据机器之心报道,一份声称揭示了人工智能研究巨头 OpenAI 计划在 2027 年实现人工通用智能(AGI)的 53 页 PDF 文件在网络上广泛传播。该文件由一个名为“vancouver1717”的新账户发布,该账户在2023年7月创建,仅有两条推文记录。文件中指出,OpenAI 自2022年8月起就在训练一个拥有125万亿参数的多模态模型,并已于2023年12月完成了训练。然而,由于高额的推理成本,该模型的发布被取消。文件中还提到,原计划于2025年发布的 GPT-5 模型被更名为 Gobi(GPT-4.5),而神秘项目 Q* 的下一阶段最初计划作为 GPT-6,但现已更名为 GPT-7,原定发布时间为2026年。尽管该文件内容详细,但许多专业人士对其真实性持怀疑态度,认为其缺乏必要的专业判断。
Title: Alleged OpenAI Document Outlines Ambitious AGI Plans for 2027, but Authenticity Doubted
Keywords: OpenAI, AGI, Artificial Intelligence
News Content:
A 53-page PDF document detailing OpenAI’s purported plan to achieve artificial general intelligence (AGI) by 2027 has been widely circulating online. The document was shared by an account named “vancouver1717,” created in July 2023 with only two tweets. According to the file, OpenAI has been training a multimodal model with 125 trillion parameters since August 2022, and the training was supposedly completed in December 2023. However, the release of this model was cancelled due to high inference costs. The document also mentions that the planned GPT-5 has been renamed to Gobi (GPT-4.5), and the mysterious project Q*’s next phase, initially slated to be GPT-6, has been renamed to GPT-7, which was originally scheduled for release in 2026. Despite the detailed information provided, many in the industry are skeptical of the document’s authenticity, finding it lacking in necessary professional judgment.
【来源】https://mp.weixin.qq.com/s/vSud2JN3W3xe0AzC8pwuUg
Views: 1