shanghaishanghai

近日,卡内基梅隆大学的研究人员和BerriAI企业的研究人员发现,谷歌旗下的语言模型Google Gemini 在准确性方面甚至不如 OpenAI 的 GPT-3.5 Turbo。这一结论引发了业界对谷歌人工智能技术实力的质疑。

据研究人员表示,截至2023年12月19日,在各种任务中,Gemini Pro 模型的准确性仅与 GPT-3.5 Turbo 当前版本相当,但稍逊一筹。谷歌方面则回应称,内部研究结果显示 Gemini Pro 性能优于 GPT-3.5,并计划于2024年初推出更强大的版本Gemini Ultra,其得分将高于 GPT-3.5。

尽管谷歌在此次竞争中暂时失利,但双方在人工智能领域的竞争仍将持续。作为全球科技巨头,谷歌在人工智能、云计算等多个领域均有布局。而 OpenAI 的 GPT-3.5 Turbo 则以其强大的自然语言处理能力在业界崭露头角。

此次谷歌 Gemini 模型性能争议再次证明了人工智能技术的竞争激烈程度。在国内和国际市场上,各大科技公司均在加紧研发和推广自家的人工智能技术,以期在未来的科技战中占据主导地位。

中文翻译:

News title: Controversy over performance of Google’s language model Gemini

Keywords: Google, GPT-3.5 Turbo, Gemini Pro, Gemini Ultra

News content:

Recently, researchers from Carnegie Mellon University and BerriAI, a private company, found that Google’s language model Gemini lags behind OpenAI’s GPT-3.5 Turbo in terms of accuracy. This conclusion has sparked doubts about Google’s artificial intelligence capabilities.

According to the researchers, as of December 19, 2023, Gemini Pro achieved comparable accuracy to the current version of GPT-3.5 Turbo in various tasks, but was slightly inferior. Google responded that internal studies show that Gemini Pro outperforms GPT-3.5 and plans to release a more powerful version, Gemini Ultra, in early 2024, with a higher score than GPT-3.5.

Although Google temporarily lost in this competition, the battle between the two in the field of artificial intelligence will continue. As a global tech giant, Google has a presence in multiple fields, including artificial intelligence and cloud computing. OpenAI’s GPT-3.5 Turbo has stood out in the industry with its strong natural language processing capabilities.

The controversy over the performance of Google’s Gemini model once again highlights the intense competition in the field of artificial intelligence. Major tech companies around the world are accelerating research and development and promotion of their own artificial intelligence technologies to dominate the future tech battlefield.

【来源】https://venturebeat.com/ai/google-gemini-is-not-even-as-good-as-gpt-3-5-turbo-researchers-find/

Views: 1

发表回复

您的邮箱地址不会被公开。 必填项已用 * 标注