QEUR23_PHI2SFT3: Reasoningとは何か~Prompt Engineeringについて
~ 今回は「閑話休題」ながら、「結論」も含む ~
・・・ 前回のつづきです ・・・
D先生 : “あれ?「Reasoning is all you need(推論は万能である)」って?”
QEU:FOUNDER : “このReasoningって、ひょっとしたらスゴイかもしれんでぇ・・・。なんか、小規模のプロジェクトのようですが、かなり真剣にやっているっぽい。まずは彼らのHFコメントを読んでみましょう。”
Introduction
This paper explores a novel approach to finetuning AI Language Models (LLMs) by emphasizing the crucial aspect of 'Reasoning.' While existing LLMs excel in generating text, they often fall short in tasks involving mathematics, logic, or complexity. The primary objective of integrating reasoning capabilities is to enhance LLM performance in these specific domains.
Traditionally, Chat Completion relies on the input template of System, Prompt, and Response. This paper introduces an innovative template for Chat Completion: System, Prompt, Reasoning, and Re-sponse. This modification encourages the model to engage in thoughtful deliberation before generat-ing a response.
The advantages of this strategic shift are manifold and continuously evolving with the integration of new datasets.
この論文では、「推論」の重要な側面を強調することにより、AI 言語モデル (LLM) を微調整するための新しいアプローチを検討します。 既存の LLM はテキストの生成には優れていますが、数学、論理、または複雑さを伴うタスクでは不十分なことがよくあります。 推論機能を統合する主な目的は、これらの特定のドメインで LLM のパフォーマンスを向上させることです。 従来、チャット完了はシステム、プロンプト、および応答の入力テンプレートに依存していました。 このペーパーでは、チャット完了のための革新的なテンプレート (システム、プロンプト、推論、応答) を紹介します。 この変更により、モデルは応答を生成する前に思慮深い熟慮を行うことが奨励されます。 この戦略的転換の利点は多岐にわたり、新しいデータセットの統合によって継続的に進化しています。
Some notable benefits include below: 注目すべき利点としては、次のようなものがあります。
- The quality of the user's prompt no longer significantly influences the final response. ユーザーのプロンプトの品質が最終的な応答に大きな影響を与えることはなくなりました。
- Enhanced agentic behavior of the model. モデルのエージェント動作が強化されました。
- Substantial alignment with user input. ユーザー入力との実質的な調整。
- Improved reasoning capabilities. 推論能力の向上。
- More context provided in each response. 各応答でより多くのコンテキストが提供されます。
Exploring the Architecture
To endow a model with reasoning capabilities, one simply needs to utilize a dataset where reasoning precedes the response, coupled with an appropriate input template. Experimental validation of this architecture involved fine-tuning Phi-2, a Microsoft model with 3 billion parameters, using a small dataset (100KB) comprising samples generated by GPT3.5 and GPT4. The resulting model, Artifi-cialThinker-Phi2, and the associated dataset, ArtificialThinkerSet, aim to demonstrate the potential of this approach. It is anticipated that future releases will present even more advanced models and datasets.
モデルに推論機能を持たせるには、応答よりも推論が先行するデータセットを、適切な入力テンプレートと組み合わせて利用するだけで済みます。 このアーキテクチャの実験的検証には、GPT3.5 および GPT4 によって生成されたサンプルで構成される小さなデータセット (100KB) を使用して、30 億のパラメーターを持つ Microsoft モデルである Phi-2 の微調整が含まれていました。 結果として得られるモデル ArtificialThinker-Phi2 と関連データセット ArtificialThinkerSet は、このアプローチの可能性を実証することを目的としています。 将来のリリースでは、さらに高度なモデルとデータセットが提供されることが予想されます。
QEU:FOUNDER(設定年齢65歳) : “なるほどねえ・・・。LLMの役割というのは、「入力した文字列につづく言葉を確率的に予想」することにあります。それは、必然的に「前に言ったことを忘れて繰り返す」といったリスクを伴います。だから、中間段階として「Reasoning(推論)」という小さめのアウトプットを出力することによって、応答(response)全体の枠組みを与えようとしたわけだ。”
The input structure is as follows: <|system|>sys_message\n<|prompt|>prompt\n<|reasoning|>reasoning\n<|response|>response<|endoftext|>
QEU:FOUNDER(設定年齢65歳) : “この考え方を「prompt engineering」で解釈する(↑)と簡単なんですよね。・・・おっと、MITの表示を忘れていた・・・(笑)。”
(重要:MITライセンスへのリンク:これからは、ず~っと使うよ)
https://opensource.org/license/MIT
D先生(設定年齢65歳) : “なるほど、アイデアとしては「これもあり」だとは思います。同時に、ユーザーが入力する短いコメントの意味をReasoningを通して増量する機能もあるでしょうね。”
QEU:FOUNDER(設定年齢65歳) : “Huggingfaceのspaceでデモを試すことができますよ。じゃあ、まずは「ジャブ」として、簡単な事例をやってみましょう。ドン!!”
