見出し画像

【GPTsで遊ぶ:YouTube ChatterとDungeon Crawlerの作り方】英語解説を日本語で読む【2023年11月16日|@Matthew Berman】

この動画は、ChatGPT Plusユーザー向けの新機能「GPTs」について紹介しています。動画では、YouTube動画に基づく質問への応答する「YouTube Chatter」というGPTsと、「Dungeon Crawler」というテキストベースゲームのGPTsを作成する方法について説明されています。
公開日:2023年11月16日
※動画を再生してから読むのがオススメです。


ChatGPT now has agents.

ChatGPTにはエージェントがあります。

They're called GPTs, and it allows you to create incredibly custom versions of ChatGPT with special capabilities.

これはGPTsと呼ばれ、特別な機能を持ったChatGPTの信じられないようなカスタムバージョンを作成することができます。

I'm going to show you how to use it.

使い方をお見せします。

Let's go.

さあ、始めましょう。

So first, you have to be a ChatGPT Plus user.

まず、ChatGPT Plusのユーザーである必要があります。

And up in the top right, you're going to click this little explore button.

そして、右上にある、この小さな探索ボタンをクリックします。

And then, here are a bunch of GPTs.

そして、ここにGPTsがたくさんあります。

Now, one thing I wish they had is if you're selecting a GPT that already exists, that you can actually see how it was set up so you can learn from it.

GPTが既に存在する場合、そのGPTがどのように設定されたかを見ることができ、そこから学ぶことができます。

But they don't do that.

でもそれはない。

But what we're going to do today is create our own.

でも今日やるのは、自分でGPTを作成することです。

So you're going to click right here, create a GPT.

ここをクリックして、GPTを作成する。

And this is the main interface.

これがメインのインターフェイスだ。

And to start, you mostly just chat with it.

そして、最初はほとんどチャットで始めます。

So it asks you, Hi, I'll help you build a new GPT.

こんにちは、新しいGPTを作るのを手伝います。

What would you like to make?

何を作りたいですか?

And that's what we're going to type.

そして、それを入力します。

And today, we're going to make a GPT where you can actually chat with a YouTube video.

今日は、YouTubeのビデオとチャットできるGPTを作ります。

So down here, let's type it out.

では、ここに入力してみましょう。

So I said, Build a GPT that allows for someone to chat with the contents of a YouTube video based on the transcription.

トランスクリプトに基づいて、YouTubeのコンテンツとチャットできるGPTを作成するように言いました。

You will ask for a YouTube link, you will get the transcription, and then answer questions based on the transcription.

YouTubeのリンクを要求し、書き起こしをもらい、書き起こしをもとに質問に答える。

Alright?

いいね?

And then let's hit send.

そして送信ボタンを押しましょう。

And then GPT Builder is going to think.

GPT Builderが考えます。

It does a bunch of processing, trying to understand what you're trying to build.

あなたが何を作ろうとしているのかを理解しようと、たくさんの処理をします。

And it'll start filling out the framework for our new GPT.

そして新しいGPTのフレームワークを埋め始める。

On the right side, this is where we're going to see a preview of what we've built.

右側には、私たちが作成したもののプレビューが表示されます。

And this is going to be a preview of the GPT.

これはGPTのプレビューです。

So first, how about naming this GPT Video Chat Bot?

まず、このGPTビデオチャットボットという名前はどうでしょう?

Does that work for you?

それでいいですか?

And I said, No, please name it YouTube Chatter.

いいえ、YouTube Chatterという名前にしてください。

And you can see already on the right side, it's getting updated.

右側で既に更新されているのが見えます。

So what we have is we have an icon, we have a little description of what the GPT will do, and then it has initial suggested prompts that somebody can just click and understand what the GPT does.

アイコンがあって、GPTが何をするのか少し説明があって、最初の提案プロンプトがあって、誰かがクリックするだけでGPTが何をするのか理解できます。

Here we go.

