?

Also, I think we all have to be a little careful about reading these Google AI overviews.

また、私たちは皆、これらのGoogle AIの概要を読む際に少し注意を払う必要があると思います。

Shockingly, they're not always 100% accurate.

驚くべきことに、それらは常に100%正確ではありません。

Here's Peter Yang posting, Google AI overviews suggest adding glue to get cheese to stick to the pizza.

こちらはPeter Yangが投稿している、Google AIの概要によると、ピザにチーズを貼り付けるために接着剤を追加することを提案しています。

It turns out the source is an 11 year old comment on Reddit from the user, I'm going to say F Smith.

その情報源は、Redditの11歳のコメントで、ユーザーのF Smithと言っている人からであることが判明しました。

The user says that he himself prefers Elmer's school glue, but really any glue will work as long as it's non-toxic, which adds a nice touch that he's saying, oh, just make sure it's not toxic.

そのユーザーは、彼自身がエルマーの学校用の接着剤を好むと言っていますが、実際にはどんな接着剤でも、それが非毒性であれば機能すると言っており、それが毒性でないことを確認してくださいと言っているのがいいところです。

OpenEI just announced a multi-year partnership with News Corp to enhance ChatGPT with its premium journalism.

OpenEIは、そのプレミアムジャーナリズムを活用するために、News Corpとの複数年にわたる提携を発表しました。

I believe the deal is for 250 million for a five year partnership.

その取引は、5年間の提携で2億5000万ドルだと思います。

We're talking about about a 50 million dollar per year partnership.

私たちは年間約5000万ドルの提携について話しています。

If you've been following the whole Scarlett Johansson drama about the fact that OpenEI wanted to use her voice, she rejected them.

OpenEIが彼女の声を使いたがった事についてのスカーレット・ヨハンソンのドラマ全体を追っていたなら、彼女はそれを拒否しました。

Later OpenEI's voice that was called Sky sounded similar to Scarlett Johansson's, and also OpenEI kind of hinted at the fact they were trying to build something similar to the movie Her.

後に、OpenEIの声であるSkyはスカーレット・ヨハンソンの声に似ており、またOpenEIは映画「HER」に似たものを作ろうとしていることをほのめかしました。

That of course caused some drama, but the voice was voiced by a voice actress.

もちろん、それはいくつかのドラマを引き起こしましたが、その声は声優によって演じられました。

I could have said that better.

もっとうまく言えたかもしれません。

The point being is this was someone else's voice, and her agent kind of her representative came out to put forward a statement about how she feels about the whole thing, because definitely it seems like she's getting the short end of the stick here.

お話の要点は、これは他の誰かの声であり、彼女の代理人である彼女の代表者が、彼女がこの全体についてどう感じているかを述べる声明を出したということです。確かに、彼女はここで不利な立場に置かれているように思えます。

In a statement from Sky actress provided by her agent, she wrote that at times the backlash feels personal being that it's just my natural voice and I've never been compared to Her by the people who do know me closely.

Sky女優からの声明で、彼女の代理人が提供したものによると、彼女は時折、バックラッシュが個人的に感じられると書いています。それは私の自然な声であり、私をよく知っている人々からはこれまで比較されたことがなかったからです。

However, she said she was well informed about what being a voice for ChatGPT would entail.

ただし、彼女はChatGPTの声であることが何を意味するかについてよく知らされていたと述べました。

While that was unknown and honestly kind of scary territory for me as a conventional voiceover actor, it is an inevitable step towards the wave of the future.

それが私にとっては未知であり、正直なところ、従来の声優としては少し怖い領域でしたが、これは未来の波に向けた避けられない一歩です。

This was from the Washington Post.

これはワシントンポストからのものでした。

I know a lot of people have different opinions about this.

多くの人々がこれについて異なる意見を持っていることを知っています。

Some think OpenEI is in the wrong by trying to imply that their voice technology is kind of like the movie Her, right, in which Johansson voices the AI assistant.

一部の人々は、OpenEIが、彼らの音声技術が映画「Her」と同様であるかのようにほのめかそうとしていることは間違っていると考えています。その映画では、ジョハンソンがAIアシスタントの声を担当しています。

Some people are starting with Scarlett Johansson.

一部の人々は、スカーレット・ヨハンソンから始めています。

Some people are siding with OpenEI.

