2024年6月29日 星期六

日本AI發明專利的適格性筆記之二 - JPO 2024年更新範例

延續前篇,本篇筆記JPO與時俱進更新的AI相關發明指南與範例:https://www.jpo.go.jp/e/system/laws/rule/guideline/patent/document/ai_jirei_e/jirei_add2024_e.pdf

定義AI-related技術是以機器學習通過IoT技術收集到的大數據,機器學習如一種深度學習演算法,用以訓練模型,可以針對已知/未知的數據產生結果。


列舉三個範例,分別是針對可實施性、可專利性,以及明確性的案例。

Case 5:
可實施性、說明書支持要件

發明提出一個通過機器學習法學習化學結構與已知成分(Fluorescent Compoun)的特性形成的訓練集,預測出一些具有特定特性的新的成分(A, B),其中成分A已經獲得驗證,但成分B沒有。

Claim 1描述的是通過模型預測出具有特定特性的新的成分,附屬項界定A與B兩個不同的成分。而這樣的發明專利被駁回,駁回理由是說明書並未敘明產生這些成分的方法,相關領域技術人員基於已知技術也無法理解,因此違反可實施性要件。也就是通過訓練得出的模型預測出的成分並非一般知識可以達成。


下圖顯示發明關於利用機器學習法學習化學結構與特性形成的訓練集,形成模型,經輸入螢光特性後,可以輸出新的成分。

核駁理由主要是以申請時該領域知識無法理解如何以訓練出的模型預測新的模型的技術,因此有部分是不符可實施性要件,然而有一項範圍是沒有遭遇核駁的,即Claim 2,主要理由是Claim 2描述的成分A有經過實驗證實,並且說明書的描述是可以據以實施Claim 2。


Case 8:專利適格性要件

一開始即說明AI相關發明中Claim 1所描述的“訓練集/training data”本身僅是資訊表示(mere presentation of information),不滿足專利適格性要求。而Claim 2描述了產生形成訓練集的影像的方法,可以理解的是運用硬體以軟體處理資訊是具體的技術,也就滿足專利適格性要求。以下圖式中的流程圖就是Claim 2特徵。

(編按,也就是說,至少日本專利局,認為只要講清楚AI如何運作就會具備專利適格性,如以下示意圖。)

Claim 1僅屬於資訊表示,因此不具專利適格性。Claim 2描述了產生訓練集的方法,其中需要軟硬體一起完成,具有專利適格性。

列舉Claim 2:
[Claim 2] A method for generating images for training data for training a discriminator that determines the area of surgical instruments in target images when the target images are input, the method comprising 
a step of obtaining radiological images including the human body and surgical instrument images indicating surgical instruments; and a step of generating images for training data corresponding to the target images by combining the surgical instrument images with the radiological images, performed by an image generation device for training data.


Case 10:明確性要件

專利請求項僅記載訓練模型,但沒有提到電腦,不清楚這個訓練集是一個程式指令,且沒有描述一個方法範圍或是產品範圍,因此不明確。

以下有專利範圍的解釋,也就是此例的Claim 1或2界定一個訓練集是不明確的。


此案例要克服明確性,請求項可以改寫:"The computer incorporated in the copier comprises a CPU and a memory that stores the trained model trained by the learning device, and if the trained model is a program module, the trained model may be stored in the memory.",這解決了專利標的不明確的問題。


Ron

2024年6月26日 星期三

日本AI發明專利的適格性筆記之一 - ChatGPT問答錄

前言:

其實說到AI發明,也僅是搭上目前很夯的話題,AI有個特色是 - "說不清楚",但本質上就是"軟體發明",也就是常說的"電腦實現的發明/computer-implemented invention/CII",因此有關軟體發明的專利適格性議題應該都適用在AI相關的發明。但無論如何,AI相關發明的專利適格性仍是值得一談的議題,畢竟AI還是有點意思。

AI相關發明在日本專利實務有何見解?大哉問,先問ChatGPT:

這些可參考,問了「ChatGPT」,它給我的答案算是很嚴肅的一般知識加上一些常識,也就是軟體發明該注意的,仍有一些可參考資料,如下標註:

