創作者 vs. AI生成工具:和解之路

議程

14:00-14:05  活動介紹
14:05-15:45  焦點座談

  • 主持人-蔡志宏 庭長(台灣士林地方法院)
  • 與談人-
    • 林思翰導演(Group.G共同創辦人)
    • 陳家駿理事長(台灣資訊智慧財產權協會)
    • 馮震宇教授(政治大學科技管理與智慧財產研究所)
    • 張智翔先生  (badbone 設計師、滷吉拉加熱滷味 負責人)

15:45-16:00  現場問答

臺灣的知名作家吳淡如在2023年初於社群媒體貼出一張「電腦繪圖」作品,卻被網友揭穿該作品是AI生成繪圖,意外引發藝術創作者、網友對於著作權歸屬的討論。到底 AI 所生成的作品算不算是原創作品,是否也可主張著作權?隨著生成式 AI 工具的熱度激增,近期成為討論焦點。台灣網路講堂2023年第一場的座談於4月20日舉辦,由台灣士林地方法院的蔡志宏庭長擔任主持人,並邀請到Group.G共同創辦人林思翰導演、台灣資訊智慧財產權協會陳家駿理事長、政治大學科技管理與智慧財產研究所的馮震宇教授以及badbone 設計師張智翔先生來擔任與談人,共同針對AI生成議題的著作權、法律規範以及相關衝擊等議題進行探討。

主持人蔡庭長於開場時談到,從人類發展的歷史來看,人類勞動力的產出量,隨著科技進步,產量越來越大,機械動力雖可以取代沉重的工作增加量能,卻無法取代「創作」勞動的部分,因為創作需要思考,包括像是繪畫、寫作等,無法藉助機械的力量來生成,直到ChatGPT及相關繪圖AI等技術產生後,這樣的門檻似乎已被跨越,且AI可以無限量的產出/創作,對於人類造成衝擊及影響。過往著作權基本架構保護的是思想表達,而思想的表達則是需要靠人進行創作勞動,現在有AI技術可以取代,那這還是傳統著作權所要保護的範圍嗎?且制度面要如何提供誘因,都是值得討論的課題。

首先,在關於AI生成作品合法擁有者的討論中,馮震宇教授表示目前市面上出現的AI工具很多,使用者也不一定只使用單一種AI技術,若要談AI生成作品的擁有者是誰,可能很難將其定義,或許可先從權利與義務來區分,不只要關注AI生成作品的權利歸屬,還要從是否能負擔法律責任的觀點來認定。而權利與義務的賦予,可透過服務中的使用者協議來定義。另外,因為AI的出現,我們現在要思考的是,需不需要改變目前對於著作權、專利等的保護方式?要依據現有的法律架構?還是給新的框架作為規範?

陳家駿理事長談到,目前國際間對於著作權的主流共識是所產出的作品只要不是經由抄襲、模仿、剽竊等方式而來,就視該作品中創作成分的多寡,給予適度的保護,但前提是這些作品都是「人為」的,否則不在保護範圍內。陳理事長引用中國的案例為說明,中國深圳法院在2021年針對騰迅透過AI所產出的新聞稿,後來法院判騰迅享有AI的著作權;歐盟專利局也曾以專利的作者必須是「人類」而非「機器」為理由,駁回2項以AI為發明人的器材專利申請;在英國的著作權法中有明文,視人為介入的程度多寡,也就是說由承作創設此作品而進行必要安排的人(如,開發此演算法或軟體有貢獻者),來決定誰享有著作權。不過,經由上述說法利用「人力」介入的多寡來決定著作權,又可能衍生出另外兩個問題,那就是人力與AI到底誰的貢獻更多?以及若真的發生侵權糾紛,又是誰要出面負責任呢?陳理事長也提醒,即便目前可透過使用者條款來規範權利與義務,不過當出現侵權案件時,法院不見得會如此採認。

林思翰導演認為「著作權」與「使用」可分開來看,他認為能合理使用AI生成工具所產出的作品於商業行為上,且能夠承擔責任是很重要的事情。目前在商業使用上,林導演建議要誠實與客戶揭露該作品有多少比例是使用AI生成工具來共同製作,並且以簽約方式來規範責任歸屬問題。張智翔設計師也呼應此觀點,他進一步將設計師廣泛蒐集各種元素創作出屬於自己的作品,和AI目前抓取既有著作再生成出作品的運作原理做了類比,業主或客戶不見得能夠感受這兩類作品之間的差異,也因此針對AI工具使用,能夠誠實揭露更為重要。

