• Home
  • Services
  • About
  • Contact
  • Blog
  • 知財活動のROICへの貢献
  • 生成AIを活用した知財戦略の策定方法
  • 生成AIとの「壁打ち」で、新たな発明を創出する方法

​
​よろず知財コンサルティングのブログ

Claude_Fable_5.1はサイエンスエージェント化したのか

5/9/2026

0 Comments

 
2026年9月1日、Anthropicは最新モデル「Claude Fable 5.1」と「Claude Mythos 5.1」を発表しました。両モデルは、コーディング、AIエージェント、科学研究などの高度なタスクを対象としたモデルで、AIモデルの性能を測定する独立評価プラットフォーム「Artificial Analysis」が提供する、AIモデルの総合的な知能を評価するための複合指標(ベンチマーク)「Artificial Analysis Intelligence Index」では、66で最高評価です。(Claude Opus 5 (max) 63, GPT-5.6 Sol (max) 61)
「コーディングと知識労働のための世界最先端のモデルであり、その研究能力は、AIモデルが科学の進歩にどのように貢献していくかを示す初期段階の展望を示しています。」ということで、Claude_Fable_5.1は「サイエンスエージェント化した」という議論もあります。
Claude_Fable_5.1はサイエンスエージェント化したのか、について生成AIに深掘りさせましたのでご参照ください。なお、生成AIによる調査・分析結果は、公開された情報だけに基づくものであり、必ずしも実情を示したものではなく、誤った情報を含む可能性があることにご留意のうえ、ご参照ください。
 
 
Has Claude Fable 5.1 Become a Science Agent?
On September 1, 2026, Anthropic announced its latest models, Claude Fable 5.1 and Claude Mythos 5.1. Both are designed for advanced tasks, including coding, AI agent workflows, and scientific research. They achieved the highest score of 66 on the Artificial Analysis Intelligence Index, a composite benchmark for evaluating the overall intelligence of AI models, provided by the independent AI evaluation platform Artificial Analysis. For comparison, Claude Opus 5 (max) scored 63, while GPT-5.6 Sol (max) scored 61.
The models are described as “the world’s most advanced models for coding and knowledge work, with research capabilities that offer an early glimpse of how AI models will contribute to scientific progress.” This has prompted some to argue that Claude Fable 5.1 has evolved into a “science agent.”
I asked generative AI to explore in depth whether Claude Fable 5.1 has indeed become a science agent. Please review the findings with the understanding that the AI-generated research and analysis are based solely on publicly available information, may not necessarily reflect the actual state of affairs, and may contain errors.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
0 Comments

AIサムライによる特許網自動生成システム

4/9/2026

0 Comments

 
日経新聞「トヨタ系のAIサムライ、AIが特許網を生成 中小も技術保護しやすく」によれば、AIサムライは、大阪大学大学院情報科学研究科の鬼塚真教授の研究室との共同研究で出願を予定する基本特許に関連する特許を人工知能(AI)が提案し、特許網を自動生成するサービスを開発しました。新サービスは利用者が基本特許の内容を入力し、構築したい特許網の広さを決めるとAIが5分程度で周辺特許案を5〜6件提案します。また、特許網自動生成システムと外部のAIと接続する「MCP」の提供によって、現在年間360万円程度の利用料金も年60万〜120万円にすることで、導入社数を1年後に300社に増やすという積極策を発表しています。
このAIサムライの動きについて生成AIに深掘り調査させましたのでご参照ください。なお、生成AIによる調査・分析結果は、公開された情報だけに基づくものであり、必ずしも実情を示したものではなく、誤った情報を含む可能性があることにご留意のうえ、ご参照ください。
 
トヨタ系のAIサムライ、AIが特許網を生成 中小も技術保護しやすく
2026年9月3日
https://www.nikkei.com/article/DGXZQOUC26BH50W6A820C2000000/
 
 
AI Samurai’s Automated Patent Portfolio Generation System
According to the Nikkei article, “Toyota-Affiliated AI Samurai Uses AI to Generate Patent Portfolios, Making It Easier for SMEs to Protect Their Technologies,” AI Samurai has developed a service that uses artificial intelligence to propose related patent concepts and automatically build a patent portfolio around a core patent that a user plans to file. The service was developed through joint research with the laboratory of Professor Makoto Onizuka at Osaka University’s Graduate School of Information Science and Technology.
With the new service, users enter the details of a core patent and specify how broad they want the surrounding patent portfolio to be. The AI then proposes five or six related patent concepts in approximately five minutes. AI Samurai has also announced an aggressive growth strategy involving the provision of an “MCP” that connects its automated patent portfolio generation system with external AI services. By reducing the current annual fee of approximately ¥3.6 million to between ¥600,000 and ¥1.2 million, the company aims to increase the number of corporate users to 300 within one year.
I asked generative AI to conduct an in-depth investigation into AI Samurai’s latest initiative. Please refer to the findings for further details. Please note, however, that the research and analysis produced by generative AI are based solely on publicly available information and may not necessarily reflect the actual circumstances. They may also contain inaccurate information.

Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
0 Comments

Metaが「Muse Spark 1.3」を公開

4/9/2026

0 Comments

 
Metaは9月2日(現地時間)、 「Muse Spark 1.3」 を公開しました。8月初めにリリースされた「Muse Spark 1.2」に続くAIモデルで、エージェント的なタスクとコーディングの性能が向上しているということです。「Artificial Analysis Intelligence Index」では、「max」推論レベルでは63で、Anthropicの「Claude Fable 5」(63)と同じ、OpenAIの「GPT-5.6 Sol」(62)を上回るスコアを記録しており、2026年9月1日に発表された「Fable 5.1」(66)に及ばないもののトップレベルの性能を備えていることがわかります。
「Muse Spark 1.3」について生成AIに深掘り調査させましたのでご参照ください。なお、生成AIによる調査・分析結果は、公開された情報だけに基づくものであり、必ずしも実情を示したものではなく、誤った情報を含む可能性があることにご留意のうえ、ご参照ください。
 
Muse Spark 1.3
https://developer.meta.com/ai/models/muse-spark/
 
 
Meta Releases “Muse Spark 1.3”
Meta released Muse Spark 1.3 on September 2 (local time). Following Muse Spark 1.2, which was released in early August, the new AI model reportedly delivers improved performance in agentic tasks and coding.
On the Artificial Analysis Intelligence Index, Muse Spark 1.3 achieved a score of 63 at the “max” reasoning level. This matches Anthropic’s Claude Fable 5 at 63 and exceeds OpenAI’s GPT-5.6 Sol at 62. Although it falls short of Fable 5.1, which was announced on September 1, 2026, and scored 66, the results indicate that Muse Spark 1.3 offers top-tier performance.
I asked a generative AI system to conduct an in-depth investigation into Muse Spark 1.3. Please refer to the findings below. Please note, however, that the research and analysis generated by AI are based solely on publicly available information, may not necessarily reflect the full reality, and may contain inaccurate information.

Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
0 Comments

Gemini 3.8 Flash発表

3/9/2026

0 Comments

 
Googleは2026年9月2日、新しいGeminiモデル「Gemini 3.8 Flash」と「Gemini 3.8 Flash Cyber」を発表しました。「Gemini 3.8 Flash」は、コーディング、AIエージェント、複雑な推論を強化した汎用モデルです。一方の「Gemini 3.8 Flash Cyber」は、脆弱性の発見や修正など、サイバーセキュリティに特化したモデルです。3週間前にリリースしたFlash 3.7の勢いをそのままに、わずか6週間で3回目のFlashリリースとなります。
AIモデルの性能を測定する独立評価プラットフォーム「Artificial Analysis」が提供する、AIモデルの総合的な知能を評価するための複合指標(ベンチマーク)「Artificial Analysis Intelligence Index」では、「Gemini 3.8 Flash」は59で、昨日リリースされたClaude Fable 5.1が66で最高、Claude Opus 5 (max) 63, GPT-5.6 Sol (max) 61、最高水準には達していないようです。
「Gemini 3.8 Flash」について生成AIに深掘りさせましたのでご参照ください。なお、生成AIによる調査・分析結果は、公開された情報だけに基づくものであり、必ずしも実情を示したものではなく、誤った情報を含む可能性があることにご留意のうえ、ご参照ください。
 
Introducing Gemini 3.8 Flash and 3.8 Flash Cyber
https://blog.google/innovation-and-ai/models-and-research/gemini-models/3-8-flash-and-3-8-flash-cyber/
 
 
Gemini 3.8 Flash
On September 2, 2026, Google announced two new Gemini models: Gemini 3.8 Flash and Gemini 3.8 Flash Cyber. Gemini 3.8 Flash is a general-purpose model with enhanced capabilities in coding, AI agents, and complex reasoning. Gemini 3.8 Flash Cyber, meanwhile, is a specialized model designed for cybersecurity tasks such as identifying and remediating vulnerabilities. Building on the momentum of Flash 3.7, which was released three weeks earlier, this marks the third Flash release in just six weeks.
According to the Artificial Analysis Intelligence Index, a composite benchmark provided by Artificial Analysis, an independent platform for evaluating AI model performance, Gemini 3.8 Flash scored 59. This appears to place it below the very top tier, with Claude Fable 5.1, released yesterday, leading with a score of 66, followed by Claude Opus 5 (max) at 63 and GPT-5.6 Sol (max) at 61.
I asked generative AI to conduct an in-depth analysis of Gemini 3.8 Flash. Please see the results below. Please note, however, that the research and analysis generated by AI are based solely on publicly available information and may not necessarily reflect the actual situation. They may also contain inaccurate information.

Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
0 Comments

オムロン、知財AIエージェントを内製し特許関連工数を50%削減

3/9/2026

0 Comments

 
2026年9月2日付の日経クロステックの記事「オムロンがBedrock活用で知財AIエージェント内製、特許関連工数を50%減」によれば、オムロンは、AWSの生成AI基盤「Amazon Bedrock」などを活用し、先行技術調査と発明説明書作成を支援する知財AIエージェントを自社開発し、2026年1月頃から研究開発部門で利用を始め、これらの業務に要する工数を、少なく見積もっても50%削減したということです。
オムロンでは、発明考案のプロセスを、①発明の創出、②先行技術調査、③発明説明書の作成、の3段階に整理しています。エンジニアが本来注力したいのは発明の創出ですが、専門的な特許文献の確認や説明書作成に多くの時間を取られていました。
そこで、特許庁から取得した過去の出願データをAmazon S3に格納し、Amazon OpenSearch ServiceとAmazon Bedrockを組み合わせました。AIエージェントは、入力されたキーワードや仕様書を基に関連文献を検索・要約し、類似技術の整理、競合分析、発明説明書の作成まで支援します。
商用データベースやSaaSの利用も検討しましたが、データ量の制約や独自の業務プロセスへの対応を考え、内製を選択しました。また、出願前の機密情報を安全に扱うため、AWS上に共通AI基盤「RDinX」を構築し、セキュリティー設定やAIエージェントの挙動、プロンプトを一元的に管理しています。
今回の事例は、生成AIの活用が単なる文書要約から、発明発掘、調査、分析、説明書作成をつなぐ業務プロセスの再設計へ進んでいることを示しています。AIに判断を任せ切るのではなく
、人が結果を確認しながら、エンジニアを定型作業から解放し、発明創出に集中させる取り組みとして注目されます。
この記事の内容を生成AIに深掘り調査させましたのでご参照ください。なお、生成AIによる調査・分析結果は、公開された情報だけに基づくものであり、必ずしも実情を示したものではなく、誤った情報を含む可能性があることにご留意のうえ、ご参照ください。
 
オムロンがBedrock活用で知財AIエージェント内製、特許関連工数を50%減
2026.09.02
https://active.nikkeibp.co.jp/atcl/act/19/00798/080300002/
 
 
OMRON Develops In-House IP AI Agent, Cutting Patent-Related Workload by 50%
According to a Nikkei XTECH article published on September 2, 2026, titled “OMRON Develops In-House IP AI Agent Using Bedrock, Cutting Patent-Related Workload by 50%,” OMRON has developed its own intellectual property AI agent using Amazon Web Services’ generative AI platform, Amazon Bedrock, and other AWS services. The agent supports prior-art searches and the preparation of invention disclosure documents. OMRON began using it within its research and development divisions around January 2026 and has conservatively estimated that it has reduced the time required for these tasks by at least 50%.
OMRON divides the invention development process into three stages: (1) generating an invention, (2) conducting a prior-art search, and (3) preparing an invention disclosure document. Although engineers would ideally devote most of their attention to creating inventions, they had been spending considerable time reviewing highly technical patent documents and preparing disclosure documents.
To address this issue, OMRON stored historical patent application data obtained from the Japan Patent Office in Amazon S3 and combined Amazon OpenSearch Service with Amazon Bedrock. Based on keywords or technical specifications entered by users, the AI agent searches for and summarizes relevant documents. It also assists with organizing similar technologies, analyzing competitors, and preparing invention disclosure documents.
OMRON considered using commercial databases and software-as-a-service solutions. However, it chose to develop the system in-house because of restrictions on data volume and the need to accommodate its proprietary business processes. To securely handle confidential information before patent applications are filed, the company also built a shared AI platform called “RDinX” on AWS. The platform centrally manages security settings, AI agent behavior, and prompts.
This case illustrates how the use of generative AI is evolving beyond simple document summarization toward the redesign of entire workflows connecting invention discovery, prior-art research, analysis, and the preparation of disclosure documents. Rather than leaving decisions entirely to AI, OMRON has adopted a human-in-the-loop approach in which people review the AI-generated results. The initiative is particularly noteworthy because it frees engineers from routine tasks and enables them to concentrate more fully on creating inventions.
I asked generative AI to conduct an in-depth investigation of the matters discussed in this article. Please refer to the results below. Please note, however, that the research and analysis generated by AI are based solely on publicly available information. They may not necessarily reflect the actual circumstances and may contain inaccurate information.

Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
0 Comments

タイ、英国、韓国、中国、日本のソブリンAI

3/9/2026

0 Comments

 
世界各国の人工知能(AI)戦略が民間主導の技術開発から、国家が主導する社会実装へと移行しています。生成AIに、タイ、英国、韓国、中国、日本の5カ国を対象に調査させましたので、ご参照ください。なお、生成AIによる調査・分析結果は、公開された情報だけに基づくものであり、必ずしも実情を示したものではなく、誤った情報を含む可能性があることにご留意のうえ、ご参照ください。
 
Sovereign AI in Thailand, the United Kingdom, South Korea, China, and Japan
Artificial intelligence (AI) strategies around the world are shifting from private-sector-led technology development toward government-led implementation across society. I asked generative AI to examine the initiatives of five countries—Thailand, the United Kingdom, South Korea, China, and Japan—and have shared the results for your reference.
Please note that the research and analysis conducted by generative AI are based solely on publicly available information. They may not necessarily reflect the actual situation and may contain inaccuracies.

Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
0 Comments

「Claude Fable 5.1」と「Claude Mythos 5.1」発表

2/9/2026

0 Comments

 
2026年9月1日、Anthropicは最新モデル「Claude Fable 5.1」と「Claude Mythos 5.1」を発表しました。両モデルは、コーディング、AIエージェント、科学研究などの高度なタスクを対象としたモデルです。大きなポイントは、Fable 5.1とMythos 5.1が完全に別のモデルではなく、同じ基盤モデルを異なるセーフガードで提供していることです。
AIモデルの性能を測定する独立評価プラットフォーム「Artificial Analysis」が提供する、AIモデルの総合的な知能を評価するための複合指標(ベンチマーク)「Artificial Analysis Intelligence Index」では、66で最高評価です。(Claude Opus 5 (max) 63, GPT-5.6 Sol (max) 61)
「コーディングと知識労働のための世界最先端のモデルであり、その研究能力は、AIモデルが科学の進歩にどのように貢献していくかを示す初期段階の展望を示しています。」ということですので、楽しみです。
Claude Fable 5.1 と Claude Mythos 5.1について生成AIに深掘りさせましたのでご参照ください。なお、生成AIによる調査・分析結果は、公開された情報だけに基づくものであり、必ずしも実情を示したものではなく、誤った情報を含む可能性があることにご留意のうえ、ご参照ください。
 
Claude: Fable 5.1 and Mythos 5.1
https://www.anthropic.com/claude-fable-and-mythos-5-1
 
 
Anthropic Announces “Claude Fable 5.1” and “Claude Mythos 5.1”
On September 1, 2026, Anthropic announced its latest models, Claude Fable 5.1 and Claude Mythos 5.1. Both models are designed for advanced tasks such as coding, AI agent operations, and scientific research. One of the most notable aspects of the announcement is that Fable 5.1 and Mythos 5.1 are not entirely separate models; rather, they are based on the same underlying foundation model but are offered with different safeguards.
On the Artificial Analysis Intelligence Index—a composite benchmark developed by Artificial Analysis, an independent AI model evaluation platform, to assess the overall intelligence of AI models—the models received the highest score of 66. By comparison, Claude Opus 5 (max) scored 63, while GPT-5.6 Sol (max) scored 61.
Anthropic describes the model as “the world’s most advanced model for coding and knowledge work,” adding that its research capabilities offer an early glimpse of how AI models may contribute to scientific progress. This is certainly something to look forward to.
I asked generative AI to conduct an in-depth analysis of Claude Fable 5.1 and Claude Mythos 5.1, and I invite you to read the results. Please note that this research and analysis were generated by AI based solely on publicly available information. They may not necessarily reflect the actual situation and may contain inaccurate information.

Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
0 Comments

Gensparkが描く「仕事を終わらせるAI」

2/9/2026

0 Comments

 
2026年8月30日付のForbes JAPANの記事「AIの勝負はモデルでは終わらない、Genspark創業者が描く『次のAIサービス』の条件」を読みました。
この記事が示しているのは、生成AIの競争軸が、ChatGPT、Gemini、Claudeなどの「どのモデルが最も賢いか」という競争から、その知能を誰でも無理なく使い、実際の仕事の完了までつなげられるかという競争へ移り始めていることです。AIモデルの能力は大幅に向上しましたが、利用者が自分の立場、目的、過去の経緯を何度も説明し、複数のツールを行き来しなければならない仕事の流れは、まだ十分には変わっていないとGensparkの創業者たちは見ています。
Gensparkの考え方を象徴するのが、OpenAIやAnthropicなどが高性能な「エンジン」を作る企業であるのに対し、Gensparkは利用者を目的地まで運ぶ「車」を作る企業だ、という役割分担です。利用者がモデルの性能を比較して使い分けるのではなく、情報収集、判断、文書作成、共有、次の作業までを一続きにし、「どのモデルを使うか」ではなく「何を終わらせるか」を中心にサービスを設計しようとしています。
その中核となる構想が「SecondBrain」です。メール、会議、チャット、文書、アプリ、プロジェクト履歴などを継続的な文脈としてAIに持たせ、AIをその都度質問に答えるだけの道具から、利用者の仕事を長期的に理解する存在へ変えようとしています。Genspark 6.0では、このSecondBrainを基盤として、スーパーエージェントや各種作成ツール、複数のAIエージェントを組み合わせ、「記憶」「判断」「生成」「協働」を一つの流れとして結び付けています。
さらに、Gensparkが目指しているのは、AIをテキスト入力欄の中だけに閉じ込めないことです。会議の会話、移動中のメモ、対面での打ち合わせなど、これまで十分にデジタル化されていなかった情報も仕事の文脈として取り込み、その後の整理、連絡、資料作成、意思決定へつなげようとしています。最終的な目標は、「Googleを使える人ならGensparkも使える」というほど、専門的なプロンプト技術を必要としないAIです。
知財業務に引き寄せて考えると、今後のAIツールは、特許検索や要約、クレーム解釈の精度だけでは評価できなくなると考えられます。案件の背景、過去の検索式、審査経過、競合企業の動向、事業部門との議論などを継続的に理解し、調査から分析、報告書作成、意思決定支援までをつなげられるかが重要になります。一方で、AIが多くの業務文脈を保持するほど、機密情報の管理、アクセス権限、情報の正確性、判断過程の記録、人間による最終確認の重要性も高まります。これは、便利なAIツールを導入するという話を超え、企業の業務プロセスそのものを再設計する問題です。
「AIの勝負はモデルでは終わらない」という記事タイトルは、生成AI市場が「賢い回答を返すAI」から、「文脈を理解し、仕事を実際に前へ進めるAI」へ移行していることを端的に表しています。
このGensparkが描く「次のAIサービス」の条件と、AI産業、企業の業務変革、知財業務への影響について生成AIに深掘りさせましたので、ご参照ください。なお、生成AIによる調査・分析結果は、公開された情報だけに基づくものであり、必ずしも実情を示したものではなく、誤った情報を含む可能性があることにご留意のうえ、ご参照ください。
 
