見出し画像

【Chat GPTのパーソナライズ化】英語解説を日本語で読む【2024年1月26日|@TheAIGRID】

この動画は、ChatGPTに追加されたパーソナライズ機能とその秘密について解説しています。新機能により、ChatGPTはユーザーの会話に基づいて応答をカスタマイズでき、ユーザーはChatGPTが記憶する情報を管理できます。この機能はランダムにテストされている一部のユーザーに限定されています。将来的には、完全にパーソナライズされたバージョンのChatGPTが提供され、単純なチャットボットを超えた性能を発揮する可能性があることが説明されています。
公開日:2024年1月26日
※動画を再生してから読むのがオススメです。


There has been some several secret features that have been added to ChatGPT that many people haven't actually realized.

ChatGPTには、多くの人々が実際に気付いていないいくつかの秘密の機能が追加されています。

In today's video, we'll do a quick deep dive on these features and why you should be paying attention because this secret update is a lot bigger than most people do realize.

今日のビデオでは、これらの機能について詳しく説明し、なぜ注目すべきなのかを説明します。なぜなら、この秘密のアップデートは、多くの人々が実際には気付いていないものよりもはるかに大きいからです。

When you hear Sam Altman's conversation, you're going to understand exactly why.

サム・アルトマンの会話を聞くと、なぜそうなるのかがわかるでしょう。

So, one of the tweets that I saw recently that was fascinating was this tweet by TestingCatalog.eth.

最近見たツイートの中で興味深かったのは、TestingCatalog.ethによるこのツイートです。

And essentially, it tweets about the personalization aspect of ChatGPT.

実際には、ChatGPTの個人化の側面についてツイートしています。

Now, if you're familiar with this, you'll know that previously there was a slightly different user interface for this and it was a different text box.

もし詳細を知っているならば、以前は少し異なるユーザーインターフェースがあり、別のテキストボックスがありました。

But there is also a new text box where you can see in the personalization tab.

しかし、個人化タブで見ることができる新しいテキストボックスもあります。

Now, I must add that this personalization is only rolling out to a select few because when OpenAI tends to test things, they test out in certain batches and they do it randomly.

ただし、この個人化は一部のユーザーにのみ展開されています。OpenAIはテストを行う際に、特定のバッチでランダムにテストを行う傾向があるためです。

So, I don't think there's a specific way you can sign up.

ですので、特定の方法でサインアップすることはできないと思います。

Sometimes, they do have the ability for you to test beta features, but features like this that are quite rare and haven't fully been fleshed out yet do randomly appear in some people's accounts and of course, do appear in sometimes in other accounts.

時々、ベータ版の機能をテストすることができる場合もありますが、このような珍しい機能はまだ完全に洗練されておらず、ランダムに一部のアカウントにのみ表示されることがあります。

So, you can see here that it says improve responses with your chats.

ここでは、チャットで応答を改善すると表示されています。

ChatGPT will become more helpful as you chat, picking up on details and prefer to tailor its responses to you.

ChatGPTは、チャットすることでより役立つ存在になり、詳細を把握し、応答をあなたに合わせるようになります。

To understand what ChatGPT remembers or to teach it something new, just sayremember that I like concise sponsors, I just got a puppy' andwhat do you remember about me and where did we leave on my last project.

ChatGPTが覚えていることや新しいことを教えるには、「私は簡潔なスポンサーが好きで、つい最近子犬を飼いました」と言ってください。また、「私について覚えていることや、前回のプロジェクトでどこで終わったかを教えてください」と言ってください。

Now, if you remember recently, there was a new update after the GPT store launch and there was a very small blog post about this feature.

最近、GPTストアのローンチ後に新しいアップデートがありましたが、この機能については非常に小さなブログ投稿がありました。

Now, I can't find the exact post, but essentially what they are doing is they are actually rolling out this feature because there were tweets by like Greg Brockman and some other OpenAI employees that do confirm that this is 100% a real thing.

具体的な投稿は見つけられませんが、基本的に彼らが行っているのは、グレッグ・ブロックマンや他のOpenAIの従業員によるツイートが100%本物であることを確認したため、この機能を展開しているということです。

So, it's not something that is going to be going away because some of the comments made by Sam Altman in a recent interview with Bill Gates prove that this is the gradual update that they were talking about and this is where they're going to be leading towards.