 

In Japan, the patentability of AI-related inventions follows the general principles of patent law, but there are specific considerations to keep in mind:

1. **Statutory Requirements**: To be patentable, an AI-related invention must meet the three basic requirements: novelty, inventive step (non-obviousness), and industrial applicability.
(常識)

2. **Technical Aspect**: The invention must be a "creation of technical ideas utilizing a law of nature." Pure mathematical algorithms, abstract ideas, or mental acts are not patentable unless they have a technical character or are implemented in a way that produces a technical effect.
(還算可以,就是要有技術效果的特徵)

3. **AI Algorithms**: AI algorithms themselves, if claimed as such, are generally not patentable unless they are part of a specific technical application or contribute to a technical improvement. For example, a machine learning algorithm used in a medical imaging device to improve diagnostic accuracy may be patentable.
(還算不錯的答案)

4. **Application Format**: The Japan Patent Office (JPO) encourages applicants to describe AI-related inventions in terms of their technical problem and solution. Detailed descriptions of how the AI contributes to solving a technical problem and the specific technical implementation can enhance the chances of patentability.
(要解決技術問題)

5. **Examples of Patentable AI Inventions**:
(我覺得是不錯的說明)
- AI used in a new type of control system for machinery.
- AI algorithms implemented in a specific hardware environment that results in a technical improvement.
- Specific applications of AI in fields like healthcare, manufacturing, or transportation that demonstrate a technical effect.

6. **Examination Guidelines**: The JPO provides examination guidelines specific to AI-related inventions. These guidelines help in assessing whether an AI invention meets the criteria for patentability, emphasizing the importance of a detailed technical description and the demonstration of a technical effect.

7. **Patent Strategy**: When filing an AI-related patent in Japan, it's crucial to:
(很不錯的結論)
- Clearly define the technical problem being solved.
- Detail the technical means by which the AI solution is implemented.
- Highlight any technical effects or improvements resulting from the AI implementation.

In summary, while AI algorithms as standalone inventions are generally not patentable in Japan, AI-related inventions that demonstrate a clear technical contribution and solve a specific technical problem can be patented. The focus should be on the practical application and technical details of the AI solution.


繼續問以上答案是否有來源?ChatGPT給我以下內容(我希望它用email寄給我,但是目前尚無此功能):


Subject: Information on Patentability of AI-Related Inventions in Japan

Hi Ronpan,

Here is the information on the patentability of AI-related inventions in Japan:

1. **Japan Patent Office (JPO) Official Website**:
- The JPO provides comprehensive guidelines and resources regarding patent examination, including specific guidelines for AI and software-related inventions.
- [Japan Patent Office (JPO)](https://www.jpo.go.jp/e/)

2. **JPO Examination Guidelines**:
- The JPO has published examination guidelines that cover various aspects of patentability, including specific sections on AI and software inventions.
- [JPO Examination Guidelines](https://www.jpo.go.jp/e/system/laws/rule/guideline/patent/index.html)

3. **Intellectual Property High Court of Japan**:
- For legal precedents and interpretations related to AI patents, you might find relevant case law from the Intellectual Property High Court of Japan.
- [Intellectual Property High Court](http://www.ip.courts.go.jp/eng/)

4. **Japanese Patent Attorneys and Law Firms**:
- Consulting with Japanese patent attorneys or law firms specializing in intellectual property can provide personalized and detailed advice.
- Some well-known firms include:
- Nakamura & Partners: [Nakamura & Partners](http://www.nakapat.gr.jp/english/)
- Aoyama & Partners: [Aoyama & Partners](http://www.aoyamapat.gr.jp/)

5. **Academic and Industry Publications**:
- Research papers, articles, and industry reports on AI and IP law can provide deeper insights.
- Websites like Google Scholar or legal databases such as LexisNexis and Westlaw can be useful.

6. **Professional Organizations and Conferences**:
- Organizations like the International Association for the Protection of Intellectual Property (AIPPI) often hold seminars and publish papers on the latest trends in IP law, including AI-related patents.
- [AIPPI Japan](https://www.aippi.org/aippi-japan/)

I hope this helps with your research and any AI-related patent endeavors!