在談到AI生成作品對創作者的衝擊時,林思翰導演認為AI生成工具多少會壓縮到創作者生存空間,但也無須太過擔憂,主要是因為,以他認知,他覺得目前著作權保護的是創意的表達,而非創意本身;張智翔設計師則認為,有了AI工具的協助,可以簡化許多原本繁複的步驟,對設計師最大的改變是可縮短設計時間,增加產能,並且在不需要熟悉各種不同繪圖技巧的情況下,也可以藉此將心中的意象更具體的呈現出來。

在關於作品風格是否也受保護的討論,馮震宇教授認為這部分還是有討論的空間。馮教授以日本的虛擬歌手–初音未來為例。初音未來的聲音是在取得日本聲優的授權後,再利用YAMAHA的VOCALOID引擎和Crypton自主研發的NT歌聲合成引擎的聲音庫軟體,進而再創作出屬於初音未來的聲音,這是使用既有的素材再去衍生創作成屬於自己風格的作品。不過,主要還是要使用者本身懂得自律、能夠拿捏分寸,不會把創作出來的作品拿去做違法的事情,在美國就已經出現盜用他人的聲音來做詐騙的事件。至於作品風格是否要也受到保護?馮教授指出當使用他人風格之目的是為造成第三人誤解,風格的保護就有必要性。

林導演則從創作者角度來探討,他指出當某人學習很多人的風格,然後融會貫通再創作成屬於他自己風格的作品,當民眾看到這個創作品時,首先會想到的是創作者本人,還是其他創作者?當作品本身不會誤導他人,那就不太會有問題。但不論是導演還是畫家,每個創作者都可能會隨著時間進步或想法不同,所謂的創作風格也可能因此而改變,所以陳導演認為,被保護的是創作者所想傳達的意志而非呈現出來的畫面/作品。現在市面上AI工具很多,重點不是在使用哪些工具比較適合,而是要怎樣利用這些工具創作出能傳達自己意志的作品。張智翔設計師認為,科技的進步主要是源起於人類的懶惰,挑選適合自己的AI工具,但不能失去自己的思考能力才是重要的事情。

主持人蔡庭長在道德與法律層面上先簡單做歸納,他提到創作者在使用AI創作時,要記得誠實揭露的原則,不要把AI生成當作是個人辛苦的創作。另外,也建議創作者可以把整個創作軌跡與歷程鉅細靡遺的保留下來,像是如何與AI互動的資料等,未來若真的需要進入法院對簿公堂時,可以將其證據作為辯論的基礎。

座談會最末,各位專家分別提出其對於未來世界的想像與展望,馮震宇教授認為AI能幫助人類持續成長,只要能夠善加利用,應可將世界導向更好的一面;陳家駿理事長表示,新科技的浪潮來襲無人能擋,張開雙手擁抱它就對了,法律面也不要設下太多限制;林思翰導演認為,新科技雖然可能對某些產業造成衝擊,但同時也會促進更多元的發展,暫時無需過度緊張;張智翔設計師表示,即便AI設計是目前的熱潮,也不代表每個人學會這項技術後就都會轉行成為設計師,他認為AI技術的不當使用,才是未來需要關注的重點。

AI Generated Images and Copyright

Agenda

14:00-14:05  Introduce
14:05-15:45  Panel Discussion

  • Moderator:
    Tsai,Chih-Hung─Division-Chief Judge, Taiwan Shilin District Court
  • Penelists:
    • Lin, Sih-Han─Director of Group.G
    • Chen, Jia-Jyun─Chairman, Taiwan Information Intellectual Property Association
    • Fong, Jerry G. ─Professor, Graduate Institute of Technology, Innovation & Intellectual
      Property Management, NCCU
    • Chang, Harley ─Designer, BAD BONE

15:45-16:00  Q&A

At the opening, Judge Tsai Chih-Hung mentioned that in the history of human development, the output of human labor has become increasingly larger with technological progress. Although mechanical power can replace heavy work and increase output, it cannot replace human creativity as creativity requires human thinking, which cannot be generated by the power of machinery. Until the emergence of ChatGPT and other generative AI imaging programs, the threshold seems to have been crossed. AI production is unlimited while human is NOT. For example, the basic idea of copyright protection used to be for the expression of ideas, with AI tool such as Midjourney to output the expression of ideas, is it still within the scope of traditional copyright protection? This is worthy of discussion.

The panel started with the discussion about the legal owner of AI-generated works, Professor Feng Jerry G said that there are many AI tools available on the market, and users may not only use a single type of AI technology. It may be difficult to define the owner of AI-generated works. Perhaps it can be distinguished from rights and obligations. We should not only focus on the attribution of rights of AI-generated works but also recognize from the perspective of whether they can bear legal responsibility. The assignment of rights and obligations can be defined through user agreements in services. In addition, because of the emergence of AI, we need to consider whether to change the current protection methods for copyright, patents, etc. Should we follow the existing legal framework, or give new frameworks as norms?