ですので、これは消えてしまうものではなく、サム・アルトマンがビル・ゲイツとの最近のインタビューで述べたコメントからもわかるように、これは彼らが話していた段階的なアップデートであり、彼らが進む方向性です。

Now, something additionally that I did see from the screenshot that I will leave a link to in the description below is, of course, the manage memory feature.

もちろん、以下の説明欄にリンクを残しておきますが、スクリーンショットから追加で確認したことは、メモリ管理機能です。

So, you can see right here that it says as you chat with ChatGPT, the details and preferences it remembers will be shown here.

ChatGPTとチャットする中で、ChatGPTが覚えている詳細や好みがここに表示されます。

So, it seems like we're going to be able to see exactly the details that ChatGPT is managing to pick up on us and that means that we're going to be able to see just how much information ChatGPT is remembering from our conversations and what kind of things it thinks about us.

ChatGPTが私たちについて把握している詳細を正確に見ることができるようです。つまり、私たちの会話からChatGPTがどれだけの情報を覚えているのか、どのようなことを考えているのかがわかるということです。

Because if it does get some things wrong, we could then go ahead and delete them and of course, edit them.

なぜなら、もしChatGPTが何かを間違えた場合、それを削除したり編集したりすることができるからです。

So, I think being able to visualize what this AI system knows about you is very, very good because it allows you to see if it's working correctly, if it manages to remember certain things based on certain conversations or if it manages to even infer certain things from your conversations.

AIシステムが自分について何を知っているかを視覚化できることは非常に良いことだと思います。それによって、それが正しく機能しているか、特定の会話に基づいて特定のことを覚えることができるか、さらには会話から特定のことを推測することができるかを確認することができます。

And I think that would be an interesting concept as well.

それも興味深いコンセプトだと思います。

Now, if you're confused about this memory thing and you're thinking, Wait a minute, I've seen this before.

もしもこのメモリのことで混乱して、「待って、これを以前見たことがある」と思っているなら。

I've actually seen the memory thing before.

実際、メモリのことを以前に見たことがあります。

This is because likely what you've seen is an earlier iteration of this screenshot.

これはおそらく、OpenAIのこのメモリ管理システムの以前のバージョンを見たことがあるからです。

So, I have this tweet from Tibor Blaho and this was a SC a video recording of what we saw from an early iteration from OpenAI in terms of this memory management system.

OpenAIのこのメモリ管理システムの初期バージョンに関するビデオ録画を、ティボー・ブラホのツイートで入手しました。

I'm not sure they're using this at the same time now, but in November we did get to see something like this and you can see that there are several things that the user is currently able to do from this system.

今は同時にこれを使用しているかはわかりませんが、11月にはこのようなものを見ることができました。このシステムからユーザーが現在行えるいくつかのことがわかります。

Now, like I said, I think I did have this for a moment, but it has now disappeared and we're now starting to see this different user interface rolled out.

私も一時的にこれを持っていましたが、今は消えてしまい、異なるユーザーインターフェースが展開されています。

But of course, you can see this personalization area has been changed a little bit.

ただし、この個人化エリアは少し変更されていることがわかります。

And I think this is why we're still getting delays on the next model, whether it be GPT-4.5 or GPT-5 because certain things like memory that we currently are seeing on screen now are currently being just weighted out and we're seeing what is, of course, the most effective method of doing this because there are many different things that we can do with the system.

そして、私たちはまだ次のモデル(GPT-4.5またはGPT-5)についての遅延を受けているのは、現在画面上で見ているメモリのような特定の要素がまだ検討中であり、最も効果的な方法が何かを見極めているからだと思います。なぜなら、私たちがシステムでできることはさまざまだからです。

So, I'm guessing they're waiting to see what is the more effective.

ですので、彼らはより効果的な方法を見極めるために待っていると思います。

And of course, you can see right here the temporary chat not in history blank state and no model training is something that was there before.