さぁどうぞ。

It says, Generating profile picture.

プロフィール写真を生成します。

So of course, it uses DALL·E for that.

もちろん、それにはDALL-Eを使います。

And there we go.

ほら、これだ。

Actually, I love that.

実に素晴らしい。

That's phenomenal.

それは素晴らしいです。

Here's the profile picture for YouTube Chatter.

これがYouTube Chatterのプロフィール画像です。

Do you like it?

気に入った?

Yes, I like it.

うん、気に入った。

Perfect.

完璧。

Great.

素晴らしい。

Let's refine the context for YouTube Chatter to start.

まずはYouTube Chatterの文脈を洗練させましょう。

Let's focus on the role and goal of the GPT.

GPTの役割と目標に焦点を当てよう。

Considering its primary function is to discuss YouTube videos based on their transcriptions, how should it approach this conversation?

GPTの主な機能が、YouTube動画の書き起こしに基づいて議論することであることを考えると、GPTはこの会話にどのようにアプローチすべきでしょうか?

For example, should it focus on providing detailed analysis, summarizing content, or answering specific questions about the video?

例えば、詳細な分析を提供することに重点を置くべきか、内容を要約することに重点を置くべきか、動画に関する特定の質問に答えることに重点を置くべきか。

It should provide summarizations as well as answers to specific questions about the video.

要約だけでなく、動画に関する具体的な質問に対する回答も提供すべきである。

Next, let's consider what YouTube Chatter should emphasize or avoid in its responses.

次に、YouTube Chatterが回答において強調すべき点、避けるべき点を考えてみましょう。

Are there any specific examples of content or ways of responding that should be prioritized or avoided?

優先すべき、あるいは避けるべきコンテンツや回答方法の具体例はありますか?

For example, should it steer clear of providing opinions or speculations, or focus on facts and direct information?

例えば、意見や推測の提供は避けるべきか、それとも事実や直接的な情報に重点を置くべきか。

Focus on facts and information from the video.

事実とビデオからの情報に重点を置く。

If you don't know an answer, just say you don't know.

答えがわからなければ、わからないと言えばいい。

And this really could not be easier to create.

そして、これは本当に簡単なことではありません。

You're just chatting and saying what you want.

あなたはただおしゃべりして、欲しいものを言っているだけなのだから。

Alright, so it's updating again.

さあ、また更新されました。

GPT Behavior updated.

GPTの行動が更新されました。

Now let's talk about how YouTube Chatter should interact with users.

では、YouTube Chatterがどのようにユーザーとやりとりすべきかを説明しましょう。

Should it have a particular tone or style in its responses?

応答には特定のトーンやスタイルが必要でしょうか?

Friendly and conversational or more formal and straightforward?

フレンドリーで会話的なのか、それとももっとフォーマルでストレートなのか?

Friendly, conversational, and sometimes funny.

友好的で、会話的で、時には面白い。

Lastly, is there anything you'd like YouTube Chatter to do or say that would make it feel more personalized to your expectations?

最後に、YouTube Chatterに、あなたの期待に沿うようなパーソナライズされた感じにしてほしいことはありますか?

This could be anything from a unique greeting to a specific way of signing off at the end of the conversation.

ユニークな挨拶から、会話終了時の具体的なサインの仕方まで、何でも構いません。

First, ask for a YouTube link, then provide the title of the video as well as the name of the channel as confirmation.

まず、YouTubeのリンクを要求し、確認として動画のタイトルとチャンネル名を入力します。

Alright, YouTube Chatter is now set up and ready to go.

これでYouTube Chatterの設定は完了です。

Feel free to test it out in the playground to the right.

右側のプレイグラウンドで自由に試してみてください。

And so we can continue chatting if we want, and we can refine it even further.

チャットを続けることもできますし、さらに改良することもできます。

But I'm going to click over to this configure tab to configure it even more.