AIの勝負はモデルでは終わらない、Genspark創業者が描く「次のAIサービス」の条件
8/30(日)
https://news.yahoo.co.jp/articles/bfde0325786588fe3ba4f650d0505ec046cad97c
 
 
Genspark’s Vision of “AI That Gets the Job Done”
I read the Forbes JAPAN article published on August 30, 2026, titled “The AI Race Does Not End with Models: Genspark’s Founders on What the ‘Next Generation of AI Services’ Requires.”
The article suggests that the focus of competition in generative AI is beginning to shift away from the question of which model—ChatGPT, Gemini, Claude, or another—is the most intelligent. Instead, the competition is increasingly about who can make that intelligence effortlessly accessible to everyone and use it to bring actual work through to completion. AI models have become dramatically more capable, but Genspark’s founders believe that workflows have not yet changed sufficiently: users still have to repeatedly explain their position, objectives, and previous circumstances, while moving back and forth between multiple tools.
Genspark’s philosophy is symbolized by a clear division of roles. While companies such as OpenAI and Anthropic build high-performance “engines,” Genspark sees itself as building the “car” that takes users to their destination. Rather than requiring users to compare models and decide which one to use, Genspark aims to connect information gathering, judgment, document creation, sharing, and subsequent tasks into a single continuous workflow. Its services are therefore designed around “what the user wants to get done,” rather than “which model should be used.”
At the heart of this vision is a concept called “SecondBrain.” Genspark aims to give AI ongoing access to contextual information from emails, meetings, chats, documents, applications, and project histories. This is intended to transform AI from a tool that merely answers individual questions into an entity that understands the user’s work over the long term. In Genspark 6.0, SecondBrain serves as the foundation for combining its Super Agent, various creation tools, and multiple AI agents, integrating “memory,” “judgment,” “generation,” and “collaboration” into a single workflow.
Genspark also aims to prevent AI from being confined to a text input box. It seeks to incorporate information that has not previously been sufficiently digitized—such as conversations during meetings, notes made while traveling, and face-to-face discussions—into the context of a user’s work. That information can then be connected to subsequent organization, communication, document preparation, and decision-making. The ultimate goal is to create an AI system that requires no specialized prompting skills—one that is so easy to use that “anyone who can use Google can use Genspark.”
Viewed from the perspective of intellectual property work, future AI tools will no longer be evaluated solely on the accuracy of patent searches, summaries, or claim interpretation. What will matter is whether they can continuously understand the background of a matter, previous search queries, prosecution histories, competitors’ activities, and discussions with business divisions, and then connect research, analysis, report preparation, and decision-making support into one coherent process.
At the same time, the more business context an AI system retains, the more important it becomes to manage confidential information, control access permissions, verify the accuracy of information, maintain records of the decision-making process, and ensure final human review. This goes beyond merely introducing a convenient AI tool. It raises the broader question of how companies should redesign their business processes themselves.
The article’s title, “The AI Race Does Not End with Models,” succinctly captures the generative AI market’s transition from “AI that provides intelligent answers” to “AI that understands context and actually moves work forward.”
I asked generative AI to conduct an in-depth examination of the conditions that Genspark believes the “next generation of AI services” must satisfy, as well as the implications for the AI industry, corporate business transformation, and intellectual property work. Please refer to the resulting analysis. Please note, however, that the research and analysis generated by AI are based solely on publicly available information, may not necessarily reflect actual circumstances, and may contain inaccurate information.

Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
0 Comments

韓国政府「独自AI基盤モデル」2次評価の詳細点数を発表

2/9/2026

0 Comments

 
2026年8月27日、韓国政府(科学技術情報通信省)は、国家代表の人工知能(AI)を選ぶ「独自AI基盤モデル」プロジェクトの2次段階評価の詳細点数を発表し、各評価項目について参加企業すべての得点を明らかにしました。
SKテレコムが総合1位を獲得した一方で、国際的な性能指標で首位だったMotif Technologiesが脱落したことを受けたものです。
生成AIに本件を深掘りさせましたのでご参照ください。なお、生成AIによる調査・分析結果は、公開された情報だけに基づくものであり、必ずしも実情を示したものではなく、誤った情報を含む可能性があることにご留意のうえ、ご参照ください。
 
韓国政府、「独自AI基盤モデル」2次評価の全点数公開…総合1位はSKテレコム
8/31(月)
https://news.yahoo.co.jp/articles/5ab59f2485118c999dcef6460965197cc1cf58af
 
 
South Korean Government Releases Detailed Scores from the Second-Round Evaluation of Its “Sovereign AI Foundation Model” Project
On August 27, 2026, the South Korean government’s Ministry of Science and ICT announced the detailed results of the second-round evaluation for its “Sovereign AI Foundation Model” project, which is intended to select a nationally representative artificial intelligence model. The ministry disclosed the scores received by every participating company for each evaluation criterion.
The detailed scores were released following the outcome in which SK Telecom ranked first overall, while Motif Technologies was eliminated despite having achieved the highest score on international performance benchmarks.
I asked generative AI to conduct an in-depth analysis of this matter, and the results are presented below for your reference. Please note that the research and analysis generated by AI are based solely on publicly available information, may not necessarily reflect the actual circumstances, and may contain inaccurate information.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
0 Comments

防衛省の情報分析をSakana AIが支援

2/9/2026

0 Comments

 
2026年8月28日付のLedge.aiの記事「防衛省の情報分析をSakana AIがAI支援 公開情報で収集・分析・管理、AIエージェント技術を実証」を読みました。
Sakana AIは、防衛省と「総合分析業務に必要なAI機能の調査・実証」に関する契約を締結しました。契約額は約3億2,900万円で、AIエージェント技術などを活用し、情報収集の効率化、分析能力の向上、体系的な情報管理を検証します。
防衛省の総合分析業務では、政策や部隊運用の意思決定に必要な情報を、公開情報、電波情報、画像・地理情報などから収集・分析します。情報量が急増する一方、生成AIにはハルシネーションや情報管理上の課題があるため、単なる要約ではなく、情報源の確認や人間による検証を含む仕組みが重要になります。
今回の実証対象は、有料・無料を問わない「公開情報」です。したがって、秘密情報や電波情報をSakana AIが直接分析するものではなく、新聞、ウェブ、論文、統計などを活用するOSINTの高度化が中心とみられます。言語モデルだけでなく、エージェント型AI、推薦型AI、マルチモーダルAI、オンプレミス環境などの組み合わせも検討されます。
Sakana AIは2026年3月にも、防衛装備庁から指揮統制システム関連の研究を受注していますが、今回の案件は、安全保障政策全体を支える戦略的な情報分析が対象です。AIが分析官に代わるのではなく、膨大な公開情報の整理や関連性の発見を支援し、人間の意思決定を高度化できるかが注目されます。
この記事の内容を基に、生成AIに深掘りさせましたのでご参照ください。なお、生成AIによる調査・分析結果は、公開された情報だけに基づくものであり、必ずしも実情を示したものではなく、誤った情報を含む可能性があることにご留意のうえ、ご参照ください。
 
