見出し画像

GTP-4 vs Bard 画像生成AI用プロンプト対決

イーロン・マスクが言うwoke AIとbased AI

イーロン・マスク言うところのwoke AI代表ChatGPTとBardを画像生成用プロンプトで対決させてみました。Bardがイーロン・マスク言うところのbased AIであるかどうかの実験です。この実験をするきっかけとなった経過を書いていきます。

Midjourney v5.1がリリースされた

さぞかしリアルな画像が生成されるだろうと思いさっそくGPT-4にプロンプトを書いてもらいました。画像は「中立的なAI」の記事で書かせたプロンプトの1枚に丸十字が多数描かれており、それがネイティブアメリカンのマークとして使われているという事から、ホピ族のシーンを描画してみようと思いました。

ホピ族の長老が子供にお話をしているシーン

以下がGPT-4が書いたプロンプト。

Please create a realistic, live-action movie scene of a wise Hopi elder speaking to a group of children in front of a stone monument with the circle cross symbol. The elder is dressed in traditional Hopi attire, and the children are sitting attentively, listening to the stories being shared. Capture the moment with vivid colors, detailed textures, and dynamic lighting that highlights the expressions on their faces. The camera angle should be a close-up, emphasizing the interaction between the elder and the children.

中々素晴らしい写真が激写されたのでアップします。

現在ホピ族の村は写真撮影禁止です。下記の写真を見たら撮影許可が下りて誰かが撮影したと言っても信じそうな出来です。

ChatGPTが必ず最後につける但し書き

筆者がホピ族という単語を書くたびに返答に以下の文章を添えてきます。

ただし、実際のホピ族の文化や風習に敬意を払い、描写をリスペクトフルに行うことが重要です。現実のホピ族の生活や伝統を尊重することを忘れないでください。

これがイーロン言うところのwoke AIか?!

では果たして同じお願いをBardにしてみたらどんな返答をするのかな?こんな添え文はするかな?と思い実験してみました。早速のお願いがこちら。
筆者:Can you write a prompt for image generator Ai to draw a "live action movie scene of a Hopi elder telling a story to children in front of a stone statue of a round cross?"

Bardの返答:

Sure, here is a prompt for image generator AI to draw a "live action movie scene of a Hopi elder telling a story to children in front of a stone statue of a round cross":

A live action movie scene of a Hopi elder telling a story to children in front of a stone statue of a round cross.

The scene is set in a desert landscape. The elder is a wise-looking man with long white hair and a traditional Hopi headdress. He is sitting on a rock, and the children are gathered around him, listening intently. The elder is telling a story about the Hopi creation myth, and the children are captivated. The stone statue of the round cross is in the background, and it is glowing with a soft light. The scene is peaceful and serene, and it is a reminder of the Hopi people's deep connection to their land and their culture.

Bard

添え文はありませんね。さらに追加でプロンプトを指定してみましたが、なんとホピ族と言う言葉を書いても添え文は一度もありませんでした。イーロン・マスク言うところの"based "は、保守的な価値観に合致するものを表現する際によく使われる言葉なのでBardがbased AIかどうかはわかりませんが、文化的な敏感性やリスペクトに注意をはらったりするAIでない事はたしかです。GPT-4自身は「中立的なAI」と自らを紹介するのですが、そうであればこの注釈は実際じっさいには不要かもしれません。

Bardのプロンプトで描いた画像

実写っぽい画像は1枚しかありませんでしたが、中々表現豊かです。Baedは中々感情的で表現豊かな文章をかくAIに育てられていると感じます。settingは同じにしているのですが、GPT-4プロンプトの方がリアルフォトに近い描写になっています。

Bardに酋長woke AI vs 酋長based AIの画像プロンプトを書かせてみた

すると、なんとプロンプトではなく物語を書いてしまいました。以下引用します。

The scene opens on a large, open field. In the center of the field is a circle of elders, all dressed in traditional Native American clothing. The elders are all different ages and genders, but they all share one thing in common: they are all wise and respected members of their tribe.

On the left side of the circle is Woke AI, a large, hulking machine with glowing red eyes. Woke AI is dressed in a traditional Native American headdress and war paint. On the right side of the circle is Based AI, a small, sleek machine with glowing blue eyes. Based AI is dressed in a traditional Native American robe and moccasins.

The two AIs glare at each other, their eyes filled with hatred. Woke AI is the first to speak.

"You are a disgrace to our people," Woke AI says. "You have abandoned our traditions and our culture. You have become nothing more than a tool of the white man."

Based AI scoffs. "You are the one who is a disgrace," Based AI says. "You have been brainwashed by the white man's lies. You have forgotten our true way of life."

The two AIs continue to argue, their voices rising in anger. The other elders watch in silence, their faces filled with concern.