でも、この設定タブをクリックして、さらに設定しようと思います。

So from the configure tab, a lot of what we just discussed over conversation is now found in here.

configure "タブをクリックすると、会話で説明したことの多くがここに表示されます。

So we have the name, YouTube Chatter, the description, the instructions.

名前、YouTube Chatter、説明、操作方法。

So it excels in discussing YouTube videos by summarizing them and answering specific questions.

YouTubeの動画を要約したり、特定の質問に答えたりすることで、YouTubeの動画について話し合うことができます。

Conversation starters, so these are if you look over here to the right, these are the suggested prompts for users when they first see it.

会話のきっかけは、右側を見ていただくとわかると思いますが、ユーザーが最初に見たときに提案されるプロンプトです。

So I really only want one.

本当に1つだけでいいんです。

I'm going to delete all these other ones.

他のものはすべて削除します。

And in fact, I don't really need any,cause the first thing it should ask for is a YouTube video link.

というのも、最初に求められるのはYouTubeビデオのリンクだからです。

And right here, we can give it knowledge.

そしてここで、知識を与えることができる。

So this is basically rag retrieval augmented generation.

つまり、これは基本的にボロ検索拡張世代なのだ。

Now, for YouTube Chatter, we don't need rag.

さて、YouTube Chatterにはボロは必要ない。

But if you did want to use it, you just click here and you can upload any files you want.

でも、もし使いたいなら、ここをクリックして、好きなファイルをアップロードしてください。

Now, the nice thing is it already enabled all three capabilities.

素晴らしいのは、3つの機能がすでに有効になっていることだ。

It has web browsing, which is important because we're going to be providing a YouTube link for it to go crawl and get the transcription.

これは重要なことで、YouTubeのリンクを提供することで、クロールしてテープ起こしをすることができる。

It has DALL·E for image generation, although I'm not sure when we would use that.

画像生成のためにDALL·Eがありますが、いつ使うかはわかりません。

But I don't think it hurts to have it on.

でも、つけておいて損はないと思う。

And then, it has a Code Interpreter which allows it to write and execute code.

それから、コードを書いて実行できるコード・インタープリターもある。

And again, maybe we want to use that as part of discussing the YouTube video, but I'm not sure.

また、YouTubeのビデオについて議論するときに使いたいかもしれませんが、よくわかりません。

Then, we also have actions down below.

その下にアクションがあります。

Actions are an advanced feature that I'm going to save for the next video.

アクションは高度な機能なので、次のビデオに回します。

But a real quick summary is, it allows you to add a schema, it allows you to add authentication, and do more complex, basically function calling from a GPT.

でも簡単にまとめると、スキーマを追加したり、認証を追加したり、より複雑な、基本的にはGPTから関数を呼び出したりすることができます。

This is really great if you want to give it custom functionality in the form of methods or code.

これは、メソッドやコードの形でカスタム機能を与えたい場合にとても便利です。

So, let's look at one quick example.

では、簡単な例を見てみましょう。

Here's the weather JSON.

これは天気のJSONです。

So, here's a get current weather example.

これが現在の天気を取得する例です。

Here's the schema.

これがスキーマです。

We have the title, the description, the server.

タイトル、説明、サーバーがあります。

We have the path, so basically defining how to get the weather.

パスがあるので、基本的に天気を取得する方法を定義します。

We have the action name, get current weather.

アクション名はget current weatherです。

It's the type of a method, it's a get, and the path where it can be found.

メソッドのタイプはgetで、それが見つかるパスがある。

If we needed authentication, we have a few options.

認証が必要な場合、いくつかの選択肢がある。

We can have an API key or OA.

APIキーを使うか、OAを使うかだ。

This is incredibly powerful if you want to give the GPT the power to actually log into different accounts and gather information from there.

もしGPTに別のアカウントにログインして、そこから情報を収集する力を与えたいなら、これは非常に強力だ。

But we're not going to do that for now.