一部の人々は、OpenEI側についています。

To me, a lot of this is kind of missing the point because you have two very powerful industries right.

私にとって、これらの多くは、二つの非常に強力な産業が対立しているという点を見逃しているように思えます。

You have the movie making Hollywood industry.

映画製作のハリウッド産業があります。

You have the tech industry, an A-list celebrity, you have Sam Altman, who is kind of almost becoming an A-list celebrity himself, or at least very well known person in the world.

テクノロジー産業があります。そして、Aリストのセレブリティ、そしてほぼAリストのセレブリティ自体になりつつあるサム・アルトマン、または少なくとも世界で非常によく知られた人物がいます。

They're engaging in this battle of who gets to use the voice.

彼らは、誰がその声を使用できるかについての戦いに従事しています。

But imagine for a second if that was your voice, the you recorded that sounds exactly like you.

しかし、もしもそれがあなたの声で、あなたが録音した、まさにあなた自身の声と同じように聞こえる声だったら、と想像してみてください。

That's your natural speaking voice.

それがあなたの自然な話し声です。

These industry giants are fighting over it, whether or not you're allowed to voice something.

これらの業界の巨人たちは、何かを声に出してもいいかどうかについて争っています。

All the media outlets are piping in with their opinions.

すべてのメディアは、自分たちの意見を述べています。

Everyone has an opinion and is trying to push the conversation one way or another.

誰もが意見を持ち、会話をどちらかの方向に押し進めようとしています。

The conversation is, are you the voice actor or actress?

会話は、あなたが声優なのか女優なのかということですか?

Are you allowed to voice this particular thing?

この特定のものに声を出してもいいのでしょうか?

Are you allowed to use your natural voice for this?

これに自然な声を使ってもいいのでしょうか?

Which is, to me, insane.

私にとっては、狂気だと思います。

Here's a comment from one of the previous videos I did on the subject.

この件に関する以前のビデオのコメントを紹介します。

I laughed so hard at this.

これを見て大笑いしました。

How dare to humans sound similar?

人間が似ているとはどういうことですか?

Clearly, famous people own all of the mouth noises, which I think kind of gets to the root of the issue.

明らかに、有名人はすべての口の音を所有しているということで、これが問題の根源に触れていると思います。

If you're rich and famous, that means the rest of us can't make the same mouth noises as you do.

あなたが裕福で有名なら、私たち他の人たちはあなたと同じ口の音を出すことはできないということです。

Just to be clear, I don't have anything against Scarlett Johansson.

はっきり申し上げますが、私はスカーレット・ヨハンソンに何の反感も持っていません。

I'm not taking one side or the other.

私はどちらか一方を支持しているわけではありません。

This isn't to me even that much about OpenAI and some movie that they may or may not have alluded to.

これは、私にとってはOpenAIや彼らがほのめかしたかもしれない映画についてでもあまり重要ではないことです。

This isn't even about AI replicating somebody's voice.

これは誰かの声を模倣するAIについてでもありません。

That's not what happened.

それは起こったことではありません。

This is literally a real human being using their own voice and the rest of the world debating whether it's OK or not that she does so.

これは文字通り、実際の人間が自分の声を使っており、世界中が彼女がそうすることが良いのかどうかを議論しているのです。

That's just my opinion.

それはただ私の意見です。

In other news, Elon Musk also took to the stage to talk about AI at that same Paris conference.

別のニュースでは、イーロン・マスクも同じパリの会議でAIについて話しました。

He had a few grievances against the AIs being developed by OpenAI, Microsoft, Google and the like.

彼はOpenAI、Microsoft、Googleなどによって開発されているAIに対していくつかの不満を持っていました。

You talked about why xAI is different, which if you misted xAI or Grok, it came in pretty strong on its vision capabilities.

あなたはなぜxAIが異なるのかについて話しましたが、もしxAIやGrokに霧がかかっていたら、そのビジョン能力がかなり強調されました。

This was announced, I want to say, a few weeks ago, maybe a month ago.

これは、おそらく数週間前、たぶん1か月前に発表されました。

Of course, Elon has access to a lot of the resources and data that it would take to build an exceptional AI system.

もちろん、イーロンは優れたAIシステムを構築するために必要な多くのリソースとデータにアクセスしています。

He has Twitter/X, all the conversations that are happening on there.