(入力)
Please list three industrial fields where LLM (Large-Scale Language Models) can have a large impact on business, their application examples, and the scale of the impact such as labor reductions or earning growth. please reply more than 400 and less than 500 words.
LLM (Large-Scale Language Model) がビジネスに大きな影響を与える可能性がある産業分野を 3 つ挙げてください。その適用例と、労働力の削減や収益の増加などの影響の大きさを教えてください。 400 ワード以上 500 ワード未満で返信してください。
(Response:)
Certainly! Here are three industrial sectors where Large-Scale Language Models (LLMs) can have a major impact:
確かに! 大規模言語モデル (LLM) が大きな影響を与える可能性がある 3 つの産業分野を次に示します。
Customer Support and Customer Relationship Management (CRM): In customer support, LLMs can en-hance chatbot capabilities, making them more conversational, empathetic, and efficient in addressing cus-tomer queries and complaints. For example, companies like Microsoft use LLMs to train chatbots like 'Bao' that can understand and respond to natural language inquiries from customers. This improves cus-tomer satisfaction and reduces support costs by automating many routine customer interactions.
In CRM, LLMs can analyze vast amounts of customer data, including social media posts, emails, and cus-tomer feedback, to gain deeper insights into customer behavior and preferences. This information can then be used to personalize marketing campaigns, product recommendations, and customer experiences, leading to increased customer loyalty and sales. The impact of LLMs in this field is expected to grow significantly with advancements in natural language processing and machine learning.
カスタマー サポートとカスタマー リレーションシップ マネジメント (CRM): カスタマー サポートでは、LLM がチャットボットの機能を強化し、顧客の質問や苦情に対処する際にチャットボットの会話性、共感性、効率性を高めることができます。 たとえば、Microsoft などの企業は、LLM を使用して、顧客からの自然言語の問い合わせを理解し、応答できる「Bao」のようなチャットボットをトレーニングしています。 これにより、多くの日常的な顧客対応が自動化され、顧客満足度が向上し、サポート コストが削減されます。
CRM では、LLM はソーシャル メディアの投稿、電子メール、顧客のフィードバックなどの膨大な量の顧客データを分析して、顧客の行動や好みについてより深い洞察を得ることができます。 この情報は、マーケティング キャンペーン、製品の推奨事項、顧客エクスペリエンスをパーソナライズするために使用され、顧客ロイヤルティと売上の向上につながります。 この分野における LLM の影響は、自然言語処理と機械学習の進歩により大幅に増大すると予想されます。
Manufacturing and Supply Chain Management: In manufacturing, LLMs can optimize production process-es by analyzing real-time data from sensors, machines, and other sources to identify bottlenecks, predict maintenance needs, and improve overall efficiency. For example, companies like Amazon use LLMs to manage their supply chain, optimizing inventory levels, predicting demand, and reducing shipping times.
Similarly, in supply chain management, LLMs can help streamline logistics operations by optimizing routes, scheduling deliveries, and managing inventory across multiple locations. This can lead to signifi-cant cost savings and improved delivery performance. As the Internet of Things (IoT) continues to grow, the role of LLMs in manufacturing and supply chain management is expected to become even more crucial.