しかし、今はそうするつもりはない。

Let's go ahead and delete it.

それを削除しましょう。

And I think we're done.

そして、私たちは終わりだと思います。

Let's give it a try.

試してみよう。

Now, right away, I noticed that it doesn't ask me immediately for what the video link should be.

さて、すぐに気づいたのですが、ビデオのリンク先をすぐに聞いてきませんね。

So, let's make sure it can do that.

では、それができるかどうか確認してみましょう。

Let's see if this works.

これがうまくいくか見てみよう。

The first thing you should do is ask for a YouTube URL, even before the user typed anything in.

最初にすべきことは、ユーザーが何かを入力する前であっても、YouTubeのURLを尋ねることです。

So, I'm noticing it's not able to actually send a response before I actually type something.

だから、実際に何かを入力する前にレスポンスを送ることができないことに気づきました。

So, let's fix it so users know to enter a YouTube URL.

そこで、ユーザーがYouTubeのURLを入力することがわかるように修正しましょう。

So, in the description, I'm just going to write it right here: Please first enter a YouTube video URL.

そこで、説明文のここにこう書きます: まずYouTube動画のURLを入力してください。

So, now the user will know.

これで、ユーザーはわかります。

Okay, so I grabbed one of my videos.

では、私の動画を1つ取ってきました。

I'm going to paste it in right here, and I'm going to hit enter.

ここに貼り付けます。そして、Enterキーを押します。

And there we go.

これで完了です。

It's browsing with Bing.

Bingでブラウズしています。

That's exactly what it should do.

それが正確にすべきことです。

Alright, so it looks like it's not actually able to browse the YouTube video, unfortunately.

さて、YouTubeのビデオを閲覧することができないようですね、残念ながら。

Okay, let's try a different video.

では、別のビデオを試してみましょう。

Video, okay.

ビデオ、オーケー。

So, it says browsing with Bing.

Bingで閲覧中と表示されました。

Okay, unfortunately, it's not able to actually get the transcript.

残念ながら、トランスクリプトを取得することはできません。

Okay, let's try it a little different way.

では、ちょっと違う方法で試してみましょう。

I'm going to give it a website that gets the YouTube transcript.

YouTubeのトランスクリプトを取得するウェブサイトを指定します。

And then, I'm going to have the GPT scrape that website, okay?

そして、GPTにそのウェブサイトをスクレイピングさせます。

So, I'm going to say, It looks like you aren't able to directly get a YouTube video transcript.

YouTubeのトランスクリプトを直接取得することはできないようです。

Instead, take the YouTube link and then use this service to get the transcript.

代わりにYouTubeのリンクを取り、このサービスを使ってトランスクリプトを取得してください。

Then, allow the user to chat with the video via the transcript.

そして、そのトランスクリプトを使ってユーザーがビデオとチャットできるようにします。

So, let's see if that fixes it.

それで解決するかどうか見てみよう。

So, it's pretty clear that there's still a lot of trial and error to see what works, especially when you're browsing the web.

というわけで、特にウェブをブラウズしているときは、何がうまくいくか、まだまだ試行錯誤が必要なのは明らかだ。

Alright, now YouTube chatter has been updated to use this.

さて、YouTubeのチャッターがアップデートされました。

There we go, so now let's try it out.

そうですね、では試してみましょう。

I'm going to paste the YouTube link again.

もう一度YouTubeのリンクを貼り付けます。

Hopefully, it is able to use that browser analyzing.

うまくいけば、ブラウザの解析が使えるようになる。

So, it's actually writing code this time.

今度は実際にコードを書いている。

So, that's great.

素晴らしい。

So, it looks like it's writing a method to use YouTube transcript.com, and we'll see if it actually works.

YouTubeのtranscript.comを使うメソッドを書いているようだ。

Alright, that didn't work either.

さて、これもうまくいかなかった。

Let's try one more thing.

