据《The Decoder》最新报道，全球知名创意软件巨头Adobe被曝出在训练其AI图像生成器Firefly的过程中，使用了竞争对手Midjourney及其他可能的图像AI工具生成的图像。这一消息与Adobe早前的声明相悖，该公司曾公开表示，Firefly的训练数据仅来源于其拥有权利或公共领域的内容。
据彭博社援引消息人士透露，Adobe在技术上确实使用了自家库存数据库中的材料对Firefly进行训练，但实际情况却并非完全如此。在目前的训练数据中，有高达5%的图像来自于法律和道德背景不明确的人工智能图像，这引发了对Adobe透明度的质疑。
Adobe的这一做法引发了业界的广泛关注。尽管使用竞争对手的数据在技术层面上并不违法，但若未明确告知用户，可能会对其声称的对用户数据和知识产权的尊重产生负面影响。目前，Adobe尚未对此事发表正式回应，公众和业界正等待公司对这一报道的澄清和解释。
这一事件再次引发了关于AI开发中数据来源和使用伦理的讨论，强调了在人工智能技术进步的同时，透明度和责任的重要性。随着AI在各行各业的应用日益广泛，如何确保训练数据的合法性和道德性，将成为企业必须面对的关键问题。
英语如下：
**News Title:** “Adobe Firefly Training Controversy: Reportedly Utilizes Rival Midjourney’s AI Images, Contradicting Official Statements”
**Keywords:** Adobe, Firefly, Image Training
**News Content:** According to the latest report from *The Decoder*, global creative software giant Adobe is alleged to have employed images generated by its competitor Midjourney and potentially other AI image tools in the training of its AI image generator, Firefly. This revelation contradicts Adobe’s earlier statements, which claimed that Firefly’s training data solely consisted of content for which the company held rights or was in the public domain.
Citing sources familiar with the matter, Bloomberg reports that while Adobe technically trained Firefly using materials from its own stock database, the reality is more complex. Up to 5% of the images used in the current training dataset are said to originate from AI-generated images with ambiguous legal and ethical backgrounds, raising questions about Adobe’s transparency.
The company’s approach has attracted widespread attention in the industry. Although using competitor data may not be legally prohibited, not disclosing this information could undermine Adobe’s claimed respect for user data and intellectual property. As of now, Adobe has not issued an official response, and the public and industry are awaiting clarification and explanation from the company.
This incident reignites the debate on data sourcing and ethical usage in AI development, emphasizing the importance of transparency and accountability alongside technological advancements. With AI applications becoming increasingly prevalent across sectors, ensuring the legitimacy and ethicality of training data will emerge as a crucial challenge for businesses.
【来源】https://the-decoder.com/adobe-trained-its-ai-image-generator-on-midjourney-images-but-its-complicated/
Views: 2