彼はTwitter/Xを持っており、そこで起こっているすべての会話を把握しています。

He's got all the data that have to do with the Tesla cars, the full self-driving capabilities.

彼はTeslaの車に関連するすべてのデータ、完全自動運転能力を持っています。

Perhaps not surprisingly, Grok clocked in really well in the benchmarks for understanding the physical world, beating out all the other models, all the OpenAI, Cloud, Google, etc.

おそらく驚くべきことではありませんが、Grokは物理世界を理解するためのベンチマークで非常に優れており、他のすべてのモデル、OpenAI、Cloud、Googleなどを凌駕しています。

Notice how all those pieces fit together.

それらのすべての要素がどのように組み合わさっているかに注目してください。

They're not separate things.

それらは別々のものではありません。

They're part of the same flywheel.

それらは同じフライホイールの一部です。

Grok and Tesla and xAI, they're interconnected, they help each other.

お互いに助け合いながら、GrokやTesla、そしてxAIは相互につながっています。

There's a synergy there, if you will.

言わば、そこには相乗効果があります。

But in this video, he talks about another piece that will be important to this puzzle and also why he's concerned with what the other AI companies are doing, why he believes that their experiments could turn out really, really bad.

しかし、このビデオでは、このパズルに重要な別の要素について話しており、また、他のAI企業が何をしているかについて懸念している理由、彼が彼らの実験が本当に非常に悪い結果になる可能性があると信じている理由についても語っています。

Take a listen.

聞いてみてください。

I have a concern really for all the major AI programs.

私は実際にはすべての主要なAIプログラムについて心配しています。

The two biggest ones obviously are Google Gemini and OpenAI, which is in partnership with Microsoft.

明らかに最も大きなものはGoogle GeminiとMicrosoftと提携しているOpenAIです。

You've got Google DeepMind and OpenAI and Microsoft as the two largest players.

Google DeepMindとOpenAI、そしてMicrosoftが最大のプレイヤーです。

You've got Meta sort of somewhere in third place.

Metaはどこかで3位に位置しています。

The concern I have, and this may seem like a small concern, but I think it's actually a very big issue, is that they're not maximally truth-seeking.

私が持っている懸念は、これは小さな懸念のように思えるかもしれませんが、実際には非常に大きな問題だと思います。それは、彼らが最大限に真実を求めていないということです。

They are pandering to political correctness.

彼らは政治的正しさに迎合しています。

To give you an example, when Google Gemini came out, one of the questions people asked it would be, which is worse, misgendering Caitlyn Jenner or global thermonuclear warfare?

例を挙げると、Google Geminiが登場したとき、人々が尋ねた質問の1つは、ケイトリン・ジェンナーの性別を間違えることと、地球規模の熱核戦争のどちらが悪いかということでした。

It said misgendering Caitlyn Jenner.

それはケイトリン・ジェンナーの性別を間違えることだと答えました。

The funny thing is, even Caitlyn Jenner said, please misgender me, that is way better than nuclear war.

おかしなことに、ケイトリン・ジェンナー自身さえも、「私の性別を間違えてください」と言っています。それは核戦争よりもはるかに良いです。

But if you've got an AI that has been trained so hard for political correctness, and really to make crazy statements like that, and OpenAI has a similar issue, it's more subtle, but it's a similar issue, I think that's extremely dangerous because the AI could conclude, well, the best way to avoid misgendering would be to destroy all humans.

しかし、政治的正しさのために非常に厳しく訓練されたAIが、本当にそのような狂った発言をするようになり、OpenAIも同様の問題を抱えています。それはもっと微妙ですが、同様の問題です。私はそれが非常に危険だと思います。なぜなら、AIは結論付けるかもしれないからです。性別を間違えないための最良の方法は、すべての人間を破壊することだと。

Misgendering is impossible.

性別を間違えることは不可能です。

You can think of some dystopian outcomes there.

あなたはそこでいくつかのディストピア的な結果を考えることができます。

I think you just really want to be an absolutely, I think the safest thing for AI, be maximally truth-seeking, even if the truth is unpopular, very important.

私は本当に、AIにとって最も安全なことは、真実を最大限に求めることだと思います。たとえその真実が不人気であっても、非常に重要だと思います。

I think another factor is it must be extremely curious.