Best regards,


整理以上來源:

[Japan Patent Office (JPO)](https://www.jpo.go.jp/e/)


[JPO Examination Guidelines](https://www.jpo.go.jp/e/system/laws/rule/guideline/patent/index.html)


[Intellectual Property High Court](http://www.ip.courts.go.jp/eng/)


Nakamura & Partners: [Nakamura & Partners](http://www.nakapat.gr.jp/english/)


Aoyama & Partners: [Aoyama & Partners](http://www.aoyamapat.gr.jp/)


[AIPPI Japan](https://www.aippi.org/aippi-japan/)



-------------------------------------------------------------------

問Google,則看到一篇不錯的內容:https://books.openedition.org/putc/15392?lang=en,看不出來本篇撰寫日期,但其中有引用2021年案例,應該還算新的知識。


-------------------------------------------------------------------


(以上僅筆記,編排有點混亂,加上內容很多,需要時間消化,會需要分幾篇來寫!)

Ron

2024年6月25日 星期二

如何評估澳洲"電腦實現發明"的專利適格性? - Commissioner of Patents v RPL Central Pty Ltd [2015] FCAFC 177

本篇討論澳洲專利適格性澳洲聯邦法院(FEDERAL COURT OF AUSTRALIA)案例之一 - Commissioner of Patents v RPL Central Pty Ltd [2015] FCAFC 177


澳洲專利局的講義(Manner of Manufacture Information and Computer Technology Patent Office Practice):


在RPL案中,申請專利範圍涉及資訊收集方法與系統,收集的資訊是關於評估個人工作能力的資訊,專利範圍包括以電腦從網路取得評估工作能力的標準、處理這些標準以自動產生評估用的問題、通過網路呈現問題給評估工作能力的對象,再從這些人接受回答問題的答案。


以上描述的系爭發明僅是由電腦執行,也就是簡單地將所述方法(business method)由一般電腦執行是不具備專利適格性,除非:此發明是以特定的方式讓電腦實現所述方法。這裡給了答案,就是發明應超越電腦做了甚麼("what"),而應描述電腦如何運作以改善電腦技術("how")。反觀系爭專利,並非主張關於電腦程式或運作的發明或是創新,其中創新是方法步驟本身,也就是一種方案或是商業方法(scheme or business method)。


如此,引述前例Myriad指出要判斷方法本質,仍應參考先前技術,判斷發明是否具備專利適格性,應整體地參照專利說明書相關先前技術來評估專利適格性。


在審查專利時,專利審查委員應先考量發明本質,通過以下步驟來看:(1)發明如何運作?(2)發明所解決的問題是?(3)執行發明的結果為何?(4)在其優先權日當下的技術水平?(5)發明相對先前技術增加了什麼?(6)發明優點為何?

接著,專利審查委員可通過以下步驟判斷發明是否是可專利標的?
(1)發明的技術貢獻為何?(2)發明是否以電腦解決技術問題?或不是以電腦執行?(3)發明是否能改善電腦功能,而非僅是資料處理而已?(4)是否發明僅是以一般目的電腦執行?(5)是否點僅是實現方法的中間過程,而沒有對發明有任何實質助益?

-----------------------------------------------------------------------
案件資訊:
上訴人/專利權人:RPL CENTRAL PTY LTD
被上訴人:COMMISSIONER OF PATENTS
判決日期:11 December 2015

從上訴人與被上訴人資訊可知,這是一個行政上訴案,系爭專利在澳洲專利局審定不予專利,理由是發明不符專利適格性,專利權人提起訴願維持原審查意見,而專利權人不服上訴澳洲聯邦法院(Federal court of Australia)(判決:https://jade.io/article/351118 / "Research Affiliates LLC v Commissioner of Patents [2014] FCAFC"),維持系爭案發明不符專利適格性的決定,案件繼續上訴澳洲聯邦法院第二審(判決:https://jade.io/article/421354)。


