谷歌近日宣布，其 Gemini 大模型的文生图功能因涉嫌刻意追求“多元化”而引发争议，目前该功能已被暂时搁置。据 BBC 和 The Verge 报道，部分用户反映 Gemini 在生成人物图像时，过度强调种族和性别的多样性，以至于生成的图片与实际历史或情境不符，存在误导的可能性。这一问题暴露了人工智能在追求多元化时可能面临的挑战，即如何在保持真实性和准确性的前提下实现公正的多元化。
谷歌对此迅速作出回应，表示正在“加紧”修复这一bug，以确保 Gemini 的图像生成功能能够更加准确、客观地反映输入信息。公司决定暂停 Gemini 的人物图像生成服务，以期在不久的将来发布一个经过改进的新版本。谷歌的这一举措旨在解决当前的问题，同时维护用户对人工智能技术的信任和依赖。
这一事件再次引发了关于人工智能道德和算法偏见的讨论，强调了科技公司在开发和应用AI技术时，必须谨慎处理文化、种族和性别等敏感问题，以避免无意中强化或制造偏见。谷歌的及时回应和修复行动表明了其对这些问题的重视，并致力于提供更加公平、准确的AI服务。
英语如下：
News Title: “Google Pauses Gemini’s Text-to-Image Function: Pursuing Diversity at the Cost of Accuracy, Fixes Underway”
Keywords: Google fix, Gemini, diverse imagery
News Content: Google recently announced that it has temporarily suspended the text-to-image feature of its Gemini large language model due to controversy surrounding its刻意 pursuit of “diversity.” As reported by BBC and The Verge, some users have alleged that Gemini exaggerates racial and gender diversity when generating人物 images, to the point of creating pictures that are historically or contextually inaccurate, potentially leading to misinformation. This issue highlights the challenges artificial intelligence may face in striving for diversity while maintaining authenticity and accuracy.
Google has promptly responded, stating that it is “accelerating” efforts to fix the bug, ensuring that Gemini’s image generation capabilities more accurately and objectively reflect input information. The company has decided to suspend Gemini’s character image generation service, with plans to release an improved new version in the near future. This move aims to address the current issue while preserving user trust and reliance on AI technology.
The incident reignites discussions about the ethics of AI and algorithmic biases, emphasizing the need for tech companies to handle sensitive issues like culture, race, and gender cautiously when developing and applying AI technologies, to prevent inadvertently reinforcing or creating biases. Google’s timely response and remedial actions demonstrate its commitment to addressing these concerns and providing more equitable and accurate AI services.
【来源】https://www.ithome.com/0/751/457.htm
Views: 1