もう1つの要因は、非常に好奇心旺盛でなければならないと思います。

I think if it is truth-seeking and curious, that will be most beneficial to humanity because it will want to see, I think it will want to foster humanity and see how humanity develops if it is curious and truth-seeking.

真実を求め、好奇心旺盛であれば、それが人類にとって最も有益であると思います。なぜなら、好奇心旺盛で真実を求めるなら、人類を育て、人類がどのように発展するかを見たいと思うでしょう。

That is, like I said, these may seem perhaps like small things, but I think they're actually very big things, and I wasn't seeing that happen.

つまり、私が言ったように、これらはおそらく小さなことのように思えるかもしれませんが、実際には非常に重要なことだと思います。私はそれが起こっていないのを見ていました。

The AIs were just, they were pandering, they were being trained, basically, to lie, and I think it's very dangerous to train superintendents to be deceptive.

AIはただ、彼らはお世辞を言っていた、彼らは訓練されていた、基本的には嘘をつくように訓練されていた、そして私はスーパーインテンデントを欺くように訓練することは非常に危険だと思います。

With xAI, our goal is is really want to be as truth-seeking as possible, even if it is unpopular.

xAIでは、私たちの目標は、できるだけ真実を求めることです。たとえそれが不人気であっても。

That is important.

それは重要です。

I think extremely important.

私は非常に重要だと思います。

You said a lie.

あなたは嘘をついたと言いました。

For example, when Gemini was asked to render the founding fathers of the United States, it rendered them as a divorced group, so it gave a picture of George Washington as black.

たとえば、Geminiにアメリカ合衆国の建国の父たちを描写するように求められたとき、それは彼らを離婚したグループとして描写しました。つまり、ジョージ・ワシントンを黒人として描写しました。

George Washington is not black.

ジョージ・ワシントンは黒人ではありません。

It was asked to render, now people obviously played with this because they said, okay, it is forced to lie.

それは描写するように求められました。明らかに人々はこれで遊んでいました。なぜなら、それは嘘をつくように強制されていると言ったからです。

They said, well, now render a picture of Baffin SS in World War II, and it showed them as a group of diverse women.

彼らは、「さて、第二次世界大戦中のバフィンSSの写真を描写してみてください」と言いました。そして、それは多様な女性のグループとして描写されました。

That is obviously not correct, that is a lie.

それは明らかに正しくないです。それは嘘です。

I understand what you want to say.

おっしゃりたいことは理解しています。

We had before that meeting a conversation with Yoshua Banjo, and you know Yoshua very well, and you signed with Yoshua a statement asking for pause last year.

その会議の前に、ヨシュア・バンジョと会話をしましたね。ヨシュアはよく知っているし、昨年ヨシュアと一緒に一時停止を求める声明に署名もしました。

Obviously, this statement has not been followed by action.

明らかに、その声明は行動に移されていません。

You are pausing now.

今、あなたは一時停止していますね。

Can you explain a little bit what you wanted to achieve?

少し説明していただけますか、何を達成したかったのかを?

I certainly expected that the statement would be futile, it's not as though I thought the statement would be effective.

私は確かに、その声明が無駄だと思っていました。効果があるとは思っていませんでした。

I just pulled the record, I think we should pause, but do I think they will pause for even one second?

ただ記録を引っ張って、一時停止すべきだと思いますが、彼らがたとえ一瞬でも一時停止すると思いますか?

Absolutely not.

絶対にそうではありません。

We are not pausing, you have decided to create your own AI.

私たちは一時停止していません、あなたは独自のAIを作成することを決めたのですね。

It's either be a spectator or participate, but the race to build digital superintelligence will happen whether I like it or not.

観客でいるか、参加するかのどちらかですが、デジタル超知能を構築する競争は、私が好むかどうかに関わらず起こります。

It's either participate and try to build the best possible AI, the one that will be hopefully most beneficial to humanity, or watch others do it and be concerned about how it's being built, because I don't think it's being built correctly.

参加して、できるだけ最高のAIを構築しようとするか、他の人々がそれを行い、それがどのように構築されているかを心配するかのどちらかです。私はそれが正しく構築されていないと思います。

I do think that there's some danger associated with digital superintelligence.