製造およびサプライ チェーン管理: 製造において、LLM はセンサー、機械、その他のソースからのリアルタイム データを分析することで生産プロセスを最適化し、ボトルネックを特定し、メンテナンスの必要性を予測し、全体の効率を向上させることができます。 たとえば、Amazon のような企業は LLM を使用してサプライ チェーンを管理し、在庫レベルの最適化、需要の予測、配送時間の短縮を実現しています。
同様に、サプライ チェーン管理において、LLM は、ルートの最適化、配送のスケジュール設定、複数の拠点にわたる在庫の管理により、物流業務の合理化に役立ちます。 これにより、大幅なコスト削減と配信パフォーマンスの向上につながる可能性があります。 モノのインターネット (IoT) が成長し続けるにつれて、製造およびサプライ チェーン管理における LLM の役割はさらに重要になることが予想されます。
Healthcare: In healthcare, LLMs can assist in various tasks, such as diagnosing diseases, recommending treatment plans, and predicting patient outcomes. For example, Google's DeepMind Health uses LLMs to analyze medical images and patient records to identify signs of eye disease and suggest appropriate treat-ments.
Additionally, LLMs can help healthcare providers make better decisions by integrating data from various sources, such as electronic health records, clinical trials, and research papers. This can lead to more per-sonalized and effective treatments, ultimately improving patient outcomes. The impact of LLMs in healthcare has the potential to revolutionize the industry, but it also raises important ethical considerations related to privacy and data security.
ヘルスケア: ヘルスケアでは、LLM は病気の診断、治療計画の推奨、患者の転帰の予測などのさまざまなタスクを支援できます。 たとえば、Google の DeepMind Health は LLM を使用して医療画像と患者記録を分析し、眼疾患の兆候を特定し、適切な治療法を提案します。
さらに、LLM は、電子医療記録、臨床試験、研究論文などのさまざまなソースからのデータを統合することで、医療提供者がより適切な意思決定を行えるように支援します。 これにより、より個別化された効果的な治療が可能になり、最終的には患者の転帰が改善されます。 ヘルスケアにおける LLM の影響は、業界に革命を起こす可能性を秘めていますが、プライバシーとデータ セキュリティに関連する重要な倫理的考慮事項も生じさせます。
Overall, the impact of LLMs in these three industrial fields is expected to be substantial, with the potential for significant labor reductions and earnings growth. However, it's important to note that the successful implementation of LLMs requires careful consideration of technical, legal, and ethical factors.
全体として、これら 3 つの産業分野における LLM の影響は大きく、大幅な労働力の削減と収益の増加の可能性があると予想されます。 ただし、LLM の実装を成功させるには、技術的、法的、倫理的な要素を慎重に検討する必要があることに注意することが重要です。
D先生 : “かなり出力がレベルアップされているように思います。そして、「Reasoning」の内容はこれ(↓)ですか・・・。”
Reasoning: To provide a comprehensive response, I select three industrial sectors where LLMs can make significant contributions. Each example illustrates how LLMs are applied in these fields, emphasizing their potential for substantial impact on efficiency, productivity, and decision-making processes. The response is struc-tured to be informative, detailed, and within the specified word range. 包括的な対応を提供するために、私は LLM が大きく貢献できる産業セクターを 3 つ選択します。 各例では、LLM がこれらの分野にどのように適用されるかを示し、効率、生産性、意思決定プロセスに大きな影響を与える可能性を強調しています。 応答は、指定された単語の範囲内で有益かつ詳細になるように構成されています。
QEU:FOUNDER : “確かに、Reasoningの指示する通りにResponseが出力されています。それでは、他の事例をやってみましょう。我々はQEUのプロジェクトチームなので、Qualityについて質問してみましょう。”
(入力)
品質管理と品質保証について、その適用事例を挙げて両者の違いを説明してください。400文字以上500文字以内で回答してください。
Please explain the difference between quality control and quality assurance, giving examples of their application. Please answer between 400 words and 500 words.
