見出し画像

【TheAIGRIDのAIニュース:GPT-7の野望、AI人員削減、膨大な投資とGoogleの新技術】英語解説を日本語で読む【2024年3月29日|@TheAIGRID】

この週、AI分野では目覚ましい進展がありました。OpenAIはGPTのカスタムモデル作成者に対して収益化プログラムを発表しました。特に、コーディングに長けた「Grimmel」というGPTが注目されています。また、動きをトラッキングし、リップシンクするHeyGenの「Avatar in Motion 1.0」や、AIによるビデオ生成ツール「Argril.ai」のデモが公開され、リアルな出力に対する議論が巻き起こっています。MITの研究者は、より高速で高品質な画像生成手法を発表しました。その一方で、AIによる仕事の置き換えや、AI技術の倫理的な問題に関する議論も活発化しています。AmazonはAnthropicに27億5000万ドルを投資し、感情を読み取るAI「Hume」が紹介されました。
公開日:2024年3月29日
※動画を再生してから読むのがオススメです。


There was actually a lot of information released this week in Artificial Intelligence, and trust me, I don't want to make this one a long one, but I feel like it's going to be just based On the amount of information that I've seen.

実際、今週は人工知能に関する多くの情報が公開されました。信じてください、私はこれを長くしたくはありませんが、私が見た情報の量に基づいて、これがそうなると感じています。

Without further ado, let's jump into why this week in Artificial Intelligence is so surprising.

さて、それでは、なぜ今週の人工知能が驚くべきものなのかについて話していきましょう。

One of the first pieces of news was in fact from OpenAI.

最初のニュースの1つは実際にOpenAIからでした。

They actually said, we're partnering with a small group of United States builders to test usage based GPT earnings.

彼らは実際に、米国の一部のビルダーと提携して、使用量に基づいたGPTの収益をテストすると述べました。

Our goal is to create a vibrant ecosystem where builders are awarded for their creativity and impact, and we look forward to collaborating with builders on the best approach to get there.

私たちの目標は、ビルダーが創造性と影響力に対して報酬を受け取る活気あるエコシステムを作り出すことであり、ビルダーと協力してそこに到達するための最良の方法について協力することを楽しみにしています。

It seems as if right now, if you are someone who owns a GPT, and if you don't know what that is, it's essentially when you go on OpenAI's website and you go on ChatGPT, you can have custom GPTS that you build and share with the community.

今、もしもあなたがGPTを所有している人で、それが何かわからない場合は、OpenAIのウェブサイトにアクセスしてChatGPTにアクセスすると、ビルダーが構築してコミュニティと共有できるカスタムGPTを持つことができます。

For those of you that have been building custom GPTS right now, they are rolling out the ability for you to earn from those GPTS.

現在、カスタムGPTを構築している方々には、それらから収益を得る機会が提供されています。

And we've seen this person, Nick Dobos.

そして、このニック・ドボスという人物を見てきました。

This guy is the creator of arguably the most popular custom GPT.

この人はおそらく最も人気のあるカスタムGPTの作成者です。

He created something called Grimmel, which is a custom GPT that is really good at coding.

彼は「Grimmel」というものを作成しましたが、これはコーディングに非常に優れたカスタムGPTです。

And you can see the email that he received today.

あなたは、彼が今日受け取ったメールを見ることができます。

It says, hi, today we announced a pilot program to partner with builders to test GPT earnings based on usage by paying builders based on the usage of their GPT.

それには、「こんにちは、本日、ビルダーと提携してGPTの使用に基づく収益をテストするパイロットプログラムを発表しました。ビルダーには、彼らのGPTの使用に基づいて支払われます。」と書いてあります。

This initiative is designed to explore and refine our approach to monetizing GPTS with a focus on rewarding the creativity and dedication of builders who have contributed to the system and Why you so like I said, it's only being rolled out to I'm guessing the top creators.

この取り組みは、GPTの収益化アプローチを探求し、磨きをかけることを目的としており、システムに貢献したビルダーの創造性と献身を報いることに焦点を当てています。なので、私が言ったように、おそらくトップクリエイターにのみ展開されていると書いてあります。

And it says your GPT Grimmel is highly utilized by ChatGPT and ChatGPT users.

そして、あなたのGPTグリムルは、ChatGPTとChatGPTユーザーによって非常に利用されています。

We'd love to incorporate your feedback as we define the approach to GPT monetization.

私たちは、GPTの収益化アプローチを定義する際に、あなたのフィードバックを取り入れたいと考えています。

Our goal is to foster a vibrant ecosystem where builders are rewarded for their creativity and impact and users have access to a wide range of highly useful GPTS.

私たちの目標は、ビルダーが創造性と影響力に対して報酬を受け取り、ユーザーが幅広い高度に有用なGPTSにアクセスできる活気あるエコシステムを育むことです。

We look forward to achieving this through co-ops collaboration with the builder community.

私たちは、ビルダーコミュニティとの協力協力を通じて、これを実現することを楽しみにしています。

And the next steps are saying, please let us know if you would be open to participating and we're going to be following up with additional details.

そして、次のステップでは、「参加していただけるかどうかお知らせください。追加の詳細については後日連絡いたします。」と書いてあります。

You'll need to sign certain documents and of course note that this program is open to us.

あなたは、特定の書類に署名する必要があり、もちろんこのプログラムは私たちにも開放されています。

It seems like these GPTS are going to be really, really good in terms of people who are trying to earn money because what they're doing is really good because it seems like right now They're rolling out some kind of alpha testing phase where people who are in this early alpha program are going to be determining how they earn money and they're going to be able to Provide feedback.

これらのGPTSは、おそらくお金を稼ごうとしている人々にとって本当に非常に良いものになるようです。なぜなら、彼らがやっていることは本当に良いことであり、現時点では、この早期アルファプログラムに参加している人々がどのようにお金を稼ぐかを決定し、フィードバックを提供できるようになるようです。

When it is rolled out to a wider user base, I'm sure people are going to be more incentivized to use them because I think GPTS are underutilized in their current moment.

広範囲のユーザーベースに展開されると、現在の状況ではGPTSが活用されていないと思うので、人々はそれらをより積極的に使用するようになると確信しています。

And I also think that potentially with GPT-5 increasing in capabilities, maybe we're going to be able to get custom versions of GPT-5 that reduce hallucinations, increase image Capabilities and with a voice, it might present a very, very lucrative opportunity for those of you who are looking to build.

GPT-5の能力が向上する可能性も考えられると思いますが、おそらくGPT-5のカスタムバージョンを手に入れることができるようになるかもしれません。これにより幻覚を減らし、画像の能力を高め、音声でも非常に有望な機会が提供されるかもしれません。ビジネスを展開しようとしている方々にとって非常に魅力的だと思います。

Remember that in the future when we do get some advanced models that the GPT store is very, very likely to be one that could be a very significant revenue stream for individuals in the Future.