そして、ここでは、一時的なチャットの履歴の空白状態とモデルのトレーニングがないことが表示されていますが、これは以前から存在していたものです。

Now, this was kind of fascinating because even before this leak, there was essentially someone that just viewed the source code of the web page.

これは非常に興味深いことでした。このリークの前にも、ウェブページのソースコードを表示しただけで、これを見ることができました。

And then, we were essentially able to see that.

そして、私たちは実質的にそれを見ることができました。

And I think that that is kind of interesting.

そして、それはなかなか興味深いと思います。

That ChatGPT and OpenAI do kind of add these things in.

ChatGPTとOpenAIは、これらの要素をいくつか追加するようです。

But one thing that I do want to show you guys before I continue talking about all of these new memory updates is this interview.

しかし、新しいメモリのアップデートについて話し続ける前に、皆さんに見ていただきたいものがあります。それはこのインタビューです。

Because in this interview, and you may have seen this interview, but there is a specific clip in which Sam Altman talks about the next two years of ChatGPT.

なぜなら、このインタビューでは、おそらくご覧になったかもしれませんが、サム・アルトマンがChatGPTの次の2年について話している特定のクリップがあります。

And the reason this video is really important is because this shows us where we're going for ChatGPT.

このビデオは、ChatGPTの進む方向を示しているため、非常に重要です。

And at the time, I didn't connect the dots in this interview that he did with Bill Gates.

そして、その時点では、私はこのインタビューでの彼とビル・ゲイツのやり取りを結びつけていませんでした。

But now it makes more sense because in this interview, he basically says that look, we're leading towards personalization where it's going to remember stuff and be much more customizable.

しかし、このインタビューでは、彼は基本的に言っているので、私たちは個人化に向かって進んでおり、それはもっとカスタマイズ可能で、ものを覚えるようになるということがより理解できるようになりました。

And then of course, we do have things like personalization being rolled out.

そしてもちろん、個人化のようなものも展開されています。

So it's clear that although the name might not be changed from GPT-4 to GPT-4.5 just yet, it's clear that smaller and smaller iterations are being loaded onto this system as a graduate update.

ですので、まだGPT-4からGPT-4.5に名前が変わるわけではありませんが、少しずつ小さな改良がこのシステムにロードされていることは明らかです。

But take a listen to this clip because I think it's really important for us to listen to Sam.

でも、このクリップを聞いてみてください。なぜなら、私たちはサムの話を聞くことが本当に重要だと思うからです。

And then of course, I will show you guys some more features.

そしてもちろん、もっと多くの機能を皆さんにお見せします。

You know, when you look at the next two years, what do you think some of the key milestones will be?

次の2年を見ると、どのような重要なマイルストーンがあると思いますか?

Multimodality will definitely be important. We start...

マルチモダリティは間違いなく重要になるでしょう。私たちは…。

Which means speech in, speech out?

それは言葉入力、言葉出力を意味しますか?

Speech in, speech out, images eventually, video clearly, people really want that.

言葉を入力し、言葉を出力し、最終的には画像、はっきりとしたビデオ、人々は本当にそれを望んでいます。

We launched images and audio, and it had a much stronger response than we expected.

私たちは画像と音声をリリースしましたが、それには予想以上の反応がありました。

We'll be able to push that much further.

私たちはそれをさらに進めることができるでしょう。

But maybe the most important areas of progress will be around reasoning ability.

でも、おそらく最も重要な進歩の領域は、推論能力の周りになるでしょう。

Right now, GPT-4 can reason in only extremely limited ways.

現時点では、GPT-4は非常に限られた方法で推論することができます。

And also reliability.

そして信頼性もです。

You know, if you ask GPT-4 most questions 10,000 times, one of those 10,000 is probably pretty good.

もしGPT-4に同じ質問を1万回しても、その中の1回はおそらくかなり良い回答が得られるでしょう。

But it doesn't always know which one.

でも、それがいつもどれかはわかりません。

And you'd like to get the best response of 10,000 each time.

毎回最高の回答を10,000回中から得たいと思います。

So that'll be that increase in reliability will be important.

それによって信頼性が向上するのは重要なことです。

Customizability and personalization will also be very important.