系爭案為澳洲創新專利申請案No. 2009100601,名稱為自動收集技能與知識證據的方法與系統(Method and System for Automated Collection of Evidence of Skills and Knowledge),相關法條為:18(1A)(a) of the Patents Act 1990(如本文後段)。



系爭案的內容關於評估每個人的已知技能、知識與經驗,以確認是否適合參加特定訓練課程,或是判斷是否滿足取得特定課程資格的要求。(編按,就專利權人RPL CENTRAL PTY LTD而言,系爭案發明就是該公司提供評估客戶就業類別的服務,如果以"美國專利"眼光來看是個十分"資訊收集、處理與表示"的"抽象"的發明。)

本案在專利審查階段,審查委員表示系爭發明不屬於專利法規定的可專利標的,發明並未形成任何有用的產品:"It appears that the active substances of the invention are known. Claims 1 to 3 are not therefore directed to any manner of manufacture in that they are claims to the mere use of known substances - which use also does not result in any vendible product."

法院從系爭案說明書理解發明內容,發明包括呈現出自動產生的問題,接收使用者回答的內容,之後藉由電腦網頁介面評估使用者的技能、知識與經驗。如下圖,系統包括評估伺服器,用以收集評估個人能力的資訊,伺服器有處理器、記憶體,其中儲存程式指令,經執行可收集資料,產生網頁介面,並提供遠端存取。其中還有一個儲存裝置,能夠長時間儲存涉及資格標準判斷的資料。最後是可藉由網頁介面提供使用者通過自動產生的問題而自動評估自己的能力。


系爭案發明流程可參考下圖:


列舉Claim 1如下:

claim 1: A method of gathering evidence relevant to an assessment of an individual's competency relative to a recognised qualification standard, including the steps of:

a computer retrieving via the Internet from a remotely-located server a plurality of assessable criteria associated with the recognised qualification standard, said criteria including one or more elements of competency, each of which is associated with one or more performance criteria;

the computer processing the plurality of assessable criteria to generate automatically a corresponding plurality of questions relating to the competency of an individual to satisfy each of the elements of competency and performance criteria associated with the recognised qualification standard;

an assessment server presenting the automatically-generated questions via the Internet to a computer of an individual requiring assessment; and

receiving from the individual via said individual's computer a series of responses to the automatically-generated questions, the responses including evidence of the individual's skills, knowledge and/or experience in relation to each of the elements of competency and performance criteria,

wherein at least one said response includes the individual specifying one or more files stored on the individual's computer, which are transferred to the assessment server.

Claim 1描述的發明包括使用電腦通過網際網路取得資料、處理資料產生評估使用者能力的問題、呈現問題給使用者,並取得使用者回答的內容。

Claim 5則界定收集關於評估個人能力證據的系統,系統包括至少一個伺服器電腦、記憶體或儲存裝置,記憶體或儲存裝置包括通過伺服器電腦中微處理器執行的指令,用於:呈現問題、接收個人回應問題的答案,以及儲存答案。

經參照系爭案說明書與各主要專利範圍內容與解釋,主審法官(primary Judge)評估本案特徵如下:
  1. 發明具有可利用性,產生有用的結果;
  2. 發明克服了判斷適當課程給使用者的困難;
  3. 發明符合「可販售商品」的測試(vendible product test)(也就是澳洲專利法/判例要求可專利的條件之一是發明應關於可販售的商品);
  4. 根據發明提供的方案是能將收集到的資料轉換為問題與答案格式,並引導使用者通過這些資訊提供訓練課程,法院判斷這個發明如人為的工作
  5. 發明包括了電腦,與發明本身關係複雜,在網路環境中執行通訊的工作;
  6. 發明的評估伺服器包括取得資訊、處理、產生問題等電腦實現的功能。

THE DECISION OF THE DELEGATE
根據上述參照說明書、申請專利範圍與先前技術的考量,法院認為,系爭案發明確實不屬於專利法規定的"manner of manufacture",但法院仍知悉,物理上/具體的可觀察的結果並不需要是具體產品,而可以是在具體裝置上運行的方法與應用,並且,僅是存在具體的效果也不足以具備可專利性!!!