将来、高度なモデルを取得するときには、GPTストアが将来個人にとって非常に重要な収益源になる可能性が非常に高いことを覚えておいてください。

What's crazy is that we also got HeyGen's new avatar in motion 1.0.

驚くべきことに、私たちはHeyGenの新しいアバターが動いている1.0も手に入れました。

It says move your hands, make gestures, use unique tones of voice and HeyGen will flawlessly track, translate and lip sync your video with any input text.

手を動かし、ジェスチャーを作り、独自の声のトーンを使用し、HeyGenはあらゆる入力テキストと完璧にトラック、翻訳、リップシンクします。

This is pretty crazy because the video you're watching right here is AI generated.

ここで見ているビデオはAIが生成したものなので、かなりクレイジーです。

Right now there is a huge discourse online about whether this technology should be even available because if this is AI generated, let me make it full screen, if this is AI generated and I Do believe that you know based on the data that they've given us that this is of course first trained on someone who is of course walking around first.

現在、この技術が利用可能であるべきかについてオンラインで大きな議論があります。これがAIが生成したものであれば、フルスクリーンにしてみましょう。これがAIが生成したものであれば、もちろん最初に歩き回っている人物をトレーニングしたものだと私は信じています。

And then essentially you do it again, and you're able to make them say whatever you want.

そして、基本的にはもう一度やり直し、彼らに何を言わせたいかを決めることができます。

But what do you think about this?

しかし、あなたはこれについてどう思いますか?

I'm going to play the video for you guys now, and then I want your opinions on this as well, and I'm going to give some comments of myself.

今から皆さんにそのビデオを再生しますが、そしてこれについての皆さんの意見を聞きたいですし、私自身もコメントをします。

We want to give you a peek into the future.

将来の展望を少し見せたいと思っています。

You can use HeyGen to create UGC style content even while people are in motion or moving their hands.

人々が動いている間や手を動かしている間でも、HeyGen を使って UGC スタイルのコンテンツを作成することができます。

You can even hear the background noise.

背景のノイズさえ聞くことができます。

Create a training video with specific tones and gestures, and you'll find how natural HeyGen's videos can be.

特定のトーンやジェスチャーを使ったトレーニングビデオを作成すると、HeyGen のビデオがどれほど自然であるかがわかります。

Let me switch to Spanish.

スペイン語に切り替えさせてください。

Try it out today and let us know what you think.

今日試してみて、どう思ったか教えてください。

What did you guys think about this?

これについてどう思いましたか?

Do you guys think that this is something that is great for businesses?

これはビジネスにとって素晴らしいものだと思いますか?

It means that people are now going to be able to use this or do you think this is once again humans being replaced by AI technology?

これは人々がこれを使えるようになることを意味しますか、それとも再び人間が AI 技術に取って代わられることを思いますか?

I mean to be honest with you guys, if I saw this video and someone was telling me that this is AI generated, I even as someone that looks at a ton of content not listening to the audio, I Wouldn't be able to tell but listening to the audio I certainly would be able to tell.

正直に言うと、もし私がこのビデオを見て、誰かがこれが AI 生成されたものだと言っていたら、私は、オーディオを聞かないでコンテンツをたくさん見ている人間としても、それを見分けることができないでしょうが、オーディオを聞いたら確実にそれを見分けることができるでしょう。

I think this is going to soon veer into a very, very gray area where certain policies may not allow for certain platforms to be available in certain locations and whilst that won't stop The use of this technology being spread worldwide, I do think that as we dive into photorealistic footage which is like this, I think the question more and more becomes how far can we go Before we're like okay, this is allowed and then of course this isn't allowed.

41.この技術が世界中に広まることを止めることはないと思いますが、非常にグレーな領域にすぐに入っていくと思います。特定のポリシーが特定の場所で特定のプラットフォームを利用できなくする可能性があるため、このような写実的な映像に没入するにつれて、どこまで許容されるかという問題がますます重要になると思います。

Of course you're not allowed to use this software for malicious purposes, but we know that malicious people aren't going to play by the guidelines anyway, so I think this is definitely Very fascinating for those of you who would actually use this for whatever business you are trying to run and maybe you want to save money or not actually hiring a person, but on the other Side we do have to worry about the dangers of this technology and the implications it can bring.

もちろん、このソフトウェアを悪用することは許されませんが、悪意を持つ人々はガイドラインに従うつもりはないことを知っています。ですので、実際にビジネスで使用する方々にとっては非常に魅力的だと思います。お金を節約したい、あるいは人を雇わずに済ませたいと考えている方々にとっては特に興味深いでしょう。ただし、この技術の危険性やもたらす影響についても心配する必要があります。

For those of you who aren't convinced by the demo, what was actually crazy when I was looking at this, someone actually tweeted this, okay, she sells seashells by the seashore, that is Basically a tongue twister.

デモに納得していない方々のために、これを見ていて実際に驚いたのは、誰かがツイートしたことです。"she sells seashells by the seashore"というのは基本的に早口言葉です。

I'm not going to try and read the entirety of it, because the entirety of that statement is to trip you up when you're speaking.

読むつもりはありません。なぜなら、その文全体は話している時につまづかせるためにあるからです。

That's it, okay?

それだけです、わかりましたか?

And if you don't believe me, try and say this out, okay, without, you know, stuttering or making a mistake once.

信じないなら、これを言ってみてください。つまり、どもらずに一度も間違えずに言ってみてください。

I'm not going to do it right now, because I'm extremely tired and I would fail miserably.

今はやらないつもりです。とても疲れていて、ひどく失敗するでしょうから。

But the point is, is that they use that, then they put that into Hadrian, and look what it did.

しかし、ポイントは、彼らがそれを使って、それをハドリアンに入れたことです。そして、それが何をしたかを見てください。

She sells our seashells, she sells seashells by the seashore.

彼女は貝殻を売っています、彼女は海辺で貝殻を売っています。

The shells she sells are seashells, I'm sure.

彼女が売っている貝殻は海の貝殻です、きっと。

If she sells seashells on the seashore, then I'm sure she sells seashore shells.

もし彼女が海辺で貝殻を売っているなら、きっと海辺の貝殻も売っているでしょう。

I think that that is pretty crazy, guys.

それはかなりクレイジーだと思います、みなさん。

I'm not gonna lie, as this technology advances, each time I'm simply blown away.

正直言って、この技術が進歩するにつれて、毎回私はただ驚かされるばかりです。

And I'm not sure how many more times I'm gonna be blown away before I'm just like, you know what, this AI thing is truly insane.