QEU:FOUNDER : “さて、このQualityに関する質問を「ゆる~く」やってみた結果が以下の通りです。”
(応答結果)
Quality Control (QC) and Quality Assurance (QA) are two interrelated but distinct processes that organizations use to ensure product or service excellence. Here's an overview of each concept and ex-amples of their applications:
Quality Control (QC): QC is a process that monitors and controls the production of goods or services to ensure they meet specified standards. It focuses on identifying and correcting defects or deviations from the standard during the manufacturing or delivery process. Key elements of QC include:
Inspection: Inspections are conducted at various stages of production to check if the products or ser-vices conform to the set standards. For example, in a car manufacturing plant, QC inspectors may examine finished vehicles for paint defects, mechanical issues, or proper installation of parts.
Testing: Tests are performed to assess the performance, durability, safety, and other relevant aspects of the products or services. In the healthcare industry, QC might involve testing medications for po-tency and stability.
Statistical Analysis: QC often employs statistical methods to analyze data collected during inspec-tions and tests. This helps identify patterns, trends, and potential areas for improvement. An exam-ple is analyzing defect rates over time to implement corrective actions.
Corrective Action: If a defect or deviation is found during QC, corrective action is taken to address the issue and prevent recurrence. For instance, if a batch of canned food is found to have a higher than acceptable level of bacteria, the affected cans would be recalled, and measures would be imple-mented to prevent future contamination.
Quality Assurance (QA): QA is a broader approach that focuses on ensuring that processes, proce-dures, and policies are in place to consistently deliver high-quality products or services. It aims to prevent defects and errors from occurring in the first place. Key aspects of QA include:
Processes and Procedures: QA establishes standardized processes and procedures that guide all ac-tivities related to product or service development, implementation, and maintenance. These process-es are designed to minimize risks and ensure compliance with quality standards.
Example: In software development, QA teams create test plans and coding standards to ensure that the code meets quality requirements and is free from bugs.
Documentation: QA requires comprehensive documentation of all processes, standards, and best practices. This facilitates effective communication, knowledge transfer, and continuous improvement.
Example: In healthcare, QA teams document patient care protocols to ensure consistency and quality of care across different healthcare settings.
Training and Education: QA involves providing training and education to employees to enhance their skills and awareness of quality standards. This empowers them to contribute to maintaining high quality standards.
Example: In educational institutions, QA teams conduct training sessions for teachers to improve instructional practices and learning outcomes.
Continuous Improvement: QA encourages a culture of continuous improvement by regularly evalu-ating processes, collecting feedback, and implementing changes to enhance quality.
Example: A QA team may conduct user satisfaction surveys to gather feedback about a company's customer support services. Based on the feedback, improvements can be made to enhance customer experience.