該代表承認,物理上可觀察到的最終結果不必是有形產品的意義上的,而可以是在物理設備中方法的應用和操作。 然而,他也觀察到,僅僅存在物理效應並不足以賦予專利性。

"The Delegate acknowledged that a physically observable end result need not be in the sense of a tangible product but may be in the application and operation of a method in a physical device.  However, he also observed that the mere presence of a physical effect is not sufficient to confer patentability."

即便是專利審查委員,仍是認同系爭案發明在電腦上提出了新的用途(new use),也明確地連結特定機器,而且沒有電腦也無法執行所描述的方法,也就是電腦整合在整個發明中。而發明是自動執行在一評估伺服器中,但說明書並未描述相關程式/流程,因此並未證明發明的獨創性(ingenuity),若伺服器中運行的步驟是一個自動程序,判定僅是"簡單地輸入文字和問題與一些標準的文字工作"。

當發明僅是將人的工作成為電腦中運行的自動程序,就會判定是以一般已知機器實現的抽象概念,就不足以超越抽象概念而判定符合"manner of manufacture"的要求。

法院結論是:發明屬於一種方案或是商業方法,不屬於可專利標的。
"We conclude that the claimed invention is to a scheme or a business method that is not properly the subject of letters patent."

my two cents:
其中很重要地引用下列前例針對可專利性判斷中"manner of manufacture"的解釋。

引用前例:("Patents—Application—Eradication of weeds from crop areas by application of chemicals—Known substances—New properties discovered—New process—Vendible product—Manner of manufacture—Statute of Monopolies 1628 (21 Jac. I, c. 3), s. 6—Patents Act 1952-1955 (Cth), s. 6.")(https://jade.io/article/65416

法院考量了專利局審查意見,也就是判定系爭案發明並非符合可專利性中"manner of manufacture"的定義,因此,法院就重新來探討何謂"manner of manufacture"?是否有更廣義的定義?

這裡有段我認為頗有意義的討論,法院討論的議題相對宏觀,認為專利法中"manufacture"並非是要簡化可專利性(patentability)問題,而是指涉所有可賦予專利壟斷權的類別,因此不能僅問"is this a manner of manufacture"就解決可專利性問題。法院認為(甚至搬出牛津英語字典的解釋),如果以此問題中"manufacture"解釋為限制任何人的發明應為以手或機器製作具體商品的話,就錯了!

"The word "manufacture" finds a place in the present Act, not as a word intended to reduce a question of patentability to a question of verbal interpretation, but simply as the general title found in the Statute of Monopolies for the whole category under which all grants of patents which may be made in accordance with the developed principles of patent law are to be subsumed. It is therefore a mistake, and a mistake likely to lead to an incorrect conclusion, to treat the question whether a given process or product is within the definition as if that question could be restated in the form: "Is this a manner (or kind) of manufacture?" It is a mistake which tends to limit one's thinking by reference to the idea of making tangible goods by hand or by machine, because "manufacture" as a word of everyday speech generally conveys that idea. The right question is: "Is this a proper subject of letters patent according to the principles which have been developed for the application of s. 6 of the Statute of Monopolies?" (at p269)"

補充資料:


--- 澳洲專利法可專利發明規定 ---
PATENTS ACT 1990 - SECT 18 Patentable inventions

Patentable inventions for the purposes of an innovation patent

 (1A)  Subject to subsections (2) and (3), an invention is a patentable invention for the purposes of an innovation patent if the invention, so far as claimed in any claim:

 (a)  is a manner of manufacture within the meaning of section 6 of the Statute of Monopolies; and

 (b)  when compared with the prior art base as it existed before the priority date of that claim:

 (i)  is novel; and

 (ii)  involves an innovative step; and

 (c)  is useful; and

 (d)  was not secretly used in the patent area before the priority date of that claim by, or on behalf of, or with the authority of, the patentee or nominated person or the patentee's or nominated person's predecessor in title to the invention.