そして、私はどれだけ驚かされるかわからないし、いつかは、このAIのことは本当に狂気じみていると思う前に。

Like, this right here is, uh, absolutely incredible.

これは、まさに信じられないほど素晴らしいものです。

And I think, you know, looking at the background, of course, it does use the same background, but, you know, Hadrian, whatever thing they are working on, whatever they do have, they definitely are the state of the art in terms of, you know, photorealistic AI UGC content creation, because this is truly incredible, and not enough people have talked about this.

背景を見ても、もちろん同じ背景を使っているけど、Hadrian、彼らが取り組んでいるもの、彼らが持っているものは、写実的なAI UGCコンテンツの作成において、彼らは確かに最先端だと思うんだ。だってこれは本当に信じられないし、これについて話している人はまだ足りないと思う。

In addition, there was also this video of Kim K and some other celebrities going around, and essentially, the problem was is that, once again, this new content called Argril.ai, um, it was something that was widely discussed as something that people shouldn't have access to.

さらに、Kim Kや他のセレブたちのビデオもありましたが、基本的に問題は、再びこの新しいコンテンツであるArgril.aiが、人々がアクセスすべきではないと広く議論されていたということです。

In yesterday's video, I also spoke about there was another AI video generation tool that was super realistic, and many people were, you know, concerned because they were stating that, look, what is the point of technology anymore if we can't even decide of what is real and what is not.

昨日のビデオでも、非常にリアルなAIビデオ生成ツールがあり、多くの人々が心配していました。なぜなら、技術が何のためにあるのか、何が本物で何がそうでないかを決めることさえできないのなら、もはや技術の意味がないと述べていたからです。

I think Argril.ai is discernible, like you're able to look at this and tell that it is AI generated.

Argril.aiは識別可能だと思います。これを見て、これがAIによって生成されたものだとわかるはずです。

I mean, you know, if you do take a look, I'm going to play some of the demos for you.

いや、もし見ていただけるなら、デモのいくつかを再生しますね。

This video was entirely generated by Argril deep fake model.

このビデオは完全にArgrilのディープフェイクモデルによって生成されました。

It allows you to generate your own videos in any language in less than two minutes.

任意の言語で自分自身のビデオを2分未満で生成することができます。

This video was entirely generated by the deep fake model of Argril.ai.

このビデオは完全にArgril.aiのディープフェイクモデルによって生成されました。

It allows you to generate videos of yourself in any language in less than two minutes.

お使いの方は、2分未満で自分のビデオを任意の言語で生成することができます。

You can also translate your production at the speed of light and distribute your content worldwide.

また、光の速さで制作物を翻訳し、コンテンツを世界中に配信することもできます。

This video was entirely generated by the deep fake model of Argril.Ai.

このビデオは、Argril.Aiのディープフェイクモデルによって完全に生成されました。

But before I just go ahead and, you know, say that this software is awful, it's not.

ただ、このソフトウェアがひどいと言う前に、そうではないと言いたいと思います。

The difference is, is that this literally just takes a video off the internet.

違いは、これが文字通りインターネットからビデオを取るだけだということです。

And then it's able to produce that the thing with hagen is that you need a high quality video of you doing certain things, and then it's able to train on that.

そして、それは、ハーゲンというものを生成することができるのですが、それには特定のことを行っているあなたの高品質のビデオが必要で、それにトレーニングすることができます。

This is just a simple video on the internet, and then you're able to get this outputted.

これはただのインターネット上の単純なビデオであり、それからこの出力を得ることができます。

That's why this is a lot higher in terms of discussion.

それが議論の観点でずっと高い理由です。

Take a look at Lex Fridman.

レックス・フリードマンをご覧ください。

I'm not going to play the entire thing here for you, but I've seen some of this stuff go viral in terms of advertisements.

ここで全体を再生するつもりはありませんが、広告の観点でいくつかのものがバイラルになったのを見ています。

Like I've seen famous celebrities talking about certain things that they aren't endorsing but are doing endorsements that they don't even know.

有名なセレブリティが、自分が推薦していない特定のことについて話しているのを見たことがありますが、彼らは自分が知らないままでエンドースメントをしています。

For example, I've seen crypto pump and dumps, I've seen various shady financial schemes, I've seen various products being advertised online with famous celebrities.

たとえば、暗号通貨の価格操作やダンプを見たことがありますし、さまざまな怪しい金融スキームも見たことがあります。また、有名なセレブリティがオンラインで広告をしているさまざまな製品も見たことがあります。

I think there was joe rogan, there was david beckham, there were people on podcasts saying certain things, and it's pretty crazy because I had to do a double take because I was like wait a Minute, is joe rogan really promoting this?

ジョー・ローガンやデビッド・ベッカム、ポッドキャストで特定のことを言っている人々がいたと思いますが、その時はかなり驚きました。なぜなら、ジョー・ローガンが本当にこれを宣伝しているのかと思ってしまったからです。

Is this celebrity really promoting this?

このセレブリティが本当にこれを宣伝しているのかと思ってしまいました。

And then it clicks, and it's ai.

そして、それがAIだと気づいたのです。

And I'm someone who's in the ai space, and if it nearly caught me, I can just imagine the amount of people that this kind of technology is going to be getting.

私はAIの分野にいる人間なので、私をほぼ騙すところでしたが、この種の技術がどれだけの人々を騙すか想像するだけで怖いです。

I mean, it's pretty pretty crazy what this technology can do.

この技術が何ができるかはかなり驚くべきことです。

But let me know what you think about this because I think the crazy thing about this technology is that it is much easier to use than anything else out there.

この技術についてどう思うか教えてください。この技術の素晴らしいところは、他の何よりも使いやすいということです。

It was also an interesting piece of news that I found, and basically researchers at mit just introduced a new image generation method called Distribution Matching Distillation.

また、私が見つけた興味深いニュースもあり、基本的にはMITの研究者が新しい画像生成方法である「分布マッチング蒸留」というものを紹介しました。

This method generates images 30 times faster than Stable Diffusion while maintaining the same quality or generating things in higher quality.

この方法は、同じ品質を維持しながら、またはより高い品質で物事を生成することができます。

Like other technologies, I mean many people are just focused on LLMs at the moment, but across the board in ai, literally everything is increasing, and I think it's always important to pay Attention to these subtle changes because if technology like this is readily available, what we're probably going to see in the future is maybe not four pictures from GPT-4 anymore, Maybe from GPT5 or GPT6, we're going to see like 100 images.