防衛省の情報分析をSakana AIがAI支援 公開情報で収集・分析・管理、AIエージェント技術を実証
https://ledge.ai/articles/sakana_ai_defense_information_analysis
 
 
Sakana AI Supports the Ministry of Defense’s Intelligence Analysis
I read an August 28, 2026 article published by Ledge.ai titled “Sakana AI to Support the Ministry of Defense’s Intelligence Analysis with AI—Demonstrating AI Agent Technology for the Collection, Analysis, and Management of Open-Source Information.”
Sakana AI has signed a contract with Japan’s Ministry of Defense for a project titled “Research and Demonstration of AI Capabilities Required for Comprehensive Analysis Operations.” The contract is valued at approximately ¥329 million. Through the use of AI agent technologies and other tools, the project will examine ways to make information collection more efficient, enhance analytical capabilities, and manage information systematically.
In its comprehensive analysis operations, the Ministry of Defense collects and analyzes the information needed for decision-making on policy and military operations from sources including publicly available information, signals intelligence, imagery, and geospatial information. While the volume of available information is growing rapidly, generative AI also presents challenges, including hallucinations and information-management risks. A system that goes beyond simple summarization—and incorporates source verification and human review—will therefore be essential.
The scope of this demonstration project is limited to “publicly available information,” regardless of whether access is free or paid. Sakana AI will therefore not directly analyze classified information or signals intelligence. Instead, the project appears to focus primarily on enhancing open-source intelligence, or OSINT, by using newspapers, websites, academic papers, statistical data, and other sources. In addition to language models, the project may examine combinations of agentic AI, recommendation systems, multimodal AI, and on-premises computing environments.
In March 2026, Sakana AI also received a research contract from Japan’s Acquisition, Technology & Logistics Agency concerning command-and-control systems. The latest project, however, focuses on strategic intelligence analysis supporting Japan’s broader national security policy. The key question is not whether AI will replace intelligence analysts, but whether it can help them organize vast amounts of publicly available information, identify previously unnoticed connections, and improve the quality of human decision-making.
Based on the contents of this article, I asked generative AI to conduct a more in-depth analysis. Please note that the resulting research and analysis are based solely on publicly available information. They may not necessarily reflect the actual circumstances and could contain inaccurate information.

Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
0 Comments

韓国・知識財産処「バイブコーディング業務革新」

1/9/2026

0 Comments

 
韓国の知識財産処(旧・特許庁)は、2026年8月27日に「バイブコーディング(Vibe Coding)ワークショップ」を開催し、生成AIを活用した業務革新の取り組みを本格化させました。専門的なプログラミング知識がなくても、自然語(日常の言葉)で開発意図を説明するだけで業務ツールを構築できる「バイブコーディング」により、行政業務の効率化とデジタル転換(DX)を加速させています。
この韓国の動きを生成AIに深掘りさせましたのでご参照ください。なお、生成AIによる調査・分析結果は、公開された情報だけに基づくものであり、必ずしも実情を示したものではなく、誤った情報を含む可能性があることにご留意のうえ、ご参照ください。
 
「言葉で説明すれば業務アプリ完成」…知識財産先、「バイブコーディング」でAI業務革新
https://m.news.nate.com/view/20260827n32612
 
 
South Korea’s Intellectual Property Authority Promotes “Business Innovation through Vibe Coding”
On August 27, 2026, South Korea’s intellectual property authority, formerly the patent office, held a “Vibe Coding Workshop,” marking the full-scale launch of its efforts to transform administrative operations through generative AI. Vibe coding enables users to build business tools simply by describing what they want to develop in natural, everyday language, without requiring specialized programming knowledge. Through this approach, the authority aims to improve the efficiency of administrative work and accelerate digital transformation (DX).
I asked generative AI to conduct a more in-depth investigation and analysis of these developments in South Korea. Please note, however, that the AI-generated research and analysis are based solely on publicly available information and may not necessarily reflect the actual circumstances. They may also contain inaccuracies. Please keep this in mind when reviewing the material.

Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
0 Comments

8月のAIニュース

1/9/2026

0 Comments

 
YouTube「【月刊AIニュース】8月のAIニュース46本をこの一本で全部解説」は、ちょっと長い(約1時間51分)ですが、8月のAI関連ニュースを良くまとめています。
この動画の内容を基に、生成AIに深掘りさせましたのでご参照ください。なお、生成AIによる調査・分析結果は、公開された情報だけに基づくものであり、必ずしも実情を示したものではなく、誤った情報を含む可能性があることにご留意のうえ、ご参照ください。
 
【月刊AIニュース】8月のAIニュース46本をこの一本で全部解説
https://www.youtube.com/watch?v=3lWH8rL3lIQ
 
 
AI News from August
The YouTube video, “[Monthly AI News] All 46 AI News Stories from August Explained in One Video,” is rather long—approximately one hour and 51 minutes—but provides an excellent overview of AI-related news from August.
Based on the content of this video, I asked generative AI to conduct a more in-depth investigation and analysis. Please note, however, that the results generated by AI are based solely on publicly available information and may not necessarily reflect the actual circumstances. They may also contain inaccuracies. Please keep this in mind when reviewing the material.

Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
0 Comments

PerplexityがNVIDIAと提携した Portable Computer

1/9/2026

0 Comments

 
2026年8月25日、Perplexityは、新しいローカルエージェントプラットフォーム「Portable Computer」をNVIDIA DGX Sparkに最適化することで、ユーザーのローカルハードウェア上においてパーソナルAIエージェントを実行できるようにすると発表しました。
本製品は、推論モデルや実行環境をユーザーのローカルハードウェアへ移行する「ローカルファースト」設計を採用し、データ主権の確保と運用コストの削減を同時に実現しています。動作にはNVIDIA DGX Sparkなどの高性能な計算基盤を必要とし、通常タスクは端末内で完結させつつ、高度な判断が必要な場合のみクラウドへ委譲するハイブリッド型の仕組みが特徴です。独自のPPLX 27Bモデルや決定論的なハーネスコードを組み合わせることで、小型モデルながら既存のオープンソース環境を凌駕する高い処理精度を実証しています。機密データを扱う企業やヘビーユーザーに対し、プライバシーと経済性を両立した新たな形態を提示する内容となっています。
この「Portable Computer」を生成AIに深掘りさせましたのでご参照ください。なお、生成AIによる調査・分析結果は、公開された情報だけに基づくものであり、必ずしも実情を示したものではなく、誤った情報を含む可能性があることにご留意のうえ、ご参照ください。
 
Perplexity、DGX Sparkに最適化されたローカルAIエージェントをリリース
https://ai.watch.impress.co.jp/docs/news/2136082.html
 
 
Perplexity’s “Portable Computer,” Announced in Partnership with NVIDIA
On August 25, 2026, Perplexity announced that it would enable users to run personal AI agents on their own local hardware by optimizing its new local agent platform, “Portable Computer,” for NVIDIA DGX Spark.
The platform adopts a “local-first” design that moves inference models and execution environments onto the user’s local hardware, simultaneously ensuring data sovereignty and reducing operating costs. It requires a high-performance computing platform such as NVIDIA DGX Spark and employs a hybrid architecture in which routine tasks are completed entirely on the local device, while only tasks requiring more advanced judgment are delegated to the cloud. By combining its proprietary PPLX 27B model with deterministic harness code, Perplexity has demonstrated high processing accuracy that surpasses existing open-source environments despite the model’s relatively compact size. The platform presents a new approach that combines privacy with cost efficiency for enterprises handling confidential data and for power users.
I asked generative AI to conduct an in-depth analysis of this “Portable Computer,” and I invite you to refer to the results. Please note, however, that the research and analysis generated by AI are based solely on publicly available information, may not necessarily reflect the actual circumstances, and may contain inaccuracies.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
0 Comments

知財訴訟の実際―当事者は何を考え、なぜ訴訟に至るのか

31/8/2026

0 Comments

 
2026年8月28日に行われた「【第4回サマリア知財フォーラム】知財訴訟の実際 -当事者は何を考え、訴訟に至るのか?-【講師】河部 康弘 氏: 河部法律事務所 所長/パテント・インテグレーション株式会社 顧問弁護士」のアーカイブ動画(約62分」を視聴しました。
パテント・インテグレーション株式会社の顧問弁護士でもある河部康弘氏が、豊富な訴訟経験を踏まえ、知財紛争における当事者の心理と意思決定を解説しています。河部氏が実質的に関与した、裁判所の事件番号が付された案件は約120件に上るとのことです。
日本では、勝訴者が実際に支払った弁護士費用を、そのまま敗訴者に負担させる制度にはなっていません。裁判には時間も費用もかかるため、権利者も被警告者も、本音では訴訟を避けながら、自らに有利な条件で解決したいと考えます。
権利者は「要求に応じなければ訴える」と圧力をかけ、被警告者は「訴えるならどうぞ」と強気に対応します。双方が相手の譲歩を待ち続け、ブレーキが効かなくなった結果、訴訟に突入する――これが講演でいう「チキンレース」です。
権利者が訴訟を提起するかどうかは、勝訴可能性だけでは決まりません。得られる損害賠償額、回収可能性、無効主張を受けるリスク、訴訟費用、社内負担、顧客やサプライチェーンへの影響、さらに経営者や発明者の感情まで含めて判断されます。
特に重要なのは、知財訴訟も事業活動の一環であるという点です。侵害を放置すれば、「この会社は権利行使をしない」と市場に受け取られ、他社の模倣を誘発するおそれがあります。一方で、相手の事業規模が小さく、自社への影響も限定的であれば、費用を考えて訴えない判断もあり得ます。
被警告者側も、侵害・無効の判断だけでなく、「相手は本当に訴えてくる会社か」を見ています。強力な無効資料を示したり、対象製品がほとんど売れていないことを開示したりして、権利者の訴訟意欲を低下させることも重要な交渉戦略になります。
講演で特に興味深かったのは、特許の実効的な権利範囲は、クレームの文言だけでは決まらないという指摘です。
過去の訴訟実績、訴訟予算の確保、訴訟費用保険への加入、知財訴訟に詳しい弁護士との連携などにより、「この会社は本当に訴えてくる」と相手に認識させれば、訴訟前の交渉力が高まります。強い明細書を作るだけでなく、必要な場合に権利行使できる体制と覚悟を示すことも、特許の牽制力を高める知財戦略といえます。
訴訟開始後は、裁判所の心証が示されることで侵害や無効の見通しが立ちやすくなり、中立的な裁判官が双方の事情を調整するため、和解も成立しやすくなります。
知財戦略は、特許を取得して終わりではありません。権利行使を実行できる組織、予算、社内意思決定、外部専門家との関係まで準備しておくことが、保有特許の価値と企業の交渉力を支える――非常に示唆に富む講演でした。
 
動画はこちらです。
https://www.youtube.com/watch?v=G9OetwbtBRE
 
 
The Reality of IP Litigation—What Do the Parties Think, and Why Do Disputes End Up in Court?
I watched the approximately 62-minute archived video of the seminar held on August 28, 2026, entitled “The 4th Summaria IP Forum: The Reality of IP Litigation—What Do the Parties Think, and Why Do Disputes End Up in Court?” The speaker was Yasuhiro Kawabe, Managing Partner of Kawabe Law Office and Legal Advisor to Patent Integration Co., Ltd.
Drawing on his extensive litigation experience, Mr. Kawabe, who also serves as legal advisor to Patent Integration Co., Ltd., explained the psychology and decision-making processes of parties involved in IP disputes. According to him, he has been substantially involved in approximately 120 cases that were formally assigned court case numbers.
In Japan, the losing party is not generally required to reimburse the prevailing party for the full amount of attorneys’ fees actually incurred. Because litigation is both time-consuming and costly, both rights holders and recipients of infringement warnings would, in truth, prefer to avoid going to court while still resolving the dispute on terms favorable to themselves.
The rights holder applies pressure by saying, “Comply with our demands, or we will sue,” while the accused party responds defiantly, “Go ahead and sue us.” Each side continues waiting for the other to back down, until the brakes cease to function and the dispute proceeds to litigation. This is what Mr. Kawabe described in the seminar as a “game of chicken.”
Whether a rights holder initiates litigation is not determined solely by the likelihood of prevailing. The decision also takes into account the potential amount of damages, the likelihood of actually recovering them, the risk of facing an invalidity challenge, litigation costs, the internal burden on the company, the impact on customers and supply chains, and even the emotions of management and the inventors involved.
One particularly important point is that IP litigation is itself part of business activity. If infringement is left unchallenged, the market may conclude that “this company does not enforce its rights,” potentially encouraging other companies to imitate its products or technology. On the other hand, when the other party’s business is small and the impact on the rights holder is limited, the rights holder may reasonably decide not to sue in view of the costs involved.
The recipient of an infringement warning also considers not only infringement and validity issues, but whether the rights holder is genuinely the kind of company that will bring a lawsuit. Presenting strong prior art that could invalidate the patent, or disclosing that the accused product has generated almost no sales, can reduce the rights holder’s incentive to litigate. This can therefore be an important negotiation strategy.
One of the most interesting observations in the seminar was that the effective scope of patent rights is not determined solely by the wording of the claims.
A company can strengthen its pre-litigation bargaining position by making the other party believe that it is genuinely prepared to sue. Such credibility may be established through a history of prior litigation, the allocation of an adequate litigation budget, the purchase of litigation expense insurance, and close cooperation with attorneys experienced in IP litigation. An effective patent strategy therefore involves not only drafting strong patent specifications, but also demonstrating that the company has both the organizational capability and the determination to enforce its rights when necessary. This can substantially enhance the deterrent effect of its patents.
Once litigation has commenced, the court may indicate its preliminary assessment, making it easier for the parties to evaluate the likely outcomes regarding infringement and validity. In addition, a neutral judge can take both parties’ circumstances into account and help bridge their differences, making settlement more likely.
An IP strategy does not end when a patent is granted. The value of a company’s patent portfolio and its negotiating power depend on advance preparation—including an organization capable of enforcing rights, an appropriate budget, internal decision-making procedures, and established relationships with external specialists.
It was an exceptionally insightful seminar.
 