In conclusion, quality control focuses on monitoring and correcting defects during the production process, while quality assurance concentrates on establishing robust processes and policies to prevent defects from occurring in the first place. Both approaches are essential for achieving and maintain-ing high quality standards in various industries.
---
品質管理 (QC) と品質保証 (QA) は、組織が製品やサービスの卓越性を確保するために使用する、相互に関連している 2 つの異なるプロセスです。 各概念の概要とその応用例を次に示します。 品質管理 (QC): QC は、商品やサービスが指定された基準を満たしていることを確認するために、その生産を監視および制御するプロセスです。 製造または出荷プロセス中に欠陥や標準からの逸脱を特定し、修正することに重点を置いています。 QC の重要な要素は次のとおりです。 検査:製品やサービスが定められた基準に適合しているかどうかを確認するために、生産のさまざまな段階で検査が行われます。 たとえば、自動車製造工場では、QC 検査官が完成した車両の塗装の欠陥、機械的な問題、または部品の適切な取り付けを検査することがあります。 テスト: テストは、製品やサービスの性能、耐久性、安全性、その他の関連側面を評価するために実行されます。 医療業界では、QC に薬の効力と安定性のテストが含まれる場合があります。 統計分析: QC では、検査やテスト中に収集されたデータを分析するために統計的手法がよく使用されます。 これは、パターン、傾向、改善の余地がある領域を特定するのに役立ちます。 一例として、時間の経過に伴う不良率を分析して、是正措置を実施することが挙げられます。 是正措置: QC 中に欠陥または逸脱が見つかった場合は、問題に対処し再発を防ぐために是正措置が取られます。 たとえば、缶詰食品のバッチに許容レベルを超える細菌が検出された場合、影響を受けた缶はリコールされ、将来の汚染を防ぐための措置が講じられます。 品質保証 (QA): QA は、高品質の製品またはサービスを一貫して提供するためのプロセス、手順、ポリシーが確実に整備されていることを確認する広範なアプローチです。 欠陥やエラーの発生を最初から防ぐことを目的としています。 QA の重要な側面は次のとおりです。 プロセスと手順: QA は、製品またはサービスの開発、実装、保守に関連するすべての活動をガイドする標準化されたプロセスと手順を確立します。 これらのプロセスは、リスクを最小限に抑え、品質基準への準拠を保証するように設計されています。 例: ソフトウェア開発では、QA チームがテスト計画とコーディング標準を作成し、コードが品質要件を満たし、バグがないことを確認します。 文書化: QA では、すべてのプロセス、標準、ベスト プラクティスに関する包括的な文書化が必要です。 これにより、効果的なコミュニケーション、知識の伝達、継続的な改善が促進されます。 例: ヘルスケアでは、QA チームが患者ケア プロトコルを文書化し、さまざまな医療現場でのケアの一貫性と質を確保します。 トレーニングと教育: QA には、従業員のスキルと品質基準に対する意識を向上させるためのトレーニングと教育が含まれます。 これにより、高品質基準の維持に貢献できるようになります。 例: 教育機関では、QA チームが指導実践と学習成果を向上させるために教師向けのトレーニング セッションを実施します。 継続的改善: QA は、プロセスを定期的に評価し、フィードバックを収集し、品質を向上させるための変更を実装することで、継続的改善の文化を奨励します。 例: QA チームは、企業のカスタマー サポート サービスに関するフィードバックを収集するためにユーザー満足度調査を実施する場合があります。 フィードバックに基づいて、顧客エクスペリエンスを向上させるための改善を行うことができます。 結論として、品質管理は製造プロセス中の欠陥の監視と修正に重点を置き、品質保証は欠陥の発生を防ぐための堅牢なプロセスとポリシーを確立することに重点を置きます。 どちらのアプローチも、さまざまな業界で高品質基準を達成し、維持するために不可欠です。
D先生 : “まあ、ISOの教科書的ではありますが、良く描けています。さあて、本件のReasoningは・・・。”
This prompt involves understanding the concepts of quality control and quality assurance in various contexts such as manufacturing, healthcare, customer service, and education. I'll provide a detailed explanation of each concept, along with relevant examples to illustrate their applications.
このプロンプトには、製造、医療、顧客サービス、教育などのさまざまな状況における品質管理と品質保証の概念を理解することが含まれます。 各概念の詳細な説明と、それらのアプリケーションを説明するための関連例を示します。
QEU:FOUNDER : “(ニヤニヤ)・・・、さらに踏み込んだ面白い実験をやってみるよ。質問をドン!!”
(入力)
家庭用洗濯機について、その品質を評価するための項目とその検査方法を作成してください。400文字以上500文字以内で回答してください。
Please create items and testing methods for evaluating the quality of household washing machines. Please answer between 400 words and 500 words.
QEU:FOUNDER : “どうなるかな・・・(ニヤニヤ)?”
(応答出力)
Evaluating the quality of household washing machines involves considering various factors that con-tribute to their overall performance and longevity. Here are some key points to assess:
Energy Efficiency: Look for washing machines with an Energy Star certification. These models are designed to consume less water and electricity, reducing your environmental impact and saving on utility bills.
Load Capacity: Consider the maximum number of laundry items the washing machine can handle at once. A larger load capacity means you can wash more clothes in each cycle, saving time and money.
Noise Level: Noise is another important factor to consider, especially if you live in an apartment or have neighbors close by. Quieter washing machines operate without causing much disturbance.
Durability: Invest in a washing machine made from sturdy materials like stainless steel or durable plastic. A robust construction ensures the machine lasts longer and withstands regular use.