カスタマイズ性と個人化も非常に重要になります。

People want very different things out of GPT-4.

人々はGPT-4から非常に異なるものを求めています。

Different styles, different sets of assumptions.

異なるスタイル、異なる仮定のセット。

We'll make all that possible.

私たちはそれをすべて可能にします。

And then, also, the ability to have it use your own data.

そして、また、自分自身のデータを使用する能力もあります。

So, the ability to know about you, your email, your calendar, how you like appointments booked, connected to other outside data sources all of that, those will be some of the most important.

それで、あなたについて知る能力、あなたのメール、あなたのカレンダー、アポイントメントの予約方法、他の外部データソースに接続する能力、すべてそれらは最も重要なものの一部になるでしょう。

So you can see here that that is pretty much exactly what we're seeing.

ここで見ることができるのは、ほぼまさに私たちが見ていることです。

And I think now that we've seen that the personalization is being changed and it's being upgraded to having your memory being managed.

そして、今私たちが見ているように、パーソナライゼーションが変更され、メモリが管理されるようにアップグレードされていると思います。

And of course, you say these very simple commands and it's able to remember them.

そしてもちろん、非常にシンプルなコマンドを言うと、それを覚えることができます。

This shows us that OpenAI are moving very quickly.

これはOpenAIが非常に速く進んでいることを示しています。

Now, one thing that you might need to understand from the new releases from OpenAI is that I do think that these updates and these changes need to be paid attention to.

さて、OpenAIの新しいリリースから理解する必要があることの一つは、私はこれらのアップデートと変更に注意を払う必要があると思うということです。

Because what most people are doing with OpenAI and their services and in terms of looking at the updates is that people will wait for an entirely new model to be released.

なぜなら、多くの人々がOpenAIとそのサービス、そしてアップデートを見るときに、完全に新しいモデルのリリースを待つ傾向があるからです。

But I think that what OpenAI are doing is very smart.

でも、私はOpenAIが非常に賢いことをしていると思います。

Because they know that if they release GPT-5 and it's a shocker in terms of capability, it's going to cause some, I guess you could say, potentially some societal unrest.

なぜなら、彼らはGPT-5をリリースして、能力が驚くほど低いということがあれば、社会的な不安を引き起こす可能性があるからです。

But what Sam Altman has talked about was the fact that they need to release things very, very gradually.

でも、サム・アルトマンが話したことは、彼らは非常に非常に徐々にリリースする必要があるということです。

And of course, adding things like managing your memory and the personalization is a very good way to do that.

そしてもちろん、メモリの管理や個人化のような要素を追加することは、それを実現する非常に良い方法です。

So this is where Sam Altman talks about the fact that we're going to be basically gradual kind of updates.

ですので、ここでサム・アルトマンが基本的に徐々にアップデートしていくと話しているのです。

One of the things that we believe very deeply is that society and this technology have to co-evolve.

私たちが非常に深く信じていることの一つは、社会とこの技術が共進化しなければならないということです。

We believe in iterative deployment a lot for the obvious reason, which is that people need time to gradually update and think and figure out what the rules should be.

私たちは明らかな理由から、反復的な展開を非常に重要視しています。それは、人々が徐々に更新し、考え、ルールをどのようにすべきかを考える時間が必要だからです。

But there's another part which is it's it that you can't separate the technology from the world.

でも、もう一つ重要な点があります。それは、技術を世界から切り離すことはできないということです。

You can't just, even if you get everything magically right, you can't build it in secret and then put it in the world all at once because the world is going to keep changing with each iteration.

たとえすべてを魔法のように正しくしても、それを秘密裏に構築して一度に世界に投入することはできません。なぜなら、世界は各イテレーションごとに変化し続けるからです。

Which means on those middle cases, I don't know what the right answer is yet, nor does anyone because no one has really like thought through and seen how the institutions and the world and society shift and reshape in response to this.

それはつまり、中間的なケースでは、まだ正しい答えがわからないし、誰もわからないということです。なぜなら、誰もがこの問題について考え抜いて、機関や世界、社会がどのように変化し再構築されるかを見ていないからです。

So there will be a lot of things that we'll have to start slowly on and iterate as we go.