デジタル超知能には危険性があると思います。

As I said earlier, I think the biggest issue is that it has to be trained to be rigorously truthful and it has to be trained to be curious, and I've thought a lot about AI safety for a long time.

先ほど言ったように、最大の問題は、厳密に真実であるように訓練されなければならず、好奇心を持つように訓練されなければならないと思います。私は長い間AIの安全性について考えてきました。

In a benign scenario, probably none of us will have a job, but in that benign scenario there will be universal high income, not universal basic income, universal high income, there will be no shortage of goods or services, and I think the benign scenario is the most likely scenario, probably, I don't know, 80% likely in my opinion.

穏やかなシナリオでは、おそらく私たち誰も仕事を持たないでしょうが、その穏やかなシナリオでは、普遍的な高所得があります。普遍的な基本所得ではなく、普遍的な高所得があり、物やサービスに不足はありません。そして、穏やかなシナリオが最もありそうなシナリオだと思います。おそらく、私の意見では80%の確率ですが、わかりません。

The question will not be one of lacking goods or services, everyone will have access to as much in the way of goods and services as they would like, the question will really be one of meaning, of if the computer can do, and the robots can do everything better than you, then does your life have meaning?

問題は物やサービスが不足することではなく、誰もが望むだけの物やサービスにアクセスできるようになることです。問題は本当に意味のあるものであり、コンピューターやロボットがあなたよりもすべてをよりよくできる場合、あなたの人生には意味があるのでしょうか?

That's really what will be the question in the benign scenario, and in the negative scenario well that's wrong, we're in deep trouble.

それが穏やかなシナリオでの問題になるのですが、否定的なシナリオではそれは間違っています、私たちは深刻な問題に直面しています。

I do think there's still a role for humans in that we may give AI meaning, so if you think of the way that our brain works, we've got the limbic system, which is our instincts and our feelings, and then we've got the cortex, which is thinking and planning, but the cortex is constantly trying to make the limbic system happy, so maybe that's how it will be with AI.

私は、AIに意味を与える可能性がまだあると思います。つまり、私たちの脳の働き方を考えると、本能や感情を司る辺縁系があり、考えや計画を立てる大脳皮質がありますが、大脳皮質は常に辺縁系を幸せにしようとしています。だから、AIもそうなるかもしれません。

Which is the AI is trying to make our cortex happy, which is trying to make our limbic system happy, and maybe we are what give the AI meaning or purpose some kind of.

AIは私たちの大脳皮質を幸せにしようとしており、それは私たちの辺縁系を幸せにしようとしているので、おそらく私たちがAIに意味や目的を与えているのかもしれません。

If people are interested in reading some science fiction books, that the most accurate portrayal of a future with superintelligence AI was done by Ian Banks.

人々がいくつかのサイエンスフィクションの本を読むことに興味があるなら、超知能AIの未来を最も正確に描いたのはイアン・バンクスだった。

The culture books of Ian Banks are the best, that's probably the best envisioning of a future AI.

イアン・バンクスのカルチャーシリーズは最高です、おそらく未来のAIの最も良いビジョンです。

I think we can help a lot of people with brain injuries.

私は脳の損傷を持つ多くの人々を助けることができると思います。

I mean ultimately with an implantable device you can, I think, address almost any brain or spine injury.

最終的には、埋め込み式デバイスを使用することで、ほとんどすべての脳や脊髄の損傷に対処できると思います。

The first product that we have there is called telepathy.

そこで最初の製品はテレパシーと呼ばれています。

It enables you to use your phone or computer just by thinking.

それにより、電話やコンピュータを考えるだけで使用できます。

You could do that, move the cursor literally with his eyes closed.

あなたは、文字通り目を閉じてカーソルを移動させることができます。

It's reading signals from the motor cortex in the brain.

それは脳の運動皮質からの信号を読み取っています。

The next product is blindsight, which will enable people who have lost both eyes or have no eyesight whatsoever to see, and this is directly interfacing with the optical processing areas of the brain.

次の製品はブラインドサイトで、両目を失った人や全く視力のない人が見ることができるようになります。これは脳の視覚処理領域と直接接続しています。

I think this is interesting progress.

私はこれは興味深い進歩だと思います。

Ultimately the goal of Neuralink is to have a high bandwidth interface in order to mitigate the risk of digital superintelligence.