Chairman Chen Jia-Jyun pointed out that the current mainstream consensus on copyright internationally is that works are protected as long as they are not obtained through plagiarism, imitation, theft, and other means. The extent of protection given depends on the degree of creative content in the work, but the prerequisite is that these works are “human-made.” Otherwise, they are not within the scope of protection. Chen cited examples from China to illustrate this point. In 2021, the Shenzhen court ruled that Tencent owned the copyright of a news release generated by AI. The European Patent Office has also rejected two patent applications for equipment inventions claiming AI as the inventor because the author of the patent must be “human” rather than a “machine.” In the UK Copyright Act, the degree of human involvement in creating the work (e.g., contributors who develop the algorithm or software necessary for the work) is used to determine who owns the copyright. However, using the degree of “human intervention” to determine copyright ownership can lead to two additional problems: Who contributed more, humans or AI? And if there is a copyright dispute, who is responsible? Chairman Chen also cautioned that even though user agreements can regulate rights and obligations, courts may not necessarily recognize them in the event of copyright infringement cases.

Director Lin Sih-han believes that “copyright” and “use” of the work can be viewed separately. He thinks that being able to use works generated by AI tools reasonably in commercial activities and being able to take responsibility is very important. Currently, in commercial use, Director Lin suggests being honest with clients about the percentage of the work that was generated by AI tools, and using a signed agreement to regulate responsibility attribution issues. Designer Chang Harley also echoed this view. He further compared the designer’s extensive collection of various elements to create their own works with the operating principle of AI currently grabbing existing works and generating new works. Owners or clients may not necessarily be able to perceive the differences between these two types of works. Therefore, it is more important to be honest about the use of AI tools.

When discussing the impact of AI-generated works on creators, director Lin believes that AI tools may compress the living space of creators to some extent, but there is no need to worry too much. He thinks that currently, copyright protection is for the expression of creativity, not creativity itself. Designer Chang Harley believes that with the assistance of AI tools, many complex steps can be simplified, and the biggest change for designers is the ability to shorten design time, increase productivity, and present concrete images without the need to be familiar with various drawing techniques.

In the discussion of whether the style of a work is also protected, Professor Feng believes that there is still room for discussion on this issue. Professor Feng used the virtual singer Hatsune Miku from Japan as an example. Hatsune Miku’s voice was created by using Yamaha’s VOCALOID engine and Crypton’s self-developed NT song synthesis engine after obtaining the authorization of a Japanese voice actor, and then creating a voice library software that belongs to Hatsune Miku. This is using existing materials to derive creations that belong to one’s own style. However, it is still important for users to have self-discipline, be able to grasp the scale, and not use their creations for illegal purposes. In the United States, there have been cases of using someone else’s voice for fraud. As for whether the style of a work should also be protected, Professor Feng pointed out that the protection of style is necessary when the purpose of using someone else’s style is to cause misunderstanding among third parties.

Director Lin discusses from the perspective of a creator, pointing out that when someone learns the styles of many artists and then incorporates them into their own work, when people see this creation, will they think of the creator himself or other creators? If the work itself does not mislead others, then there is less of a problem. However, whether it is a director or a painter, every creator may progress over time or have different ideas, and the so-called creative style may change as a result. Therefore, Director Lin believes that what is protected is the will that the creator wants to convey, rather than the image/artwork presented. There are many AI tools available on the market now, and the key is not which tool is more suitable to use, but how to use these tools to create works that can convey one’s own will. Designer Chang believes that the progress of technology mainly stems from human laziness. Choosing suitable AI tools is important, but not losing one’s own thinking ability is the most important thing.

Judge Tsai summarized briefly on the moral and legal aspects, mentioning that creators using AI for their creations should remember the principle of disclosure and not present AI-generated works as personal creations. Additionally, he suggested that creators should keep a detailed record of the entire creative process and interaction with AI, including data on how they interacted with AI. This evidence could be used as a basis for argument in court, if necessary, in the future.

At the end of the symposium, each expert presented their own imagination and outlook for the future world. Professor Feng believed that AI could help humans continue to grow, and as long as it is well utilized, it should be able to lead the world towards a better direction. Chairman Chen stated that the wave of new technology is unstoppable, and we should embrace it with open arms, without setting too many legal restrictions. Director Lin believed that although new technology may impact certain industries, it will also promote more diverse development, so there is no need to be overly anxious. Designer Chang pointed out that even though AI design is currently a trend, it does not mean that everyone who learns this technology will switch careers to become a designer. He believes that improper use of AI technology is the key issue that needs to be focused on in the future.