だから、私たちはゆっくりと始めて、進行しながら改善していく必要があります。

And there will be a lot of middle cases.

そして、中間的なケースがたくさんあるでしょう。

But we do want to support the US government and other governments too.

しかし、私たちはアメリカ政府や他の政府をサポートしたいと思っています。

And like I find...

そして、私は...

You can see that this is why we know that the iterative deployment is going to be a thing.

これが、反復的な展開が重要である理由を理解できると思います。

So one thing that I'm going to be doing now is I'm going to be making sure that I do take a look at every single thing that is going on with GPT in terms of these small updates.

今後、私が行うことの一つは、これらの小さなアップデートに関してGPTのすべての出来事を注意深く確認することです。

So one way that I'm going to be doing that is I'm going to be making sure that I scour Twitter and Reddit to see if there are any instances of new things just slowly being rolled out.

それをする方法の一つは、TwitterやRedditをくまなく調べて、新しいことが徐々に展開されているかどうかを見ることです。

Because this stuff is fascinating in terms of it shows us the entire direction that we're moving.

なぜなら、これらのことは私たちが進んでいる方向性を示してくれるからです。

Because like I said before, it was very evident that the track GPT system was just a base test for a very larger AI system and ecosystem that we're going to be moving into.

なぜなら、前にも言ったように、GPTシステムは非常に大きなAIシステムとエコシステムのベーステストであることが明らかでした。そして、私たちはそれに移行していくつもりです。

So I think using the GPT-4 system as it is in the future is going to be a thing of the past.

だから、将来的にはGPT-4システムをそのまま使うことは過去のことになると思います。

And in fact, we're going to have completely personalized versions.

実際、私たちは完全に個別のバージョンを持つことになるでしょう。

And that's what we've had a little bit with this GPT system.

そして、これはこのGPTシステムでも少し経験しました。

And I think this is also going to be, quite similar to MemGPT, which is essentially a personalized AI with infinite context windows.

そして、私はこれがMemGPTにもかなり似ていると思います。それは本質的には無限のコンテキストウィンドウを持つ個別のAIです。

And essentially, MemGPT was something that was devised to enhance the performance of Large Language Models by introducing a more advanced management scheme in terms of the memory, hoping to overcome the challenges posed by fixed context windows.

そして、実際には、MemGPTは、固定されたコンテキストウィンドウによって引き起こされる課題に対処するために、より高度なメモリ管理スキームを導入することで、大規模言語モデルのパフォーマンスを向上させるために考案されたものです。

And essentially, it was basically giving the AI system the ability to manage its own memory, having a virtual context management, and having Interruption handling.

そして、実質的には、AIシステムに自己のメモリを管理させ、仮想的なコンテキスト管理と中断処理を持たせることでした。

So I think something like this has inspired the next wave.

だから、私はこのようなものが次の波に影響を与えたと思います。

And of course, like I said before, with how systems are revolving and the kind of reception that we did see rabbit r1 device get, I think we all know that personalized AI systems, if they are done correctly, will be even better than a simple chatbot that we can just talk to.

そしてもちろん、前にも言ったように、システムがどのように進化しているか、rabbit r1デバイスがどのような反応を得たかを考えると、個別のAIシステムは、正しく行われれば、ただ話すだけの単純なチャットボットよりもさらに優れていることを皆さんも知っていると思います。

And then, just get a simple response out from.

そして、単純な応答を得るだけです。

So now, if you do want to be able to test out these features when they do go live, what you want to be able to make sure is number one, that you are subscribed to ChatGPT's plus tier.

今、もしもこれらの機能を実際にテストしたい場合は、まずChatGPTのプラスティアに加入していることを確認したいですね。

So, you want to make sure that you upgrade your plan to the $20 a month tier because that is the only tier where you do get access to experimental features that may change.

ですので、月額20ドルのティアにアップグレードすることを確認してください。なぜなら、それが実験的な変更が行われる可能性のある機能にアクセスできる唯一のティアだからです。

For example, if you then go into your settings and then you go to Beta features, you'll see this tab.