可取得專利的條件:製作方法、新穎性、進步性、可利用性,在優先權日以前未秘密實施。

 (2)  Human beings, and the biological processes for their generation, are not patentable inventions.人類與其生物方法不得取得專利。

Certain inventions not patentable inventions for the purposes of an innovation patent

 (3)  For the purposes of an innovation patent, plants and animals, and the biological processes for the generation of plants and animals, are not patentable inventions.產生動植物生物方法不可專利。

 (4)  Subsection (3) does not apply if the invention is a microbiological process or a product of such a process.

微生物方法或其產品例外。

--------------------------


參考資料:

Ron

2024年6月18日 星期二

101答辯筆記 - 答辯說明請求項存在有意義的額外元件

Claim Rejections Under 35 U.S.C. §101






In the Office Action, claims 1-20 were rejected under 35 U.S.C. §101 because the claimed invention is directed to an abstract idea without significantly more. On page 3 of the office action, the Examiner indicated that, as per step 2A, the claim limitation under its broadcast reasonable interpretation covers interactions between people, then it falls within the “certain methods of organizing human activity” that is judicial exception, and furthermore the judicial exception is not integrated into a practical application because the additional elements “sensors” and the “data processing unit” amount no more than merely using a generic computer.

審查委員認為claim中即便有sensor, processor等元件,並沒有超越一般電腦能做的事,判定發明是抽象概念。

The applicant submits that the amended claim 1 is now directed to a care system and amended claim 11 has been amended to incorporate clarifications that reflect the scope of an automatic care method. According to the amended claim 1, the care system includes a data processing unit, a plurality of sensors and a user interface. Applicant submits that the amended claim 1 provides a solution that is able to train an intelligent model (i.e., the care-taking warning prediction model) by learning the data generated by the sensors and performing a machine-learning method  through meaningful additional elements/specific-purpose devices recited in the claim beyond the judicial exceptions.

答辯說發明為通過學習感測器產生的資料訓練智能模型,其中是通過有意義的額外元件執行機器學習方法。

Further, the sensors of the care system includes two different groups of sensors including a first group of sensors operating at all times and generating a first group of detection data related to a care recipient and a second group of sensors being activated to generate a second group of detection data related to the care recipient when the first group of detection data reaches a first threshold. The technical correlation between the first and second groups of sensors forms the claim limitation that not falls within the subject matter grouping the abstract idea. Furthermore, the claim recites the additional elements such as the sensor and the user interface for receiving a warning notification generated by the care recipient that are applied to the care system in a meaningful way beyond generally linking the use of the judicial exception to a particular technological environment.

在此說明系統中的感測器的特殊有意義之處,例如其中有兩組感測器,之間具有關聯性,且分別產生不同型態的資料,加上其中使用者介面,形成claim中的額外元件,以有意義的方式應用在系統中。

Still further, the data generated by the first and second groups of sensors and the timing generating the warning notification are particularly provided to be learned by a machine learning method performed by the data processing unit of the care system so as to establish a warning prediction model. It is appreciated that those data are transformed into the warning prediction model that is provided for the care system to determine whether or not a status of the care recipient satisfies a condition of generating the warning notification. The applicant asserts that additional elements recited in the claim limitation are sufficient to amount to significantly more than the judicial exception.

上述兩組感測器產生的數據以及產生警告通知的時間都提供機器學習,這些內容都形成claim中的額外元件,已經足以超越抽象概念。

Accordingly, Applicant respectfully submits that the amended claims now adequately reflect patent eligible subject matters that incorporate practical integration, whose patentability shall be determined substantially through novelty and non-obviousness analysis. Withdrawal of the 35 U.S.C. 101 rejections is respectfully requested.

因為審查委員這樣想,這裡也用相同的邏輯回應審查委員,因為以上特徵描述通過新穎性與進步性分析後,證明具備實質應用的專利適格性。

Ron

Petapator - Revolutionize Patent analysis with AI-powered Insights

新版Petapator - 整合大型語言模型(LLM),可以直接應用在專利檢索,增加效率,將LLM產生的檢索結果生成檔案,並且提供專業專利檢索人員以事先定義好的prompts快速而深入分析檢索結果。

特色包括:
- LLM integration - Tailor your patent search experience with LLM directly on patent search sites, effortlessly extracting insights and enhancing efficiency. 
- Streamline data organization by seamlessly downloading LLM responses into CSV files for future analysis.
- Leverage pre-defined prompts specifically created for patent professionals, enabling quick and deep analysis of search results.