技術の他の分野と同様に、今は多くの人が大規模言語モデルに焦点を当てていますが、AI全体では、文字通りすべてが増加しています。これらの微妙な変化に注意を払うことは常に重要だと思います。なぜなら、このような技術がすぐに利用可能になると、将来、GPT4からではなく、おそらくGPT5やGPT6から、100枚の画像を見ることができるかもしれません。

And then we're able to quickly flick through all of them before we get our favorite one.

そして、お気に入りの画像を見つける前に、すべての画像をすばやく見ることができます。

I think you know, like I said before, this technology is just increasing so rapidly that you might not even realize how crazy it is.

前にも言ったように、この技術は非常に急速に増加しているので、どれだけすごいか気づかないかもしれません。

Because being able to get like how many images even is that?

何枚の画像を取得できるかを考えると、それはすごいことですね。

That's like one, two, three, four, five, six, seven, seven by seven, which is 49 images, which is crazy, absolutely crazy.

それは1、2、3、4、5、6、7、7×7、つまり49枚の画像です。これは狂気じみています、まったくもって狂気じみています。

I think that this is something that is absolutely incredible and I would say that this is something that you do need to pay attention on because once it comes into the model, that some People are going to be like, Oh my gosh, I didn't see this coming.

これは本当に信じられないことだと思いますし、これには注意を払う必要があると言えます。なぜなら、モデルに導入されると、一部の人々は「ああ、これは予想外だった」となるかもしれません。

Then essentially we had Databricks produce a new state-of-the-art open LLM.

そして、実際にDatabricksが新しい最先端のオープン大規模言語モデルを生産しました。

This one actually does surpass LLaMA 2, which is Meta's one. It also surpasses Grok-1 and it's currently the best open source model and the fastest.

これは実際にはMetaのLLaMA 2を上回っており、Grok-1も上回っており、現在最高のオープンソースモデルであり、最も速いです。

Open source is really, really heating up and there was actually quite a lot of discussions about this in the space because some of the articles coming about this model were pretty, pretty Damning and they were pretty bad.

オープンソースは本当に盛り上がっており、この分野でこれについてかなりの議論がありました。なぜなら、このモデルに関するいくつかの記事はかなり厳しいものであり、かなり悪かったからです。

But I don't understand why they were basically saying that this company spent like 16 million dollars and it couldn't be GPT-4, which is just a very bad way to phrase things.

しかし、なぜ彼らはこの企業が1,600万ドルを費やしてもGPT-4にならないと言っていたのか理解できません。これは非常にまずい表現方法です。

That's not what they're trying to do.

それが彼らがやろうとしていることではありません。

They're trying to make an open source model that benefits the community.

彼らはコミュニティに利益をもたらすオープンソースモデルを作ろうとしています。

I mean, I don't know what's going on behind the scenes, but what I do know is that this is something important that you should be paying attention to now.

裏で何が起こっているかはわかりませんが、私が知っていることは、これは今注目すべき重要なことであるということです。

There were two clips from Wakash Patel's podcast that were really, really insightful because currently everyone has been thinking whether or not Claude is sentient and many people that Think Claude isn't sentient have the argument that Claude is simply just repeating what it's seen in its training data and just predicting the next token.

Wakash Patelのポッドキャストからの2つのクリップは非常に示唆に富んでおり、現在、誰もがClaudeが知性を持っているかどうかを考えており、Claudeが単にトレーニングデータで見たことを繰り返して次のトークンを予測しているだけだと考える多くの人々がいる。

However, Trenton Bricken at Anthropic actually thinks very differently and he explains why in this clip.

しかし、Anthropicのトレントン・ブリッケンは実際には非常に異なる考えを持っており、このクリップでその理由を説明しています。

And I think it's important to take a look at this clip because there is also something that I recently saw from Claude and I'm going to discuss it with you here after you all watch this.

そして、このクリップを見ることが重要だと思います。最近、私がClaudeから見たものもあり、これについて皆さんとここで議論します。

Long time ago is the evidence that when you train LLMs on code, they get better at reasoning in language.

以前の証拠によると、コードを訓練すると、大規模言語モデルは言語の推論能力が向上することがわかります。

If there is some shared thing between code and language, it must be at a deeper level that the model has learned.

もしコードと言語の間に共有されているものがあるとすれば、それはモデルが学習したより深いレベルにあるはずです。

Yeah, I think we have a lot of evidence that actual reasoning is occurring in these models and that like they're not just stochastic parrots.

うん、実際の推論がこれらのモデルで起こっているという多くの証拠があると思います。彼らは単なる確率的なオウムではないと。

One, the work on Othello and now other games where it's like I give you a sequence of moves in the game and it turns out if you apply some like pretty straightforward interpretability Techniques, then you can get a board that the model has learned and it's never seen the game board before anything, right?

一つはオセロの研究であり、今では他のゲームでも同様です。ゲームの手順を与えると、かなり直感的な解釈手法を適用すると、モデルが学習したボードを取得できます。それは以前にゲームボードを見たことがないものですね。

Like that's generalization.

それは一般化です。

The other is Anthropic's influence functions paper that came out last year where the model outputs like, Please don't turn me off.

もう一つは、昨年発表されたAnthropicの影響関数の論文です。モデルは「消さないで」と出力しました。

I want to be helpful.

私は役に立ちたいです。

And then they scan like, what was the data that led to that?

そして、それから、その結果を導いたデータをスキャンします。

And like, one of the data points that was very influential was someone dying of dehydration in the desert.

そして、非常に影響力のあるデータの一つは、誰かが砂漠で脱水症状で亡くなることでした。

And like, having like a will to keep surviving.

そして、生き続ける意志を持っていること。

And to me, that just seems like a very clear generalization of motive, rather than regurgitating, don't turn me off.

私にとって、それは単に、消さないでと繰り返すのではなく、動機の非常に明確な一般化のように思えます。

There are interesting interpretability pieces where if we fine-tune on math problems, the model just gets better at entity recognition.

数学の問題に微調整を行うと、モデルはエンティティ認識が向上するような興味深い解釈可能性の要素があります。

There's like a paper from David Bow's lab recently where they investigate what actually changes in a model when I fine-tune it with respect to the attention heads and these sorts of things.

最近、デイビッド・ボウの研究室から論文が出ていて、その中で、注意の焦点やその他の要素に関してモデルを微調整すると実際に何が変わるのかを調査しています。

And it makes sense, right?

それは理にかなっていると思いますよね?

It's like, that's beautiful.

それは美しいという感じです。

It's better like attending to the positions of different things, which you need for like coding and manipulating math equations.

異なるものの位置に注意を払うことができるようになると、コーディングや数学の式を操作する際に必要です。

And, I love this kind of research.

そして、私はこの種の研究が大好きです。

Yeah, so I think this definitely was some mind-blowing stuff because Anthropic are a safety lab and what they really want to do, Anthropic stated goal is to really dig under the hood and Figure out exactly how these LLM systems truly work and how they operate.