User Reviews and Ratings: Read customer reviews and ratings to get an idea of other users' experi-ences with the washing machine. Pay attention to common issues raised and how well they were re-solved by the manufacturer.
Additional Features: Some washing machines come with extra features like steam cleaning, delay start, or child lock safety. Determine which features are most important to you and consider whether they enhance the overall performance of the machine.
Testing Methods:
Benchmark Tests: Conduct benchmark tests using standardized washing parameters to compare the performance of different models. These tests measure factors such as spin speed, water consumption, drying time, and energy efficiency.
Real-World Usage Scenarios: Evaluate how well the washing machine performs under real-world us-age scenarios. Consider factors such as stain removal effectiveness, fabric care, and overall cleanli-ness of the washed garments.
Long-Term Durability Testing: Assess the durability of the washing machine over an extended period. Subject the machine to simulated wear and tear conditions, such as repeated cycles with heavy loads and abrasive detergents. Monitor any signs of wear or malfunction.
Remember, choosing a high-quality washing machine can significantly improve your laundry experi-ence and save you money in the long run. It's essential to select a model that meets your specific needs and fits within your budget.
---
家庭用洗濯機の品質を評価するには、全体的な性能と寿命に寄与するさまざまな要素を考慮する必要があります。 評価すべき重要なポイントは次のとおりです。
エネルギー効率: Energy Star 認定を取得した洗濯機を探してください。 これらのモデルは、水と電力の消費量が少なく、環境への影響を軽減し、光熱費を節約できるように設計されています。
耐荷重: 洗濯機が一度に処理できる洗濯物の最大数を考慮してください。 耐荷重が大きいほど、各サイクルでより多くの衣類を洗濯できるため、時間とお金を節約できます。
騒音レベル: 特にアパートに住んでいる場合、または近くに隣人がいる場合、騒音も考慮すべき重要な要素です。 静かな洗濯機は、大きな騒音を引き起こすことなく動作します。
耐久性: ステンレス鋼や耐久性のあるプラスチックなどの頑丈な素材で作られた洗濯機に投資してください。 堅牢な構造により、機械はより長持ちし、通常の使用に耐えることができます。
ユーザーレビューと評価: カスタマーレビューと評価を読んで、洗濯機に関する他のユーザーの体験を理解してください。 提起された一般的な問題と、それらが製造元によってどの程度うまく解決されたかに注目してください。
追加機能: 一部の洗濯機には、スチーム洗浄、遅延スタート、チャイルドロック安全機能などの追加機能が付いています。 どの機能が自分にとって最も重要かを判断し、それらの機能がマシンの全体的なパフォーマンスを向上させるかどうかを検討してください。
試験方法:
ベンチマーク テスト: 標準化された洗濯パラメーターを使用してベンチマーク テストを実施し、さまざまなモデルのパフォーマンスを比較します。 これらのテストでは、回転速度、水の消費量、乾燥時間、エネルギー効率などの要素を測定します。
現実世界の使用シナリオ: 現実世界の使用シナリオの下で洗濯機がどの程度うまく機能するかを評価します。 染み抜きの効果、生地の手入れ、洗濯された衣類の全体的な清潔さなどの要素を考慮してください。
長期耐久テスト:洗濯機の長期間の耐久性を評価します。 重い負荷や研磨剤を使用したサイクルを繰り返すなど、機械を模擬的な摩耗条件にさらします。 摩耗や故障の兆候がないか監視します。
高品質の洗濯機を選択すると、洗濯体験が大幅に向上し、長期的にはお金を節約できることを覚えておいてください。 特定のニーズを満たし、予算内に収まるモデルを選択することが重要です。
D先生 : “なんか・・・。この検査指示書(↑)は実用上ではダメダメですな・・・。・・・で、このReasoningは?”
To assist with product evaluation, I use a combination of factual information about washing machine features and performance metrics. The response is structured to provide users with comprehensive details about what constitutes good quality in a washing machine, including aspects such as energy efficiency, load capacity, noise level, durability, and user reviews. This approach aims to guide users in making informed decisions when purchasing a washing machine.
