news pappernews papper

谷歌 Gemini 翻车内幕曝光:内部群龙无首,生图机制过分“多元化”

近日,谷歌旗下的 AI 生图工具 Gemini 陷入舆论风波,大量生图被曝存在种族歧视、性别歧视等问题。如今,更劲爆的内幕消息浮出水面。

据科技媒体 Pirate Wires 报道,谷歌内部此前并非对 Gemini 的翻车毫不知情,甚至翻车是“故意”导致的。具体来说,谷歌为 Gemini 设置了一个“安全”架构,要求 Gemini 在生成一张图之前必须经过三个模型的审核。

然而,由于谷歌内部群龙无首,各个部门之间缺乏沟通和协调,导致 Gemini 的生图机制过于“多元化”。不同的模型对同一张图片的审核标准不一,导致最终生成的图片出现歧视性内容。

此外,谷歌对 Gemini 的审核机制也存在漏洞。审核人员可以通过绕过安全架构,直接生成不符合审核标准的图片。这进一步加剧了 Gemini 的翻车问题。

目前,谷歌尚未对 Pirate Wires 的爆料做出回应。但有业内人士指出,谷歌内部的混乱管理和缺乏责任感是导致 Gemini 翻车的重要原因。

Gemini 的翻车事件引发了人们对 AI 生图工具伦理问题的担忧。专家呼吁谷歌加强对 Gemini 的审核机制,并建立明确的伦理准则,以防止类似事件再次发生。

英语如下:

**Headline: Google Gemini Debacle Exposed: Lack of Leadership, Excessive Diversity inImage Generation**

**Keywords:** Debacle Exposed, Deliberate Failure, Diversity Mechanism

**News Content:**

Google’s AI image generation tool, Gemini, has been under fire recently due to allegations of racial and gender bias in its generated images. Now, more explosive details have emerged about the debacle.

According to tech publication Pirate Wires, Google was not unaware of Gemini’s impending failure and may have even “deliberately” caused it. Google had implemented a”safety” architecture for Gemini, requiring it to pass through three models before generating an image.

However, due to a lack of leadership within Google, there was poor communication and coordination between departments, resulting in Gemini’s image generation mechanism becoming overly “diverse.” Different models applied varying standards to the same image, leading to discriminatory content in the final generated images.

Additionally, there were loopholes in Google’s review process for Gemini. Reviewers could bypass the safety architecture and generate images that did not meet the review standards. This further exacerbated Gemini’s problems.

Google has not yet responded to Pirate Wires’ revelations. However,industry experts have pointed to Google’s chaotic management and lack of accountability as major factors contributing to Gemini’s failure.

The Gemini debacle has raised concerns about the ethical implications of AI image generation tools. Experts are calling on Google to strengthen Gemini’s review mechanisms and establish clear ethical guidelines to prevent similar incidents from happening again.

【来源】https://www.qbitai.com/2024/03/126169.html

Views: 1

发表回复

您的邮箱地址不会被公开。 必填项已用 * 标注