安裝:
此範例直接安裝為Chrome的擴充功能。




使用步驟:
(1)用Google Patents檢索:
在Google Patents用"Retrieval-Augmented Generation/RAG"關鍵字檢索。


其中立即可以看到新的功能:"enter your prompt", "predefined prompts", "LLM analysis", "download patent info", "download LLM analysis"。

事先建立的prompt,真是挺口語的。

如何在檢索到的專利使用AI LLM?



(2)輸入LLM prompts進行分析:

選擇整頁檢索的結果,輸入prompt: "please provide the most useful patent for "retrieval augmented generation""。

(3)直接在Google Patents查看:

(4)下載分析結果:

按下"download patent information":


p.s.有一些功能需要付費,在此僅小試一下。

感謝來自petapator團隊的信息。

- Revolutionize Patent analysis with AI-powered Insights

Ron

2024年6月11日 星期二

使用名人的聲音造成損害賠償或是不公平競爭? - Midler v. Ford Motor Co. (9th Cir. 1988)

前言,聽到張大春news98節目來賓「葉郎-異聞筆記版主(Dr. Strangenote)」(編按,兩周一次的這個"影劇"系列十分精采)講到最近「Scarlett Johansson (史嘉蕾·喬韓森)」槓上「OpenAI」用了她的聲音的新聞事件,同時提到可能如1988年"Midler v. Ford Motor Co."事件,兩者有不小相似之處,這也意味著生成式AI將面對許多與"真人"維權的爭議,許多問題待解決,或許是人與機器之間永遠的戰爭。

在此僅"表面上"看一下這個Midler v. Ford Motor Co., 849 F.2d 460 (9th Cir. 1988)訴訟:
原告:Bette Midler
被告:Ford Motor Company、Young & Rubicam, Inc.

案件涉及未經同意將"名人/Bette Midler"的聲音用在商業用途上的爭議,被告Ford公司與其廣告商在電視汽車廣告上使用了"像Bette Midler的歌聲"(因為原唱不想接商業廣告),意圖是想將消費者帶回在大學"雅痞/Yuppies"時期,但卻使用了"像那個年代"的歌聲。

使用的歌曲是「Do You Want To Dance/"The Divine Miss M."」。因為Bette Midler拒絕接廣告,廣告商(列為本次被告)就找了長期擔任Bette Midler伴唱歌手(backup singer)"Ula Hedwig"來唱,並要求就唱得像原唱那樣唱「Do You Want To Dance("sound like Bette Midler's recording")

廣告播出後,Bette Midler被告知這個廣告,並有許多人覺得這就像是Bette Midler唱的,甚至在訴訟中有專業人士作證覺得這就如Bette Midler唱的。

雖廣告商有獲得使用歌曲的授權,但卻"如同"使用了Bette Midler的聲音,在地方法院審理此案時,認為被告就如"普通小偷/of the average thief"那樣偷了Bette Midler的聲音

但是,儘管如此認為,在法律上的判斷卻沒有"禁止模仿名人/Bette Midler聲音的規定",因此在簡易判決中判定Bette Midler敗訴。

Bette Midler上訴第九聯邦巡迴法院。

美國憲法第一修正案保護以相似物/重製聲音(reproduction)媒體的工作,這是一種言論與新聞自由的權利(freedom of speech and press),如果重製的目的是傳達資訊與文化,就可免責如果不是用在以上免責的功能,就不能免責

另一方面,在著作權法中,即便當事人要盡可能精準模仿原唱/原表演者的表演,模仿這件事並不構成侵權

前例一(涉及損害賠償):
列舉另一個訴訟(涉及損害賠償),相似地,有其他歌手(Nancy Sinatra)狀告廠商在廣告中模仿她的聲音,包括聲音與穿著,當時訴訟理由是"不公平競爭/unfair competition",主張其個人形成的二次含意(secondary meaning)的商標保護,並要求損害賠償。但法院的判決是,因為被告已經取得歌曲的使用授權,原告並不能再要求損害賠償,因此原告敗訴