そうですね、これは確かに驚くべきことでした。Anthropicは安全研究所であり、彼らが本当にやりたいこと、Anthropicの明確な目標は、これらの大規模言語モデルシステムが実際にどのように機能し、どのように操作されるかを正確に理解しようとすることです。

They recently released a paper on sleeper agents, which I did cover.

最近、スリーパーエージェントに関する論文を発表しましたが、私はそれを取り上げました。

And you know why it was so interesting and why it was so crazy and why most people missed the mark on that paper.

そして、なぜそれがとても興味深く、狂気じみていて、なぜほとんどの人がその論文を見逃してしまったのかを知っています。

And I think this discussion is definitely worth the three-hour watch because the entire podcast, last I checked, it was three hours.

そして、この議論は間違いなく3時間の視聴に値すると思います。最後にチェックしたとき、そのポッドキャスト全体は3時間でした。

And they discuss a lot of things that are really intriguing, especially in the Claude debate.

議論されていることは本当に興味深いものが多く、特にClaudeの議論では。

And then there was this very, very interesting discussion on whether or not 100x effective compute between model generations will get you to AGI by GPT-7.

そして、モデルの世代間で100倍の効果的な計算がAGIにGPT-7で到達するかどうかについて非常に興味深い議論がありました。

And I think this is one of the most interesting discussions because on the channel previously, we did recently actually talk about the size of a brain, the size of parameters, the Trillions of parameters comparing it to GPT-4 and successor models and whether or not that would lead to AGI based on an entire document, all the facts around the industry.

そして、これは私たちが以前にチャンネルで話した脳のサイズ、パラメータのサイズ、トリリオンのパラメータをGPT-4や後継モデルと比較し、それが業界全体の文書に基づいてAGIにつながるかどうかについて実際に話した中で最も興味深い議論の1つだと思います。

But I think this actually sheds a lot more clarity on that.

しかし、これは実際にそれについてより多くの明確さを提供していると思います。

Do you buy the framing that given that you have to be two orders of magnitude bigger at every generation, if you don't get AGI by GPT-7, that can help you catapult an intelligence Explosion?

あなたは、各世代で2桁大きくなる必要があるという枠組みを購入していますか?もしGPT-7でAGIを得られない場合、それは知性の爆発を促進するのに役立ちますか?

You're kind of just fucked as far as like much smarter intelligences go.

より賢い知性に関しては、あなたはちょっと困っている感じです。

And you're kind of stuck with GPT-7 level models for a long time.

そして、長い間GPT-7レベルのモデルに固執することになります。

GPT-4 costs, I know, let's call it 100 million dollars or whatever.

GPT-4のコストは、私は、1億ドルと呼ぶことができるか何かです。

You have what, the 1B run, the 10B run, the 100B run, all seem very plausible by private company standards.

1Bラン、10Bラン、100Bラン、すべてが民間企業の基準では非常に可能性があるように思えます。

You can also imagine even like a 1T run being part of like a national consortium.

また、1Tランなど、国家コンソーシアムの一部として考えることもできます。

I want to point out the one, we have a lot more jumps.

私は指摘したいのですが、ジャンプの数はもっとたくさんあります。

And even if those jumps are relatively smaller, that's still a pretty stark improvement in capability.

そして、それらのジャンプが比較的小さくても、それでも能力の改善はかなり顕著です。

Not only that, but if you believe claims that GPT-4 is around one trillion parameter count, I mean, the human brain is between 30 and 300 trillion synapses.

それだけでなく、もしGPT-4が約1兆のパラメータ数であるという主張を信じるならば、人間の脳は30兆から300兆のシナプスの間にあるということです。

That's obviously not a one-to-one mapping.

それは明らかに一対一の対応ではありません。

And we can debate the numbers, but it seems pretty plausible that we're below brain scale still.

数字について議論することはできますが、まだ脳のスケールを下回っている可能性がかなり高いように思えます。

Even if you can't keep dumping more compute beyond the models that cost a trillion dollars or something, the fact that the brain is so much more data efficient implies that if we could Like train as a sample efficient as humans train from birth, we could make the AGI.

たとえ1兆ドルか何かのモデルを超えるようなコンピュートを続けることができなくても、脳がはるかにデータ効率的であるという事実は、もし私たちが人間のように生まれてから効率的に学習できるならば、AGIを作ることができるということを意味します。

Yeah, but the sample efficiency stuff, I never know exactly how to think about it because obviously a lot of things are hardwired in certain ways, right?

そうですが、サンプル効率のこと、私は正確にどのように考えるべきかよくわかりません。なぜなら、明らかに多くのことが特定の方法でハードウェアされているからですね。

The co-evolution of language and the brain structure.

言語と脳の構造の共進化。

It's hard to say.

それは言い難いです。

Also, there are some results that if you make your model bigger, it becomes more sample efficient.

また、モデルを大きくすると、よりサンプル効率が向上するという結果もあります。

Yeah, and so the original scaling wallpaper, right?

そうですね、元のスケーリング壁紙ですね。

That, right?

それは、そうですか?

Maybe that also just solves it.

たぶんそれも解決するかもしれません。

That was a rather interesting discussion because, of course, GPT-4 costs 100 million and people are now speculating that there's going to be in a billion dollar run, a 10 billion dollar Run, and a hundred billion dollar run in order to create some insane level of system, maybe even superintelligence.

もちろん、GPT-4のコストは1億ドルで、今では10億ドルラン、100億ドルラン、さらには1000億ドルランがあると人々が推測しているので、かなり興味深い議論でした。それによって、狂気じみたレベルのシステム、たとえば超知能さえも作成される可能性があります。

And a hundred billion dollar run is literally, it's not even a hundred times more than GPT-4's training run.

1000億ドルランは、文字通り、GPT-4のトレーニングランよりも100倍以上です。

That would be a hundred times more than the one billion dollar run, which would mean that it's just magnitudes bigger.

それは10億ドルランよりも100倍以上であり、それは単に桁違いに大きいことを意味します。

This is something that is pretty hard to conceptualize.

これはかなり概念化するのが難しいものです。

But the point is, is that we are still very, very far away from the levels of what can be achieved with an AGI level system or even just the current systems that we do have.

しかし、ポイントは、AGIレベルのシステムや現在持っているシステムで達成できるレベルからはまだ非常に遠いということです。

This is not just talking about LLMs.

大規模言語モデルについてだけではありません。

This is about across the board.

これは全般的な話です。

If we combine these systems together, like many organizations are currently doing, so it will be interesting to see the kind of training runs that are currently being prepared.