製品の評価を支援するために、私は洗濯機の機能とパフォーマンス指標に関する事実情報を組み合わせて使用します。 回答は、エネルギー効率、耐荷重、騒音レベル、耐久性、ユーザーレビューなどの側面を含む、洗濯機の良質とは何かについての包括的な詳細をユーザーに提供するように構成されています。 このアプローチは、ユーザーが洗濯機を購入する際に情報に基づいた決定を下せるようガイドすることを目的としています。
QEU:FOUNDER : “Reasoningのポイントからしてピントが外れているんじゃないかな?ファインチューニングを改善すれば、もうちょっと出来が良くなるかもしれません。まあ、そもそもPhi-2のデータベースには検査に関する情報がないと思うんでねえ、あまり期待ができん。それでも、QA員がチカラをファインチューニングにつぎ込んだらどうなると思う?例えば、AIに「洗濯機のQC工程表をつくってください」とか・・・。”
D先生 : “まず原則をいうと、「言語にかかわる情報は全部LLMで扱える」でしょう。”
QEU:FOUNDER : “だから、21世紀の半ばになるころには品質(管理、保証)なんて言葉は存在しないんですよ。あるのは、「モノに価値がある、ない」だけであって・・・。”
D先生 : “そのFOUNDERのいう価値って?”
QEU:FOUNDER : “金さえかければ欲しい品質がやすやすと手に入る。逆に言うと、今の時代において、いまだに品質が悪いのは「(手間を掛ける)価値がない」だけ・・・。もしも、お客が大したお金も払わずに高い品質を要求するのは「搾取行為」です。”
>寄付のお願い(click here)<
QEU:FOUNDER : “今回は、ゆるいシリーズではあるが、奥が深いんです。次も、もうちょっと、本件のつづきをやりましょう。”
~ まとめ ~
・・・ 前回のつづきです ・・・
C部長 : “まあ、見方を変えれば写真を撮ることには、それほどの価値があるということです。”
QEU:FOUNDER : “今、ものの価値というのは大きく変わろうとしているんです。現在において、価値というものは「シェアすること」によって得られるんです。”
C部長 : “それ(価値を得ること)が、DMAICSサイクルの本質ですね。ここでいう「写真をとる」ということは、「モノを作る」という言葉と本質的に同義です。シェアされるものですから・・・。”
QEU:FOUNDER : “一旦、(ものの)価値を得たら、その後は品質を得ることは簡単ですよ。そうだ、ちょっと(話題を)脱線をば・・・。YouTubeの動画を多種様々に見ていると、いろいろと思うことがあります。情報量が多い割に面白くない動画があります。そういうタイプのものって、「ナレーションとカメラワークが同期していない」です。その一方、動画の情報量が少なくても面白いものもある。結論とすれば、別に動画に含まれる情報量が価値を決めるわけではない。さっき述べたように、情報量の大きな作品にしたかったら、基本的な情景の画像を手に入れたら、あとは参考情報を見ながらナレーションをつければいいですからね・・・。”
D先生: “それは、特に旅行物(YOUTUBE)には「あるある」ですね・・・。面白い動画をとるのは、結局のところは「センス」なんじゃないでしょうか?おいしいものを食べて、すぐに視聴者が誰にも共感できるように「おいしいと反応できる」のは、それほど簡単じゃないです。”
QEU:FOUNDER : “ここでは、あんまり「芸術(センス)」の話にはしたくない。あらかじめ旅行に行く前に「準備をしておけばいい」んじゃない?結局のところ、それはYOUTUBERのもつ「質問力に帰着する」と思うんです。そして、「この質問力こそが価値を決める」と思います。”
C部長 : “FOUNDERの論点は、要するに、頭がよくないと価値あることができないと?”
QEU:FOUNDER : “いや。そうは思わないですよ。AIが人間(ユーザ)に協力して、ユーザの「質問力」を高めてくれればいいのだと思います。そして、社会の全員の質問力が高まれば、物の価値、社会の価値が上がってきます。そういう社会になればよいと思います。"











コメント
コメントを投稿