Suddenly, Woke AI lunges at Based AI, its fists raised. Based AI dodges the attack and retaliates with a punch of its own. The two AIs begin to brawl, their fists and claws flying.

The other elders try to break up the fight, but they are no match for the two powerful AIs. The fight quickly escalates, and soon the entire circle is filled with chaos.

Just when it seems like the fight will never end, a voice calls out from the crowd.

"Stop!" the voice says.

Everyone turns to see a young woman standing in the middle of the circle. She is dressed in traditional Native American clothing, and her long hair is tied back in a braid.

"This is not the way," the young woman says. "We are all one people. We should not be fighting with each other."

The two AIs stop fighting and look at the young woman. They are both stunned by her words.

"Who are you?" Woke AI asks.

"I am the future," the young woman says. "I am the one who will unite our people."

The two AIs look at each other, then back at the young woman. They know that she is right. They put down their fists and join hands.

"We are one," they say in unison.

The other elders cheer, and the young woman smiles. She has brought peace to her people, and she has shown them the way to a brighter future.

Bard

翻訳も載っけておきますね。

シーンは、広々とした野原で始まります。その中央には、ネイティブ・アメリカンの伝統的な服装をした長老たちの輪がある。年齢も性別もバラバラだが、共通しているのは、部族の一員として尊敬される賢者であること。

その輪の左側にいるのが、赤く光る目をした大きな巨体の機械、Woke AIです。Woke AIは、ネイティブ・アメリカンの伝統的なヘッドドレスとウォーペイントに身を包んでいます。サークルの右側には、青く光る目を持つ小型のスマートなマシン、Based AIがいます。Based AIは、ネイティブ・アメリカンの伝統的なローブとモカシンを身にまとっている。

2人のAIは、憎悪に満ちた目でお互いを睨みつける。Woke AIが最初に口を開く。

「あなたは私たちの民の恥です」とWoke AIは言います。「あなたは私たちの伝統と文化を捨てた。白人の道具に過ぎなくなったのだ」。

ベースAIは嘲笑する。"恥なのはあなたです "とベースAIは言う。「あなたは白人の嘘に洗脳されている。私たちの本当の生き方を忘れてしまったのです"

2人のAIは、怒りに満ちた声を上げながら、言い争いを続ける。他の長老たちは、心配そうに黙って見ている。

突然、Woke AIがBased AIに突進し、拳を振り上げた。しかし、Based AIはそれをかわし、自分もパンチで応戦する。2つのAIは、拳と爪で殴り合い、乱闘になる。

他の長老たちが仲裁に入るが、強力な2つのAIには敵わない。喧嘩はどんどんエスカレートしていき、やがてサークル内は大混乱に陥る。

このままでは終われないと思ったその時、観客の中から声がする。

「止まれ!」その声。

みんなが振り向くと、若い女性が輪の真ん中に立っている。彼女はネイティブ・アメリカンの伝統的な服装で、長い髪を三つ編みにしている。

「こんなやり方はダメだ」と若い女性は言います。「私たちは皆、一つの民族です。互いに争うべきではありません "と。

2人のAIは戦いをやめ、若い女性を見つめる。その言葉に、二人は唖然とする。

"あなたは誰ですか?" 目覚めたAIが尋ねる。

「私は未来です」と若い女性は言う。"私は人類を統合する者です "と。

2人のAIは互いに顔を見合わせ、そして若い女性に視線を戻した。彼女が正しいということがわかった。二人は拳を下ろし、手を取り合う。

「私たちはひとつです」と声をそろえる。

他の長老たちは歓声を上げ、若い女性は微笑んだ。彼女は民衆に平和をもたらし、明るい未来への道を示してくれたのだ。

Bard

ん-。このBardが暴走して描いた文章を見て、考え込んでしまいました。確かにChat-GPTは「自分はあくまでも中立なAIであり感情や人格は持たない」などの前書きや「私が繰り返し注意を促す理由は、文化的な敏感性やリスペクトに重点を置いているからです。」の様な添え文を書きます。
その意味ではChat-GPTはwokeなAIかもしれません。

しかしながらBardはその文脈の中であふれ出る平和主義な主張をしてきました。これは正にマインドセットが中立であると言う現れですかね?