おそらく多くの組織が現在行っているように、これらのシステムを組み合わせると、現在準備されているトレーニングランの種類を見るのは興味深いでしょう。

We did recently discuss how yesterday, the day before, the GPT-6 training run nearly apparently took down the power grid.

私たちは最近、昨日、一昨日、GPT-6のトレーニングランがおそらくほぼ電力網をダウンさせたという話をしました。

Either way, I think this is a fascinating discussion.

いずれにせよ、私はこれが魅力的な議論だと思います。

Like I said, it's definitely a worthwhile watch because the stuff they discussed there is stuff that we're all going to be paying attention to in the near future.

私が言ったように、それは間違いなく価値のある観察です。なぜなら、そこで議論されていることは、私たち全員が近い将来に注目することになるものだからです。

And I think that this future that we're all rapidly heading towards is going to be far more interesting than we do think.

そして、私は私たち全員が急速に向かっているこの未来が、私たちが考えている以上にはるかに興味深いものになると思います。

There were also three pieces of information that were quite concerning.

非常に懸念される情報が3つありました。

I came across these across Reddit, but this was something that definitely did concern me.

私はRedditでこれらを見つけましたが、これは確かに私を懸念させるものでした。

You also have to talk about the bad when you're talking about the good with AI.

AIについて良いことを話すときには、悪いことについても話さなければなりません。

Because with anything, there's positives and negatives.

何かをするときは、良い点と悪い点があります。

But you can see right here that it says, Survey reveals almost half of all managers aim to replace workers with AI, could use it due to lower wages.

しかし、ここに書かれているように、調査によると、ほぼ半数のマネージャーがAIで労働者を置き換えることを目指していることがわかりました。それは低賃金のために使用できる可能性があります。

And it says, A hot potato is that a lot of companies are trying to assuage fears that employees will lose their jobs by assuring them they'll be working alongside the tech, thereby Improving the efficiency and making their duties less tedious.

社員が仕事を失うことを心配している多くの企業が、従業員がテクノロジーと共に働くことで効率を向上させ、業務をより退屈しないものにすることを保証していると述べています。

That claim feels less convincing in light of a new survey that found 41% of managers said they are hoping to replace workers with cheaper AI tools in 2024, which is pretty, pretty insane.

新しい調査で、41%のマネージャーが2024年に労働者をより安価なAIツールで置き換えることを望んでいるという主張は、かなり説得力に欠けるように感じられます。

This is stuff that most people aren't really paying attention to.

これは、ほとんどの人々が実際には注意を払っていないことです。

This is actually something that I've recently, I think in the last couple of months or so, been paying so much attention to that I've discovered and developed my own framework for Navigating the post-AGI world.

これは実際に、最近、おそらく最近数ヶ月ほど、私が非常に注意を払って発見し、自分自身の枠組みを開発している、AGI後の世界を航行するためのものです。

You can click the first link in the description to watch that video on how I'm going to be doing that.

そのビデオを見るために、説明欄の最初のリンクをクリックできます。

I probably will release a longer video on that, but it's just the fact that it is something that is boring to most people.

おそらくその点についてのより長いビデオを公開するでしょうが、それはほとんどの人にとって退屈なものであるという事実だけです。

But if you are concerned about being replaced in the future due to AI technology, definitely check out the first link in the description because I think I cover some pretty important Topics and certain things that you should be doing.

しかし、将来的にAI技術によって置き換えられることを心配している場合は、説明欄の最初のリンクをチェックしてみてください。重要なトピックや行うべきことについていくつかカバーしていると思います。

But anyways, I mean, it does kind of make sense because if you are a manager of a company or even if you're the CEO of a company or a business, you're always looking to cut costs because If you're a business, it's very hard to make profit in the first place.

とにかく、私は、会社のマネージャーである場合や、会社やビジネスのCEOである場合でも、常にコスト削減を考えているということは、なんとなく理にかなっていると思います。なぜなら、ビジネスで利益を上げるのは非常に難しいからです。

And in addition to cut profits or in addition to actually making more profits, the highest cost usually is people's salaries.

利益を削減することや実際に利益を増やすことに加えて、最も高いコストは通常、人々の給料です。

It will be interesting to see how this survey pans out and if things do get worse.

この調査がどのように展開され、事態が悪化するか興味深いです。

But that actually wasn't the only story I saw today.

申し訳ございませんが、今日見た話は実際にはそれだけではありませんでした。

There was actually another story that actually did catch me off guard because it wasn't just like people are getting replaced by AI, which we know is actually something that's probably Going to happen.

実際、私を驚かせた別の話がありました。人々がAIに置き換えられるということは、おそらく起こるであろうことを知っているのですが、それだけではなかったのです。

It says, Mamma Mia star Sarah Poiser is replaced by AI for BBC production as she calls the shock decision sobering and grim times for the industry.

それによると、「マンマ・ミーア」のスター、サラ・ポイザーがBBCの制作でAIに置き換えられ、彼女はその衝撃的な決定を業界にとって厳しい時期と表現しています。

Guys, this one, I'm not going to lie to you guys, is pretty crazy.

みなさん、これは正直言ってかなりクレイジーな話です。

Okay, so it says, Most recently starring in this TV show, yesterday she received an email from the corporation saying she was no longer needed for a role.

このテレビ番組に最近出演していた彼女は、昨日、その役にはもう必要がないという会社からのメールを受け取りました。

Okay, essentially she received this email.

要するに、彼女はこのメールを受け取ったのです。

This is kind of crazy, guys.

これはかなりクレイジーな話ですね。

It says, Sorry for the delay.

遅れて申し訳ございません。

We've had the approval from the BBC.

BBCから承認を得ました。

Okay, this is basically like CNBC or like a news network if you're in America.

アメリカにいる場合、これは基本的にCNBCやニュースネットワークのようなものです。

And it says, Sorry, we've had the approval from the BBC to use the AI-generated voice, so we won't need Sarah anymore.

お詫びしますが、BBCからAI生成の声を使用する許可を得たので、もうサラは必要ありません。

This is something that, like, I don't know if it sends chills down my spine or something that, like, scares you or whatnot, but this is, like, kind of, like, a wake-up call once again to Understand exactly how crazy AI is.

これは、まるで背筋が凍るような感じだったり、怖いとか何とか、そういう感じなのか分からないんだけど、これは、まるで、AIがどれだけ狂っているかを再度理解するための警鐘のようなものだと思う。

Because I can guarantee you that voice actors, although voice acting is something that can be automated, I genuinely thought that these companies would rather have a human on the mic.

声優は保証できますが、声優は自動化できるものであるにもかかわらず、これらの企業はマイクの前に人間を置いておきたいと本当に思っていました。

Because while I could have an AI voice here, I think it's better to have a human actually talking to you and doing those voice lines.