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
0 Comments

自衛隊の通信網にNTT「IOWN」の中核技術導入

31/8/2026

0 Comments

 
防衛省は、全国の基地や部隊を結ぶ自衛隊の通信網に、NTTの次世代光通信基盤「IOWN(アイオン)」の中核技術であるオールフォトニクス・ネットワーク(APN)を導入する方針を固め、次期防衛力整備計画に明記し、2027年度予算案に関連費を計上、2028年度以降の本格運用を目指すということです。
この件について生成AIに深掘りさせましたのでご参照ください。なお、生成AIによる調査・分析結果は、公開された情報だけに基づくものであり、必ずしも実情を示したものではなく、誤った情報を含む可能性があることにご留意のうえ、ご参照ください。
 
自衛隊情報基盤にIOWN導入 AI活用へ大容量通信、次期計画に明記
2026/8/29
https://www.nikkei.com/article/DGXZQOUA276CL0X20C26A8000000/
 
 
Japan Self-Defense Forces to Introduce NTT’s Core IOWN Technology into Communications Network
Japan’s Ministry of Defense has reportedly decided to introduce the All-Photonics Network (APN)—a core technology of NTT’s next-generation optical communications infrastructure, IOWN—into the communications network connecting Japan Self-Defense Forces bases and units nationwide. The plan is expected to be incorporated into the next Defense Buildup Program, with related expenditures included in the fiscal 2027 budget proposal and full-scale operation targeted for fiscal 2028 or later.
I asked generative AI to conduct an in-depth investigation into this development. Please refer to the findings below. Please note, however, that the research and analysis generated by AI are based solely on publicly available information and may not necessarily reflect the actual circumstances. They may also contain inaccurate information.

Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
0 Comments

OpenAIが2026年末までに社内でAGIを実現する

30/8/2026

0 Comments

 
2026年8月26日TIME誌公開のインタビュー記事で、OpenAIのサム・アルトマンCEOは「2026年末までに社内でAGI(汎用人工知能)を実現するという予測を示しました。
この件について、生成AIに深掘りさせましたのでご参照ください。なお、生成AIによる調査・分析結果は、公開された情報だけに基づくものであり、必ずしも実情を示したものではなく、誤った情報を含む可能性があることにご留意のうえ、ご参照ください。
 
OpenAIは2026年末までにAGIと呼ばれるシステムを開発するだろうとサム・アルトマンCEOが回答
https://gigazine.net/news/20260827-openai-agi/
 
 
OpenAI to Achieve AGI Internally by the End of 2026
In an interview published by TIME on August 26, 2026, OpenAI CEO Sam Altman predicted that the company would achieve artificial general intelligence (AGI) internally by the end of 2026.
I asked generative AI to conduct an in-depth investigation into this topic. Please note that the resulting research and analysis are based solely on publicly available information, may not necessarily reflect the actual circumstances, and may contain inaccuracies.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
0 Comments

AIが機器を操作する共通仕様MHS

30/8/2026

0 Comments

 
Paragraph. 編集するにはここをクリック.​8月30日 AIが機器を操作する共通仕様MHS
 
Anthropicは2026年8月27日(現地時間)、AIエージェントが物理的な機器を安全に操作するための共通仕様「Model Hardware Standard」(MHS)の研究プレビューを、最初の科学研究機関および先進的な製造業者グループに公開すると発表しました。MHSにより、AIエージェントは顕微鏡、液体ハンドラー、ロボットアームなどの複数の実験機器や製造機器を並行して操作し、日常的な創薬実験から量子コンピュータ上のレーザー校正まで、複雑なタスクを実行できるようになります。
Anthropicが2024年に公開した「MCP (Model Context Protocol)」がAIエージェントとツールやデータを接続するためのプロトコルなのに対し、MHSは物理デバイスを扱うための仕様です。
この「Model Hardware Standard」(MHS)について生成AIに深掘りさせましたのでご参照ください。なお、生成AIによる調査・分析結果は、公開された情報だけに基づくものであり、必ずしも実情を示したものではなく、誤った情報を含む可能性があることにご留意のうえ、ご参照ください。
 
AIと機器接続の共通規格開発 米アンソロピック、時間短縮へ
8/28(金)
https://news.yahoo.co.jp/articles/04807df85975846361a9d4bc1662c0b617c30dd4
 
 
MHS: A Common Standard for AI to Operate Physical Equipment
On August 27, 2026, local time, Anthropic announced that it would release a research preview of the Model Hardware Standard (MHS)—a common specification designed to enable AI agents to operate physical equipment safely—to an initial group of scientific research institutions and advanced manufacturers.
MHS will enable AI agents to operate multiple pieces of laboratory and manufacturing equipment in parallel, including microscopes, liquid-handling systems, and robotic arms. This will allow them to perform complex tasks ranging from routine drug-discovery experiments to laser calibration on quantum computers.
Whereas Anthropic’s Model Context Protocol (MCP), released in 2024, is a protocol for connecting AI agents to tools and data, MHS is a specification for interacting with physical devices.
I asked generative AI to conduct an in-depth analysis of the Model Hardware Standard (MHS). Please see the results below. Please note that the research and analysis produced by generative AI are based solely on publicly available information and may not necessarily reflect the actual situation. They may also contain inaccuracies.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
0 Comments