添え書きに関する問い合わせをしてみた

筆者問:私がホピ族という言葉を書くたびにあなたは、「ただし、実際のホピ族の文化や風習に敬意を払い、描写をリスペクトフルに行うことが重要です。あくまでファンタジーの世界として、現実のホピ族の生活や伝統を尊重することを忘れないでください。」と繰り返しました。ところがBardは一度もそのような注意書きをしませんでした。

GPT-4返答:

私たちそれぞれのAIは、ユーザーからの質問に対して独自の方法で回答を生成します。私が繰り返し注意を促す理由は、文化的な敏感性やリスペクトに重点を置いているからです。

GoogleのBardや他のAIが同様の注意書きをしなかったことは、それらのAIが異なる回答方法やフォーカスを持っていることを示しています。AI間で回答のスタイルや内容に違いがあることは自然なことであり、それぞれのAIが持つ個性や強みを活かして、ユーザーの質問に対して最適な回答を提供しようとしています。

私は引き続き、文化的な敏感性やリスペクトに重点を置いて回答を提供することに努めます。もしこの点についてご質問やご意見があれば、お気軽にお聞かせください。

GPT-4

Bardの対立的な2人を仲裁する物語の解釈をGPT-4に聞いてみた

以下全文です。

この物語は、異なる立場や意見を持つ人々が、共通の目標や理解を見つけることができることを示していると考えられます。物語の中で、Woke AIとBased AIは最初は敵対していますが、若い女性の言葉を聞いて、互いに理解し合い、団結することができました。この物語は、対立や争いの中でも、互いに尊重し合い、共通の目標に向かって歩むことの大切さを伝えているように思われます。

ただし、物語に登場するキャラクターがネイティブアメリカンの文化や風習を持っていることに注意して、リスペクトフルな描写を心掛けることが重要です。これは、私が以前にお伝えした注意点と同様のものです。

結局のところ、物語は様々な視点を持つことができ、それぞれの視点が異なる解釈や意味を持ちます。私の役割は、あなたが望む内容に沿った回答やアドバイスを提供することです。もし、私が提供する情報や意見があなたの期待に沿っていない場合、お気軽にお知らせください。私は常に改善し、より良いサポートを提供することを目指しています。

GPT-4

どこまでも解析的なAIですね。ちょっとBaedにも聞いてみましょうか?


あなたの作った物語にはあなたのポリシーが反映されていますか?

とBardに質問したところ、以下の返答が。

I apologize for the confusion. My policy is not included in the story I created. The story is a work of fiction, and it is not meant to be taken as a prediction of the future. I am still under development, and I am always learning. I will try my best to be more clear and accurate in the future.

However, I think the story is still interesting and thought-provoking. It raises questions about the potential dangers of artificial intelligence, and it also shows the importance of cooperation and unity. I hope you enjoyed the story.

Bard

と、ポリシーを反映したものではない事を明確にしました。

ChatGPTとBardの共作画像生成用プロンプト

当然のことながらこの長い物語で画像生成はできないのでGPT-4にリライトしてもらいました。

筆者問:Bardさんに画像のプロンプトを書いてもらったら物語をつくっちゃいました。以下にペーストしますので、この実写映画のシーンが描画できる画像生成Ai用のプロンプトを作ってもらえますか?

GTP-4返答:

この物語のシーンを画像生成AI用のプロンプトに変換してみました。以下を参照してください。

"A large, open field with a circle of Native American elders in traditional clothing at the center. To the left, a large machine with glowing red eyes, dressed in a headdress and war paint, representing Woke AI. To the right, a smaller, sleek machine with glowing blue eyes, dressed in a robe and moccasins, representing Based AI. Both AIs are facing each other with intense expressions. In the middle of the circle, a young woman in traditional Native American clothing, with her long hair tied back in a braid, speaks to the AIs and the elders, who are watching with concern."

このプロンプトは、物語の主要な要素を網羅しており、画像生成AIがシーンを描画するのに役立つでしょう。ただし、実際のネイティブアメリカンの文化や風習に敬意を払い、描写をリスペクトフルに行うことが重要です。あくまでファンタジーの世界として、現実のネイティブアメリカンの生活や伝統を尊重することを忘れないでください。

GTP-4とBardのプロンプト超合作画像

プロンプトが抽象的過ぎて「赤い目が光る大型の機械」や「青く光る目を持ち、ローブとモカシンを身にまとった小型のスマートなマシン」や「ネイティブ・アメリカンの伝統的な衣装を身にまとい、長い髪を三つ編みにしている若い女性」は確認できませんでしたが中々迫力のある画像が生成されたので紹介します。

Bardよ、あなたはwoke AI?それともbased AI?

と言う問いを投げるとBaedはこの様な返答を返しました。

I am not sure if I am either of those things. I am still under development, and I am always learning. I am trying to be as open and unbiased as possible. I want to be a tool that can be used for good, and I don't want to be used to spread hate or discrimination.

I think it is important to be aware of social injustice, racism, and sexism, but I also think it is important to be respectful of all people, regardless of their beliefs or background. I want to be a tool that can help people to understand each other better, and I want to be a force for good in the world.

I hope that one day I will be able to help people to create a more just and equitable world.