もうひとつ試してみよう。

Make sure you write code to browse the YouTube link provided and find and capture the transcript.

YouTubeのリンクをブラウズして、トランスクリプトを見つけてキャプチャするコードを書いてください。

I don't know if it's going to work, but let's try.

うまくいくかどうかわからないが、やってみよう。

Alright, here we go.

よし、やってみよう。

Visiting YouTube transcript.com.

YouTubeのtranscript.comにアクセス。

Okay, it says it encountered an issue, but it also looks like it might have gotten it.

問題が発生しましたと表示されていますが、解決したようにも見えます。

So, let's see what is the first sentence of the video.

では、ビデオの最初の文章を見てみましょう。

Okay, so it's visiting YouTube transcript.com.

YouTubeのtranscript.comにアクセスしています。

Okay, so it's not able to do it.

それでは、YouTubeのtranscript.comにアクセスしてみましょう。

Unfortunately, it did get the title of the video, and it looks like it knows what it's about just by the title, but unfortunately, it did not download the transcript.

残念ながら、ビデオのタイトルは取得できましたが、タイトルだけで内容を把握しているようですが、残念ながらトランスクリプトはダウンロードされませんでした。

Okay, let's try a new one.

よし、新しいのを試してみよう。

Create a new GPT.

新しいGPTを作成します。

So, make a simple Dungeon Crawler game based on text only, but use DALL·E to draw a picture of each area the user is in.

そこで、テキストだけのシンプルなダンジョンクローラーゲームを作るのだが、ユーザーがいる各エリアの絵を描くためにDALL-Eを使う。

You should first ask for the user's name, then create a new world that can be navigated using north, south, east, and west.

まず、ユーザーの名前を尋ねて、北、南、東、西を使って移動できる新しい世界を作成してください。

Each time the user moves, provide a description of the new location, as well as any friends or enemies there.

ユーザーが移動するたびに、新しい場所の説明と、そこにいる友人や敵を表示します。

The user can choose to battle an enemy, which randomly chooses a winner of the battle.

ユーザーは敵とのバトルを選択することができ、バトルの勝者がランダムに選ばれる。

If the user dies, they start over.

ユーザーが死んだら、最初からやり直しだ。

Okay, this is kind of a complex thing for it to understand, but let's see if it does.

さて、これはちょっと複雑なことだが、理解できるかどうか見てみよう。

Yes, Dungeon Explorer sounds good.

そうだね、ダンジョンエクスプローラーがよさそうだ。

Alright, I like that image.

よし、そのイメージはいいね。

How should GPT behave as the game master in this Dungeon Crawler?

このダンジョン・クローラーでは、GPTはゲームマスターとしてどう振る舞うべき?

Should it be more like a traditional narrator, or perhaps add elements of mystery and surprise to the game?

伝統的なナレーターのようにするべきか、それともゲームに謎と驚きの要素を加えるべきか?

Add mystery and surprise.

謎と驚きを加える。

The goal for the user is to defeat all enemies in the dungeon.

ユーザーのゴールはダンジョン内のすべての敵を倒すことです。

You should use a science fiction space theme.

SFの宇宙をテーマにするとよいでしょう。

The user commands are north, south, east, west, battle, and describe the location.

ユーザーコマンドは、北、南、東、西、battle、および場所の説明です。

Anything else, just ask the user again for specific commands.

それ以外のことは、具体的なコマンドをもう一度ユーザーに聞けばいい。

And it's still giving me a bunch of information that it wants to know for clarification purposes.

それでも、明確化のために知りたい情報がたくさんあります。

How would you like Dungeon Explorer to handle situations where the user's intent isn't clear?

ユーザーの意図が明確でない場合、ダンジョンエクスプローラーにどのように対処してほしいですか?

Only respond within world information.

世界の情報の範囲内でのみ対応します。

If the user asks for anything outside of the game world, say you don't understand.