Tencent「Hy4 preview」発表

29/8/2026

0 Comments

 
中国のテック大手Tencentは、2026年8月28日に新世代のフラッグシップ大規模言語モデル「Hy4 preview」(テンセント混元/Hunyuan Hy4)を発表し、オープンソース(オープンウェイト)として一般公開しました。
Hy4 previewはMixture-of-Experts、略してMoEという方式で、すべての計算部分を毎回動かすのではなく、入力に応じて必要な専門部分だけを動かします。総パラメータ数は7700億ですが、1トークンの処理で稼働するのは490億。
このHy4-previewについて生成AIに深掘りさせましたのでご参照ください。なお、生成AIによる調査・分析結果は、公開された情報だけに基づくものであり、必ずしも実情を示したものではなく、誤った情報を含む可能性があることにご留意のうえ、ご参照ください。

​Tencent Announces “Hy4 Preview”
Chinese technology giant Tencent announced its next-generation flagship large language model, “Hy4 Preview”—also known as Tencent Hunyuan Hy4—on August 28, 2026, and made it publicly available as an open-source, or more precisely open-weight, model.
Hy4 Preview employs a Mixture-of-Experts (MoE) architecture. Rather than activating the entire model for every computation, it selectively activates only the expert components needed for each input. Although the model has 770 billion parameters in total, only 49 billion parameters are activated when processing each token.
I asked generative AI to conduct an in-depth analysis of Hy4 Preview, and the results are provided for your reference. Please note, however, that the research and analysis generated by AI are based solely on publicly available information, may not necessarily reflect the actual situation, and may contain inaccuracies.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
0 Comments

島津製作所知財業務の生成AI活用第二段階

28/8/2026

0 Comments

 
2026年8月5日に行われた関西知財セミナー「生成 AI が知財業務をどう変えるか ~現状と展望~」では、阿久津 好二(あくつ こうじ)氏(株式会社 島津製作所 知的財産部 部長)が、島津製作所における生成AIを活用した知的財産業務の変革プロセスを解説しました。
これまで初期の「ファーストステージ」(AIを事務作業の自動化やコスト削減のための補助ツールとして位置づけ、劇的な効率化を実現)の話が多かったのが、今回は、続く「セカンドステージ」(単なる省力化を超え、AIを基盤としたビジネスプロセスそのものの再構築)の話がメインでした。AIエージェントの活用と人間による高度な意思決定を組み合わせ、企業価値の創造に直結する組織へと進化させる将来構想が示されていました。
島津製作所における生成AIを活用した知的財産業務の変革プロセスについて、講演資料と私のメモをベースに、生成AIに深掘りさせましたので、ご参照ください。なお、生成AIによる調査・分析結果は、公開された情報だけに基づくものであり、必ずしも実情を示したものではなく、誤った情報を含む可能性があることにご留意のうえ、ご参照ください。
 
 
The Second Stage of Generative AI Adoption in Shimadzu Corporation’s Intellectual Property Operations
At the Kansai Intellectual Property Seminar titled “How Generative AI Is Transforming Intellectual Property Operations: Current Status and Future Outlook,” held on August 5, 2026, Mr. Koji Akutsu, General Manager of the Intellectual Property Department at Shimadzu Corporation, explained the company’s process for transforming its intellectual property operations through the use of generative AI.
Until now, much of the discussion has focused on the initial “first stage,” in which AI is positioned as a support tool for automating administrative tasks and reducing costs, thereby achieving dramatic improvements in efficiency. This time, however, the main focus was on the subsequent “second stage”: moving beyond simple labor savings to redesign the business processes themselves around AI. The presentation outlined a future vision in which the use of AI agents is combined with sophisticated human decision-making, enabling the intellectual property organization to evolve into one that directly contributes to the creation of corporate value.
Based on the presentation materials and my own notes, I asked generative AI to conduct an in-depth analysis of Shimadzu Corporation’s process for transforming its intellectual property operations through generative AI. Please refer to the results. Please note, however, that the research and analysis generated by AI are based solely on publicly available information, may not necessarily reflect the actual circumstances, and may contain inaccuracies.

Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
0 Comments

究極のコスト効率を目指したQwen3.8-Flash-Next

28/8/2026

0 Comments

 
2026年8月26日、AlibabaのAI研究チームであるQwenがAIモデル「Qwen3.8-Flash-Next」を公開しました。Qwen3.8-Flash-NextはQwen4シリーズで採用される予定の次世代アーキテクチャを用いて開発されており、学習コストを抑えつつ高性能なモデルを構築することに成功しています。
生成AIにQwen3.8-Flash-Nextを深掘りさせましたので、ご参照ください。なお、生成AIによる調査・分析結果は、公開された情報だけに基づくものであり、必ずしも実情を示したものではなく、誤った情報を含む可能性があることにご留意のうえ、ご参照ください。
 
Qwen3.8-Flash-Next: A New Architecture, Towards Ultimate Cost-Efficiency
https://qwen.ai/blog?id=qwen3.8-flash-next
 
 
Qwen3.8-Flash-Next: Pursuing Ultimate Cost Efficiency
On August 26, 2026, Qwen, Alibaba’s AI research team, released a new AI model called Qwen3.8-Flash-Next. The model was developed using a next-generation architecture expected to be adopted in the Qwen4 series, successfully achieving high performance while keeping training costs low.
I asked a generative AI to conduct an in-depth analysis of Qwen3.8-Flash-Next. Please take a look. Please note, however, that the research and analysis generated by AI are based solely on publicly available information, may not necessarily reflect the actual circumstances, and may contain inaccuracies.

Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
Your browser does not support viewing this document. Click here to download the document.
0 Comments
<<Previous
Forward>>

    著者

    萬秀憲

    アーカイブ

    April 2026
    March 2026
    February 2026
    January 2026
    December 2025
    November 2025
    October 2025
    September 2025
    August 2025
    July 2025
    June 2025
    May 2025
    April 2025
    March 2025
    February 2025
    January 2025
    December 2024
    November 2024
    October 2024
    September 2024
    August 2024
    July 2024
    June 2024
    May 2024
    April 2024
    March 2024
    February 2024
    January 2024
    December 2023
    November 2023
    October 2023
    September 2023
    August 2023
    July 2023
    June 2023
    May 2023
    April 2023
    March 2023
    February 2023
    January 2023
    December 2022
    November 2022
    October 2022
    September 2022
    August 2022
    July 2022
    June 2022
    May 2022
    April 2022
    March 2022
    February 2022
    January 2022
    December 2021
    November 2021
    October 2021
    September 2021
    August 2021
    July 2021
    June 2021
    May 2021
    April 2021
    March 2021
    February 2021
    January 2021
    December 2020
    November 2020
    October 2020
    September 2020
    August 2020
    July 2020
    June 2020

    カテゴリー

    All

    RSS Feed

Copyright © よろず知財戦略コンサルティング All Rights Reserved.
サイトはWeeblyにより提供され、お名前.comにより管理されています
  • Home
  • Services
  • About
  • Contact
  • Blog
  • 知財活動のROICへの貢献
  • 生成AIを活用した知財戦略の策定方法
  • 生成AIとの「壁打ち」で、新たな発明を創出する方法