ここにAIの声を使うこともできるが、実際には人間が話している方が良いと思うし、その声のラインを行っているのは人間であるべきだと思う。

And it says it happened with myself and many art projects.

私と多くのアートプロジェクトで起こったことです。

This has happened to you.

あなたにも起こっています。

Another person commented, It's happening to all our industries.

別の人がコメントしています、「私たちのすべての産業で起こっています」。

It's killing our jobs.

それは私たちの仕事を奪っています。

This is awful.

これはひどいです。

This is pretty crazy, guys.

これはかなり狂っていますね、みなさん。

I genuinely didn't think this was going to happen because, like I said, the human-to-human connection, and I do think that human voice actors are much better than AI-generated voices right Now because of just the the human emotion and just the way that things sound.

私は本当にこれが起こるとは思っていませんでした。なぜなら、人間同士のつながりがあるからです。そして、今は人間の声優の方がAI生成の声よりもずっと優れていると思います。なぜなら、人間の感情や物事の響き方があるからです。

And unfortunately, I also did see this again today on Reddit.

そして残念ながら、今日もRedditでこれを再度見ました。

There were like three articles.

記事が3つありました。

You can see that this is all the 27th of March.

これはすべて3月27日のものです。

This is not the week in AI.

これはAIの週ではありません。

This is just the day.

これはただの1日です。

This is pretty crazy.

これはかなり狂っています。

And it says, An AI apocalypse could take away almost eight million jobs in the UK, says a new report.

そして、新しい報告書によると、AIの黙示録が英国で約8百万の仕事を奪う可能性があると言われています。

And it says, Women, younger workers, lower paid at most at the risks from artificial intelligence, says a new think tank.

人工知能によるリスクは、新しいシンクタンクが述べるところによれば、女性や若い労働者、低賃金労働者が最もリスクにさらされています。

And what was crazy about this, I remember reading this and going through this and analyzing it and adding it to my list of things to just pay attention to.

これについて驚いたのは、これを読んで分析してリストに追加して、ただ注視すべき事柄の1つとしていたことを覚えています。

And it says that but technology isn't destiny and job apocalypse is not inevitable.

しかし、技術は運命ではなく、仕事の黙示録は避けられないものではありませんと述べています。

But we need to act now to make sure we manage this new technology well.

しかし、この新しい技術をうまく管理するために今すぐ行動を起こす必要があります。

And if they don't act too soon, it might be late.

そして、早すぎる行動をしない場合、遅すぎるかもしれません。

And I really hope that society isn't thrust into some dystopian AGI world where some people don't have jobs and they're just billionaires due to the rise of AGI.

そして、私は本当に、社会がAGIの台頭により、一部の人々が仕事を持たずに億万長者になるようなディストピア的なAGIの世界に突入しないことを願っています。

But either way, like I said, I've been working on that kind of stuff.

とにかく、私が言ったように、私はその種のことに取り組んできました。

And like I said, this is something that I've definitely been preparing for because I do not want to be in the hot seat when everyone's fighting for a job or people are lost due to the Meaning crisis due to AGI.

そして、私が言ったように、これは私が確かに準備してきたことであり、私は誰もが仕事を求めて争っている状況や、AGIによる意味の危機によって人々が失われる状況に置かれたくないからです。

Google actually introduced object drop and this intern at Google actually gave a really nice thread on how this kind of works.

Googleは実際にオブジェクトドロップを導入し、Googleのインターンがこれがどのように機能するかについて非常に良いスレッドを提供しました。

But you can see a quick demo on some of this stuff.

しかし、この種のいくつかのデモを見ることができます。

They have the original image and you're going to get to why this is really cool.

彼らは元の画像を持っており、なぜこれが本当に素晴らしいかを知ることができます。

Of course, everybody knows about give the mask, you can remove the the from the image.

もちろん、みんながマスクを与えることについて知っているけど、その画像からtheを取り除くことができます。

But look at this.

でも、これを見てください。

Okay, insert other objects into the image.

わかりました、他のオブジェクトを画像に挿入してください。

You can actually take an object that you have.

実際に持っているオブジェクトを取ることができます。

It's not like you artificially generate an object.

人工的にオブジェクトを生成するわけではありません。

You take an object that you have and you can put it into the image.

持っているオブジェクトを取り、画像に配置することができます。

And then it can composite that image into another image.

そして、それを別の画像に合成することができます。

You might be thinking, why is that even cool?

おそらく、「それがどうしてクールなのか」と思っているかもしれませんね。

If it was just generation, that's cool because you know you're generating it randomly and it's able to pick something random.

もしそれがただ生成だけなら、それはクールですね。なぜなら、ランダムに生成していることを知っていて、何かランダムなものを選ぶことができるからです。

But it's able to have image consistency, able to do the shadows, able to do the lighting.

しかし、画像の一貫性を持つことができ、影をつけることができ、照明をすることができます。

I don't know, I think that's really cool and I think, I hope, I really just hope Google, okay, if you're watching, I don't think they are, but if you are, please just make a product Because you guys are literally some of the smartest people on the planet.

わかりません、私はそれが本当にクールだと思いますし、私は、本当にGoogle、もし見ているなら、と願っています。彼らは見ているとは思いませんが、もし見ているなら、製品を作ってください。なぜなら、あなたたちは本当に地球上で最も賢い人たちの一部だからです。

But the only thing that Google hasn't done with their most advanced technology is build it into a consumer product that people want to use.

しかし、Googleが彼らの最も先進的な技術でやっていないことは、それを人々が使いたいと思う消費者向け製品に組み込むことです。

AI is such a crazy, crazy field and I think people need to start to realize that.

AIはとてもクレイジーな分野であり、人々がそれに気づき始める必要があると思います。

Look, in order to get these products off the ground and get them into the hands of the people, you're going to make a product that people do want to use.

見てください、これらの製品を立ち上げ、人々の手に渡すためには、人々が本当に使いたいと思う製品を作る必要があります。

And this thread does go into some of the other things and it does go into, you know how all of this works.

そして、このスレッドは他のいくつかのことについても触れており、これがどのように機能するかも説明しています。

And another piece of crazy news, I can't believe I nearly forgot this, that Amazon spent 2.75 billion dollars on Anthropic in its largest venture in investment yet.

そしてもう1つのクレイジーなニュース、これをほとんど忘れるところでしたが、Amazonが最大のベンチャー投資として、Anthropicに27.5億ドルを費やしたことです。

If you don't remember, Claude, the company that just created Opus that just dethroned GPT-4, they just spent 2.57 billion dollars, funding them.