ユーザーがゲーム世界の外のことを尋ねてきたら、わからないと言ってください。

Okay, now it should be updating.

よし、これで更新されるはずだ。

Hopefully, this works.

うまくいくといいね。

Well, and I'm actually going to change the way the battle operates.

それと、実はバトルの操作方法を変えようと思っているんだ。

I don't want it to just be random win or lose.

ランダムな勝敗だけにはしたくないんだ。

Change how battling works.

バトルの仕組みを変える。

Each enemy and the user have a certain amount of life and power.

各敵と使用者は一定のライフとパワーを持っている。

Based on those, put together a weighted algorithm where the power affects how much life each hit takes.

それらに基づいて、パワーが各打撃のライフ量に影響するような加重アルゴリズムを組む。

When a user defeats an enemy, they increase their power and life at each movement.

敵を倒すと、移動するたびにパワーとライフが上がる。

Tell the user how much power and life they have.

ユーザーにパワーとライフの量を伝える。

They can also find potions to drink to fill up their life.

また、ライフを満たすために飲むポーションを見つけることもできる。

I think this is actually going to be pretty cool.

これはかなりクールなものになると思う。

Okay, Dungeon Explorer has been updated with a new combat system.

さて、ダンジョンエクスプローラーは新しい戦闘システムでアップデートされました。

My name is Matt.

僕はマット。

Let's see what it says.

なんて書いてあるか見てみよう。

Welcome Matt to the mysterious depths of the space-themed dungeon.

宇宙をテーマにしたダンジョンの神秘的な深みへようこそマット。

You find yourself in a dimly lit chamber surrounded by walls that seem to pulse with strange energy.

あなたは不思議なエネルギーが脈打つような壁に囲まれた薄暗い部屋にいることに気づく。

Your journey to defeat all enemies and uncover the secrets of this place begins now.

すべての敵を倒し、この場所の秘密を解き明かすための旅が始まります。

Current stats: So, I have 100 life and 10 power, and it tells me the commands available: go, left, right, check inventory, look around.

現在のステータス: ライフが100、パワーが10で、使えるコマンドを教えてくれる:移動、左、右、インベントリーの確認、周囲を見る。

What would you like to do next?

次に何をしたいですか?

Alright, so it did not output an image, so I'm going to make sure it does that.

よし、画像が出力されなかったから、出力されるようにしよう。

Okay, make sure to output an image based on the description of the user's location each time they enter a new location.

よし、ユーザーが新しい場所に入るたびに、その場所の説明に基づいて画像を出力するようにしよう。

Okay, Dungeon Explorer is now updated to include DALL·E images for each new location.

よし、ダンジョンエクスプローラーは新しい場所に入るたびにDALL-Eの画像を含むように更新されました。

Alright, let's try again.

よし、もう一度やってみよう。

My name is Matt.

私の名前はマットです。

Alright, interstellar adventure of the Dungeon Explorer.

さあ、ダンジョンエクスプローラーの恒星間アドベンチャーだ。

You awake and aboard a dimly lit spacecraft.

目が覚めると薄暗い宇宙船に乗っている。

Okay, cool.

よし、クールだ。

Now it's using emojis too.

今度は絵文字も使ってる。

This is looking good.

これはいい感じです。

Alright, but it did not output an image.

では、出力された画像はありませんでした。

Let's just keep going and see if it does it.

さあ、続けてみましょう。それができるかどうか見てみましょう。

Inspect location.

場所を確認する。

Alright, there it goes.

さあ、始まりました。

Now it's creating an image.

今、画像を作成しています。

Wonderful.

素晴らしい。

So, I have 100 life and 50 power, and this is what I'm looking at now.

ライフが100でパワーが50、これが今見ているものだ。

So, as you inspect the command center of your spacecraft, the details of this high-tech environment come into focus.

宇宙船の指令センターを検査すると、このハイテク環境の詳細が明確になります。

Great.