在本案Midler v. Ford中,法院認為如果原告同樣地主張其secondary meaning商標權,應該也是敗訴,但與上述案例不同的是,本案原告Bette Midler並非要求損害賠償,並且人聲(有不準確因素)也不是著作權保護標的

前例二(涉及不公平競爭):
另一件比較像本案例的前例"Bert Lahr v. Adell Chemical Co. ",此案原告認為被告在廣告中使用像其獨特的聲音(pitch, accent, inflection, and sounds),原告以"不公平競爭/unfair competition"提出告訴,因為被告使用他的聲音讓他相關市場上的發展受到影響。法院認為此案例與本案比較相似,只是此案例同樣地判決原告敗訴,理由是,原告沒有拍電視廣告,廣告上使用他的聲音並不會有"不公平競爭"的問題

很精彩地,回到本案例,加州民法(California Civil Code)中規定如果他人使用某人的名字、聲音、簽名、照片或是其他相似物(name, voice, signature, photograph or likeness),某人可主張損害賠償。不過,本案例中,被告並未使用Bette Midler的名字、簽名與照片,聲音也不是用她的,所述likeness是指影像,非人聲。

也就是說,Bette Midler無法主張損害賠償,因為她不拍廣告,也不會構成不公平競爭。

前例三(涉及財產權):
名字、聲音、簽名、照片或相似物,都算是人民的財產權。案例"Motschenbacher v. R.J. Reynolds Tobacco Co. (9th Cir. 1974)"中,被告使用著名賽車手的照片在香菸廣告中,原告Lothar Motschenbacher在照片中的車輛裡,但沒有出現他的影像,有些人看了這個廣告會認為車是他的,且他在車中,因此會認同這個產品,於是此案例判決被告侵害原告的專屬特徵,即財產權。

在此案例中,被告將原告的車輛與本人(雖看不到,但暗示他在車中)都用在廣告中,目的是希望消費者因為相信原告而認同產品;在本案中,被告並沒有將原告置入廣告,但被告廣告商使用像Bette Midler的歌聲也是希望向消費者傳達如Midler唱歌的那樣,在這點是相像的

本案法官用了一個挺哲學的概念來看本案事實,認為"人聲"如人臉一樣的可識別性,如我們通過聲音來辨識朋友,因此認為仿冒一個人的聲音就是用了他的身分

"A voice is as distinctive and personal as a face. The human voice is one of the most palpable ways identity is manifested. We are all aware that a friend is at once known by a few words on the phone. At a philosophical level it has been observed that with the sound of a voice, "the other stands before me.""

法院判決:
法院不覺得所有模仿聲音的情況都可以當作提告的理由,但判定是,一個職業歌手的聲音是消費者可辨識的聲音,如果被模仿來執行商業行為,販售者就是使用了不是他們的東西,在加州是違法的,因此判決本案被告使用了Bette Midler的聲音,也就是她的身分,去販售產品,因此推翻地方法院作出的簡易判決

"We need not and do not go so far as to hold that every imitation of a voice to advertise merchandise is actionable. We hold only that when a distinctive voice of a professional singer is widely known and is deliberately imitated in order to sell a product, the sellers have appropriated what is not theirs and have committed a tort in California. Midler has made a showing, sufficient to defeat summary judgment, that the defendants here for their own profit in selling their product did appropriate part of her identity."

my two cents:
本案顯示,「名氣」確實會帶給人利益,因為本案判決僅能適用在「名人」身上,可能還要"夠有名"才會發生這類事情。

回到最近史嘉蕾·喬韓森v.OpenAI案上,OpenAI退縮、道歉,就是因為史嘉蕾·喬韓森夠有名氣,聲音有點像(事實上史嘉蕾·喬韓森聲線確實很特別)就要下架!!!

逼得OpenAI作出聲明:



(OpenAI不用賠償,卻也賺到名聲)




Ron