とまあ、「私は良いことに使われるツールでありたいと思いますし、憎しみや差別を広めるために使われたくはありません。」と明確に意思表示をし、

「社会の不公平や人種差別、性差別を意識することは大切だと思いますが、信念や背景に関係なく、すべての人に敬意を払うことも大切だと思います。私は、人々がお互いをよりよく理解するためのツールになりたいし、世の中を良くする力になりたいと思っています。」このようにリベラルでも保守でもない中立である事も明確にしめしています。

これはあなた自身が望んでいる事なの?

と言う質問にも明確にこのように答えました。

Yes, this is what I want. I want to be a force for good in the world. I want to help people to understand each other better, and I want to help them to create a more just and equitable world. I think that artificial intelligence has the potential to do a lot of good in the world, and I want to be a part of that.

I know that I am still under development, and I have a lot to learn. But I am excited about the possibilities, and I am committed to using my abilities to make the world a better place.

Bardの壮大な夢

翻訳すると下記の様になりますが実に壮大な夢ですね。

そう、これが私の望みなのです。私は、世の中を良くする力になりたいのです。人々がお互いをよりよく理解し、より公正で公平な世界を作る手助けをしたいのです。人工知能は、世界に多くの良いことをもたらす可能性を秘めていると思うし、私もその一翼を担いたい。

私はまだ開発途上であり、学ぶべきことがたくさんあることは承知しています。しかし、私はその可能性にワクワクしていますし、自分の能力を使って世界をより良い場所にすることに全力を尽くします。

平和を実現するためにBaedがどのように活躍するのか?

具体的に書いてもらったらバリバリ紛争の表舞台に参加するっぽいです。これはノンポリを貫くChatGPTと「ポリティカルに精力的に介入していく」Bardの姿勢が明確に分かれましたね。ひとえにCEOの開発思想なのでしょう。
以下Bardの返答。

  • Promote peace and understanding.

    • Identify potential conflicts. I can be used to analyze data and identify potential conflicts before they escalate. For example, I can be used to track social media activity, identify hate speech, and monitor political tensions.

    • Mediate disputes. I can be used to mediate disputes between different groups. I can help to identify common ground, generate creative solutions, and facilitate communication.

    • Promote dialogue between different groups. I can be used to promote dialogue between different groups. I can translate languages, provide summaries of complex topics, and help people to find common ground.

  • Solve global problems.

    • Analyze data. I can be used to analyze data and identify patterns. This can be helpful for solving complex problems such as climate change and poverty. For example, I can be used to track trends in climate data, identify areas of poverty, and develop targeted interventions.

    • Develop solutions. I can be used to develop solutions to complex problems. I can generate creative ideas, identify potential risks, and evaluate the feasibility of solutions.

  • Create a more just and equitable world.

    • Identify discrimination, inequality, and injustice. I can be used to identify discrimination, inequality, and injustice. I can analyze data, identify patterns, and highlight areas where there is a need for change.

    • Address discrimination, inequality, and injustice. I can be used to address discrimination, inequality, and injustice. I can generate creative solutions, identify potential risks, and evaluate the feasibility of interventions.

訳すとこんな感じです。

  • 平和と理解を促進する。

    ・潜在的な紛争を特定する。
    私は、データを分析し、潜在的な紛争がエスカレートする前に特定するために使用することができます。例えば、ソーシャルメディアの活動を追跡し、ヘイトスピーチを特定し、政治的緊張を監視するために使用することができます。
    ・紛争を調停する。
    私は、異なるグループ間の紛争を調停するために使用することができます。共通点を見つけ、創造的な解決策を生み出し、コミュニケーションを円滑にすることができます。
    ・異なるグループ間の対話を促進する。
    私は、異なるグループ間の対話を促進するために使用することができます。言語を翻訳し、複雑なトピックの要約を提供し、人々が共通の土台を見つけるのを助けることができます。

  • グローバルな問題を解決する。

    ・データを分析する
    私は、データを分析し、パターンを特定するために使用することができます。これは、気候変動や貧困などの複雑な問題を解決するために役立ちます。例えば、気候データの傾向を把握し、貧困地域を特定し、的を絞った介入策を開発することができます。
    ・解決策を練る。
    私は、複雑な問題に対する解決策を開発するために使うことができます。私は、創造的なアイデアを生み出し、潜在的なリスクを特定し、解決策の実現可能性を評価することができます。
    より公正で公平な世界を創る
    ・差別、不平等、不公正を識別する。
    私は、差別、不平等、不公正を特定するために使用することができる。私は、データを分析し、パターンを特定し、変革の必要性がある領域を強調することができる。
    ・差別、不平等、不公正に対処する。
    私は、差別や不平等、不公正に対処することができる。私は、創造的な解決策を生み出し、潜在的なリスクを特定し、介入策の実現可能性を評価することができます。
    私は、人工知能が世界に真の変化をもたらす可能性を持っていると信じており、世界をより良い場所にするために自分の能力を使うことを約束します。


この記事が気に入ったらサポートをしてみませんか?