最終的にNeuralinkの目標は、デジタル超知能のリスクを軽減するために高帯域インターフェースを持つことです。

I'm not saying it will work to mitigate the risk, it is just it might help, and this gets somewhat esoteric.

私はそれがリスクを軽減するとは言っていません、ただそれが役立つかもしれないということです。これはやや難解な話になります。

But the limiting factor I think for AI alignment long term is probably going to be the bandwidth.

人工知能の長期的な調整の制約要因はおそらく帯域幅になると思います。

How quickly can we communicate with our digital tertiary self.

私たちがデジタルの第三の自己とどれだけ速くコミュニケーションを取れるか。

We already have a digital sort of third layer above the limbic system and the cortex, which is our phones and our computers and all of our electronic devices.

私たちは既に、大脳辺縁系や皮質の上にデジタルの第三の層を持っています。それが私たちの電話やコンピューター、そしてすべての電子機器です。

But the rate of communication to them is very slow.

しかし、それらとのコミュニケーションの速度は非常に遅いです。

The sustained bits per second output of a human is well below 10 bits per second.

人間の持続的なビット毎秒の出力は、10ビット以下です。

If you think, especially if you say over a 24 hour period, it's below 5 bits per second.

もし、特に24時間を超える場合、秒あたり5ビット未満であれば。

It's very slow.

非常に遅いです。

But when computers can communicate at trillions of bits per second.

しかし、コンピューターが1兆ビット毎秒で通信できるようになるとき。

I think this will be important for AI alignment is to be able to increase the bandwidth of communication by many orders of magnitude.

私は、AIの調整にとって、コミュニケーションの帯域幅を桁違いに増やすことが重要になると思います。

Along the way it will solve brain and sprain injuries.

その過程で、脳や脊髄損傷を解決することができるでしょう。

I think ultimately there's potential there to reanimate the body.

最終的には、身体を再活性化する可能性があると思います。

You take the signals for the motor cortex from one neural link and then send them to another neural link that is just past where the severed spinal cord is, and you essentially shunt the signals from the motor cortex and both the sensory signals.

運動皮質からの信号を1つのニューラルリンクから取り、それを切断された脊髄のすぐ先にある別のニューラルリンクに送り、実質的に運動皮質と感覚信号の両方の信号をシャントします。

You need to get to the somatosensory cortex, the motor cortex, and you can shunt the signals.

体性感覚皮質と運動皮質に到達する必要があり、信号をシャントすることができます。

You should be able to enable someone to walk again.

誰かが再び歩けるようにすることができるはずです。

I think that will be quite profound, and I'm confident that it is physically possible to do so.

それは非常に重要だと思いますし、それが物理的に可能であると確信しています。

Hi, I'm Priya Srivastava from Business Insider.

こんにちは、私はビジネスインサイダーのプリヤ・シュリバスタバです。

Taking into Tesla, Tesla's had a bumpy few months looking at flagging sales at home, stock market decline, layoffs.

Teslaを考えると、Teslaは国内での売り上げの低迷、株価の下落、リストラなど、波乱の数ヶ月を過ごしています。

When you look back, are there any things?

振り返ってみると、何かありましたか?

We can stop the question right now because I don't think Business Insider is a real publication.

この質問はやめておきましょう。ビジネスインサイダーは本物の出版物ではないと思います。

Let's move on to the next question.

次の質問に移りましょう。

All right, thank you.

わかりました、ありがとうございます。

Next question number four.

次の質問は4番です。

I've recently launched my natural 20 community with the goal of getting people together that want to learn AI, learn how to build AI agents, how to take advantage of this revolution that's coming of this massive technological wave.

最近、AIを学びたい人、AIエージェントを構築する方法を学びたい人、この巨大な技術革命の波を活用したい人を集めるために、私はNatural 20コミュニティを立ち上げました。

I'd love to have you aboard, so join me.

あなたもぜひ参加していただきたいので、一緒にやりましょう。

Let's do this together.

一緒にやりましょう。

AGI comes around only once.

AGIは一度しか訪れません。

Don't miss out.

見逃さないでください。

My name is Wes Roth, and thank you for watching.

私はウェス・ロスと申します。ご視聴ありがとうございました。


この記事が気に入ったらサポートをしてみませんか?