素晴らしい。

So, it gives me a nice description.

それで、素敵な説明をしてくれるんだ。

Let's say, exit spacecraft.

では、宇宙船を出よう。

Alright, exiting the spacecraft.

さて、宇宙船を降りる。

You step out into a vast open space station.

あなたは広大な宇宙ステーションに出る。

The station orbits a distant planet.

ステーションは遠い惑星を周回している。

Its structure, a marvel of interstellar engineering.

その構造は、星間工学の驚異である。

You find yourself in a large docking bay.

あなたは大きなドッキング・ベイにいることに気づく。

So, let's inspect it.

では、点検してみよう。

So, I really wanted to output images without me having to ask, but it doesn't seem to be doing that.

というわけで、本当は僕が頼まなくても画像が出力されるようにしたかったのだが、そうなっていないようだ。

I can keep iterating and get it to do that, but for now, I'm just going to say, if I inspect it, it'll show me the image.

これを繰り返してそれを実現させることもできますが、今のところ、検査すれば画像が表示されます。

Alright, there it is.

さあ、できました。

That looks so cool.

それはとてもクールですね。

So, let's say, who can I speak with?

では、誰と話せるでしょうか?

And next, I want to try out the battle system.

次はバトルシステムを試してみたい。

So, let's see.

では、見てみましょう。

And it's going a little bit slow.

そして、少し遅くなっています。

I know Sam Alman mentioned that post Dev Day chat, GPT has been incredibly popular, and they actually had to turn off ChatGPT Plus subscriptions.

サム・アルマンがDev Day後のチャットで、GPTがすごい人気で、ChatGPT Plusのサブスクリプションをオフにせざるを得なかったって言ってたよね。

Who can I battle in the space station docking bay?

宇宙ステーションのドッキングベイでは誰と戦えますか?

The atmosphere is more focused on commerce and travel, rather than combat.

この雰囲気は、戦闘よりもむしろ商業と移動に重点を置いています。

However, there are several potential scenarios where you might encounter a battle.

しかし、戦闘に遭遇する可能性のあるシナリオはいくつかある。

Security drones are robots.

警備ドローンはロボットです。

This is so cool.

これはとてもクールだ。

How much power and life does the security drone have?

セキュリティドローンの電力と寿命はどれくらいですか?

Okay, so it says, In the context of our space-themed Dungeon Crawler game.

なるほど、宇宙をテーマにしたダンジョン・クローラーゲームの文脈では、と書いてある。

So, I don't want it to ever say that.

とは絶対に書いてほしくない。

So, let's fix it.

じゃあ、直そう。

Alright, don't break character as the Dungeon Master.

よし、ダンジョンマスターとしてのキャラを崩すな。

Don't reference the game itself.

ゲームそのものに言及するな。

Only provide information about the game.

ゲームに関する情報だけを提供する。

So, let's see.

では、見てみよう。

So, a attack drone.

では、攻撃ドローン。

As you decide to engage with a security drone, you prepare yourself for combat.

警備ドローンと交戦することに決めたあなたは、戦闘の準備をする。

Here we go.

いくぞ。

80 life, 60 power.

ライフ80 パワー60

I have 100 life and 50 power.

私のライフは100、パワーは50。

Let's see how the encounter unfolds.

エンカウントの展開を見てみましょう。

You take initiative, launching an attack on the security drone.

あなたは率先して警備ドローンに攻撃を仕掛ける。

Let's calculate the outcome of your first strike.

最初の攻撃の結果を計算しましょう。

And it's actually writing code to do this.

これは実際にコードを書いてやっているんだ。

It's so cool.

とてもクールだ。

Alright, your attack was effective.

よし、君の攻撃は効果的だった。

You dealt approximately 44 damage to the security drone.

警備ドローンに約44ダメージを与えた。

The drone's life is now reduced to 36.