例えば、設定に進み、ベータ機能に移動すると、このタブが表示されます。

And essentially, this tab says that as a plus user, enjoy Early Access to experimental features, experimental new features which may change during development.

基本的に、このタブには、プラスユーザーとして、開発中に変更される可能性のある実験的な新機能に早期アクセスを楽しむということが書かれています。

And I think that's what we're seeing.

そして、私はそれが私たちが見ているものだと思います。

So, I likely think that the roadmap for OpenAI is to simply test out all of this stuff.

だから、OpenAIのロードマップは、単にすべてのこれらのことをテストすることだと思います。

And I guess to add a more balanced output in terms of the memory and in terms of the personalization to test these things before we do get a future release, whether it be GPT-4.5 or whether it be GPT-5.

そして、私はメモリと個別化のバランスの取れた出力を追加し、これらのことをテストしてから、将来のリリース、GPT-4.5またはGPT-5を得ることだと思います。

I think that that memory aspect is where we're going to be heading towards overall in this niche.

私はこのニッチ全体で、このようなメモリの側面が進む方向だと思います。

Now, for those of you who are familiar with MemGPT, you'll understand and know how the system works.

今、MemGPTに詳しい方は、このシステムがどのように機能するかを理解しているでしょう。

But for those of you who aren't familiar with this new style of update that we're going to be getting with ChatGPT, I think it's important for me to cover some of the demonstrations of how this system updates its memory and how it changes its interactions with the user based on this memory GPT paper.

しかし、新しいアップデートのスタイルに慣れていない方々のために、ChatGPTが提供するこの新しいアップデートについて、このメモリGPT論文に基づいてシステムがどのようにメモリを更新し、ユーザーとの対話をどのように変更するかについてのデモンストレーションをいくつか紹介することが重要だと思います。

So, in the paper, we can see that MemGPT says, Hello Chad, welcome. I'm excited to embark on this journey with you. As a PhD student in computer science, I can only imagine the fascinating ideas you're juggling. Can you share what you're currently working on? I'm all ears.

そのため、論文では、MemGPTは「こんにちは、チャド、ようこそ。私はあなたとのこの旅に興奮しています。コンピュータサイエンスの博士課程の学生として、あなたが取り組んでいる魅力的なアイデアを想像することしかできません。現在取り組んでいることを共有していただけますか?私は耳を傾けます。」と言っています。

And then, you can see, I took the day off today.

そして、それから、私たちは、今日は休みを取りました。

My mom Brenda baked me a birthday cake.

私の母ブレンダが誕生日ケーキを焼いてくれました。

It was my favorite, chocolate lava.

それは私のお気に入りのチョコレートラバケーキでした。

And then, you can see it updates its context to say, Birthday 11th October, its favorite cake is chocolate lava made by its mom and Brenda.

そして、それから、それはコンテキストを更新して、誕生日が10月11日であり、お気に入りのケーキはチョコレートラバで、母親の名前はブレンダであることを示しています。

And of course, you can see that it it just continues on the conversation, trying to get out some small information.

そしてもちろん、会話は続けられ、少しの情報を提供しようとしています。

And I think this is likely where these AI systems are going to go in terms of being much more useful.

そして、私はこれがこれらのAIシステムがより役立つようになるであろう場所だと思います。

Because the very big problem that we do have with these AI systems is that every time we have to get a conversation out, we have to tailor it specifically to our user interest.

なぜなら、私たちが現在のシステムから最大の出力を得るための唯一の制限は、人々が適切にプロンプトする方法を知らないこと、そしてそれに適切な情報を与えて、成功するために必要なコンテキストを提供する方法を知らないことだと思います。

And that is going to make these AI systems so much more powerful.

それによって、これらのAIシステムは非常に強力になるでしょう。

Because the only limitation that I've truly seen from these models, in terms of getting the maximum output that you can from these current systems, is the fact that people don't know how to prompt them properly.

なぜなら、これらのモデルにおいて、現在のシステムから最大の出力を得るための唯一の制限は、人々が適切にプロンプトする方法を知らないという事実です。

And they don't know how to give it all of the correct information that gives it the context that will provide you with the type of information that will help you succeed and whatever it is whatever kind of task you're doing.