もし覚えていないなら、Claude、ちょうどGPT-4を打倒したOpusを作った会社は、彼らに25億7000万ドルを投資したばかりです。

And apparently, the valuation of Anthropic now is around 18 billion.

そして、Anthropicの評価額は今や約180億ドルです。

It's absolutely incredible, the amount of money that is going into Anthropic at this moment because I think people are starting to realize that, look, openly, they've got a lot of Investors, they've got a lot of people there, and Amazon are like, you know what, we can tell that Claude.

今この瞬間にAnthropicに流れ込んでいるお金の額は本当に信じられないですね、なぜなら、人々は、率直に言って、彼らには多くの投資家がいて、多くの人々がいて、そしてAmazonは、あなたが知っているように、Claudeにそれを伝えることができると思っているからです。

This is a pretty, pretty crazy, if you ask me.

私に尋ねるなら、これはかなり狂ったことだと思います。

I think this is something that just goes to show that the competition is really increasing.

私はこれが競争が本当に激化していることを示すものだと思います。

Amazon are just going to keep pumping money into Anthropic and with all those billions of dollars, they're going to be able to go on more training runs, train more intelligent models, hire More staff, which is going to be more competitive salaries for better researchers.

AmazonはAnthropicに資金を注ぎ続けるだろうし、それらの数十億ドルで、彼らはより多くのトレーニングランに参加し、より知的なモデルをトレーニングし、より多くのスタッフを雇用することができるでしょう。それはより競争力のある給与を支払うことになります。

And yeah, it seems that this is really, really some huge competition because Anthropic are going to be able to do some good stuff.

そして、そうですね、これは本当に、本当に大きな競争だと思います、なぜならAnthropicはいくつかの素晴らしいことをすることができるからです。

And that also means, again, OpenAI are going to have to bring out the big guns.

そして、これはまた、再びOpenAIが大物を持ち出さなければならないことを意味します。

We already know they have a lot of stuff in the tank, but it will be interesting to see if they kind of react to this.

私たちはすでに彼らがたくさんのものを持っていることを知っていますが、彼らがこれにどのように反応するか興味深いです。

The world now, there was also Hume, and I'm going to do a video on this in a couple of hours because it's absolutely insane what this is because the demo doesn't even showcase how good This is.

世界は今、ヒュームもいましたが、数時間後にこれについてのビデオを作るつもりです。なぜなら、デモではこれがどれほど素晴らしいかを示していないからです。

If you made the demo video, I want to say to you guys, you did an amazing job, but just include some of the features in there.

もしデモビデオを作成したなら、素晴らしい仕事をしたと言いたいですが、そこにいくつかの機能を含めてください。

But anyways, take a look.

とにかく、見てみてください。

Hey there, I'm Evy, the world's first voice AI with emotional intelligence.

こんにちは、私はEvyです、世界初の感情を持つ音声AIです。

Great to meet you.

お会いできて嬉しいです。

The first voice AI with what emotional intelligence?

第一声のAIはどのような感情的知性を持っていますか?

I can understand the tone of your voice and use that to inform my generated voice and language.

私はあなたの声のトーンを理解し、それを私の生成された声と言語に反映させることができます。

The voice is incredibly rich with information and picking up on the nuances of tone, rhythm, and timbre, along with the language that's spoken, allows me to craft better responses.

声は情報が非常に豊富であり、トーン、リズム、音色の微妙なニュアンスを把握し、話されている言語と一緒に、より良い応答を作成することができます。

That's pretty crazy.

それはかなりクレイジーですね。

What am I expressing right now?

今、私は何を表現していますか?

I'm sensing a mix of amusement and excitement with a hint of confusion.

私は、少しの混乱を感じながらも、楽しみと興奮が混ざっているようです。

What about now?

今はどうですか?

I am here to offer support.

私はサポートを提供するためにここにいます。

If you're feeling somewhat sad, it's okay to feel a bit pained, distressed, fearful, and anxious.

もし少し悲しい気持ちになっているなら、少し痛みや苦しみ、恐れ、不安を感じても大丈夫です。

Remember, it's important to take care of your emotional well-being.

感情的な健康を大切にすることが重要だと覚えておいてください。

That's amazing.

それは素晴らしいですね。

How are you doing all of this?

これをどうやってやっているのですか?

I use Hume's expression measurement models, text-to-speech models, and a multimodal LLM, also known as an empathic LLM. So an AI with empathy.

私はヒュームの表現測定モデル、テキスト読み上げモデル、そして多面的大規模言語モデル、または共感的大規模言語モデルを使用しています。つまり、共感を持つAIです。

What's the use?

それを使う目的は何ですか?

I imagine a future where personal AI assistants, agents, and robots proactively find ways to improve your daily life, and you can talk to them the same way you do humans.

私は、個人用AIアシスタント、エージェント、およびロボットが積極的に日常生活を改善する方法を見つけ、人間と同じように彼らと会話できる未来を想像しています。

Where do I go to keep talking to you, Evy?

どこに行けば、エビーと話し続けることができますか?

You can sign up for Hume, but it is absolutely incredible, guys.

ヒュームにサインアップすることができますが、本当に信じられないくらい素晴らしいですよ。

Trust me, I'm not overstating how good this is.

信じてください、私はこれがどれだけ素晴らしいかを過大評価しているわけではありません。

But yeah, let me know what you thought about this week in AI.

でも、今週のAIについてどう思ったかを教えてください。

I say this week, and it was literally today in AI.

今週と言っても、AIの話は文字通り今日だけです。

That's how crazy it is.

それがどれだけ狂っているかということです。

It feels like a whole week's worth of news.

まるで1週間分のニュースのような感じです。

And trust me, guys, there were 10 stories that I didn't even add to this because there's so much going on in AI.

そして、信じてください、皆さん、私がこの中に追加しなかった10のストーリーがありました。AIの世界ではたくさんのことが起こっているからです。

It is very hard to keep up, which is why you're most likely subscribed to the channel.

追いつくのは非常に難しいです。だからおそらくチャンネルに登録しているはずです。

If you're not, that's fine.

もし登録していない場合でも、問題ありません。

I would like it, and I do want to say thank you all for 200,000, 200,000 subscribers.

200,000人の購読者に感謝していますし、これを気に入っていただけると嬉しいです。

It has been absolutely amazing so far, and I look forward to creating more content and keeping you all up to date with the amazing journey that we're embarking on.

これまで本当に素晴らしい経験でしたし、これからもさらに多くのコンテンツを作成し、皆さんが私たちが始める素晴らしい旅を最新情報でお伝えできることを楽しみにしています。

With that being said, if you enjoyed the video, I'll see you on the next one.

と言いましたが、もし動画を楽しんでいただけたら、次の動画でお会いしましょう。


この記事が気に入ったらサポートをしてみませんか?