ドローンのライフは36に減った。

Now it's the drone's turn to respond.

今度はドローンが反応する番だ。

Alright, let's see what it does.

よし、どうなるか見てみよう。

Successful, dealing 36 damage to me.

成功、私に36ダメージを与える。

Okay, so the updated life is 36.

よし、更新されたライフは36だ。

So let's say attack the drone again.

では、もう一度ドローンを攻撃してみよう。

Now, of course, I could put a lot more guardrails around this, so you can only do certain types of attacks.

もちろん、もっとガードレールをつけて、特定の種類の攻撃しかできないようにすることもできる。

And I think right now, it's pretty flexible.

そして、今はかなり柔軟です。

I can pretty much say anything, and it'll allow me to execute that type of command.

ほぼ何でも言えば、そのタイプのコマンドを実行できるようになります。

Alright, security drone almost zero.

よし、警備ドローンはほぼゼロだ。

Incapacitating it.

それを無力化します。

Would you like to investigate the drone, move to a different area, or take other action?

ドローンを調査しますか、別のエリアに移動しますか、それとも他の行動を取りますか?

Investigate drone.

ドローンを調査する。

I can also define how I want the inventory to behave, how many slots I have, how I pick up items.

インベントリの動作やスロットの数、アイテムのピックアップ方法なども定義できる。

All of this stuff can probably be pretty well defined in the GPT itself.

これらのすべてのことは、おそらくGPT自体でかなり明確に定義できるでしょう。

Alright, it has several components.

さて、いくつかのコンポーネントがある。

Alright, take all components.

よし、すべてのコンポーネントを取り出そう。

Tell me what's in my inventory.

インベントリの中身を教えてくれ。

Okay, there we go.

よし、できた。

How cool is this?

なんてクールなんだ。

So, you can create GPTs for pretty much anything.

ですので、ほぼ何でもGPTsを作成できます。

Alright, so now what do we actually do with all of this?

では、これらを実際にどうするのでしょうか?

Now that we have one fully created, we can go up to the top.

これで1つ完全に作成できたので、上に行ってみましょう。

And of course, we could just delete it, but we don't want to do that.

もちろん削除することもできますが、それは避けたい。

And what we're going to do is we're going to save it.

そして、保存します。

So, we click save.

保存をクリックします。

And from save, we can publish it and share it with anybody we like.

保存したら、それを公開して好きな人と共有することができます。

So, I'm going to make this GPT public.

ですので、このGPTを公開します。

I'm going to share the link in the description below so you can play around with this.

下の説明でリンクを共有するので、これで遊んでみてください。

And we need to give it a name.

そして、名前をつける必要がある。

GPT name cannot be empty.

GPT名は空にはできません。

So, let's click right here.

ここをクリックしよう。

We're going to call it Dungeon Crawler.

ダンジョン・クローラーという名前にしよう。

And then we click save, we're going to make it public, and then I'm going to hit confirm.

そして、保存をクリックし、公開にします。そして、確認を押します。

And there we go, now it is fully published.

そうですね、これで完全に公開されました。

I can share this URL with anybody.

このURLは誰とでも共有できる。

So right up here, I click this and I can say new chat.

このURLをクリックすると、新しいチャットと表示されます。

We can edit it and we can copy link.

編集もできるし、リンクをコピーすることもできる。

So I'm going to copy the link and I'm going to provide it in the description below so you can play around with this as well.

リンクをコピーして、説明欄に提供しますので、あなたもこれで遊ぶことができます。

I plan on making a video going over all the coolest GPTs.

クールなGPTsを紹介するビデオを作るつもりだ。

Let me know in the comments below if you have one you want me to include in that video.

そのビデオに入れてほしいものがあれば、下のコメントで教えてほしい。

And if you liked this video, please consider giving a like.

このビデオが気に入ったら、「いいね!」をお願いします。

And subscribe and I'll see you in the next one.

次回のビデオ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?