そして、人々はそれに適切な情報をすべて与える方法を知らないため、それには文脈を与える正しい情報をすべて与えることができず、成功に役立つ情報の種類を提供することができません。

And then, of course, you can see here it says, Hello Chad, it's a pleasure to finally have a conversation with you.

そして、もちろん、ここには「こんにちは、チャドさん、お話できて嬉しいです」と書かれています。

I understand that you're studying computer science and have a keen interest in dynamic areas like Formula 1 and sailing.

あなたがコンピュータサイエンスを勉強しており、フォーミュラ1やセーリングなどのダイナミックな分野に興味を持っていることを理解しています。

What draws you to those interests?

それらの興味を引かれる理由は何ですか?

And it says, Speed, of course, and thrill and adrenaline.

そして、「もちろん、スピードやスリル、アドレナリンです」と書かれています。

And then here, he says, The conversation history will soon reach its maximum length and be trimmed. Make sure to save any important information from the conversation to your memory.

そして、ここでは、「会話履歴はまもなく最大の長さに達し、トリミングされます。会話から重要な情報をメモリに保存してください」と言っています。

And then, of course, it actually manages to do that.

そして、もちろん、それを実際に行うことができます。

So it says, Key personality trait: enjoys high-speeded adrenaline rush activities, likes Formula 1 racing, and intense gaming sessions in CS:GO.

そして、ここでは、「主要な性格の特徴:高速のアドレナリンラッシュアクティビティを楽しむ、フォーミュラ1レースやCS:GOでの激しいゲームセッションが好き」と書かれています。

And another thing you can see here is that this is how it actually updates.

そして、ここでもう一つ見ることができるのは、これが実際にどのように更新されるかということです。

So, for example, it says, Welcome back, Bia. Let's chat about horror movies.

例えば、「おかえりなさい、ビアさん。ホラー映画についておしゃべりしましょう」と書かれています。

And he says, Actually, I don't like horror movies. I'm more into romantic comedy.

そして、彼は「実は、私はホラー映画が好きではありません。もっとロマンティックコメディが好きです」と言っています。

And then the system essentially says, It's going to update its memory, saying,I watch horror movies' replaced this section withI like romantic movies. So the system is able to learn about how you change over time.

そして、システムは本質的には、「私はホラー映画を見る」という部分を「私はロマンチックな映画が好きです」と置き換えることで、その記憶を更新すると言っています。

And it says, You know, apologies for the mix-up earlier, yada yada yada.

そして、それは「前回の混乱について申し訳ありません、などなどなど」と書かれています。

So you can see here that MemGPT is essentially conversation-like agents, our virtual companions, and personalized assistants aim to engage users in natural, long-term interactions, potentially spanning weeks, months, or even years.

ここでは、MemGPTは本質的には会話のようなエージェントであり、仮想の相棒や個人のアシスタントであり、自然な長期的な対話に従事し、数週間、数ヶ月、または数年にわたる可能性があります。

And I think this kind of paper really goes to show us what we're going to be getting from these future models.

そして、私はこのような論文が私たちにこれからのモデルから得られるものを示していると思います。

And I think everyone is overlooking this personalization feature because it's not some big announcement.

そして、私はみんながこの個人化の特徴を見落としていると思います。なぜなら、それは大きな発表ではないからです。

But I think GPT and OpenAI know that while the GPT store is very good, personalization is just something that you can't get away from because the personalized nature of our friends is what makes the connection so good.

しかし、私はGPTとOpenAIが、GPTストアは非常に良いものの、個人化はついつい避けられないものであり、友人の個人的な性格がつながりを良くする要素であることを知っていると思います。

Because we have all these unique shared experiences and memories.

なぜなら、私たちはすべてこれら独自の共有体験や思い出を持っているからです。

And a personalized AI system that's able to get deeply ingrained in terms of remembering key details about your life, like your birthday, is going to be something that is far more valuable to you.

そして、あなたの誕生日のようなあなたの人生に関する重要な詳細を深く覚えている個人化されたAIシステムは、あなたにとってはより価値のあるものになるでしょう。


この記事が気に入ったらサポートをしてみませんか?