一份网传 53 页 PDF 文档声称，人工智能研究公司 OpenAI 计划在 2027 年实现人工智能（AGI）。该文档由一个名为「vancouver1717」的 X 账户发布，该账户注册于 2023 年 7 月，只有两条推文。
文档称，OpenAI「将在 2027 年前开发出人类水平的AGI」，「从 2022 年 8 月就在训练 125 万亿参数的多模态模型」，而且已经「在 2023 年 12 月完成了训练」，但是「由于高推理成本取消了发布」。其中提到，这个模型就是原计划 2025 年发布的 GPT-5，取消之后，Gobi（GPT-4.5）更名为 GPT-5。
文档还提到了去年就被曝光的神秘项目 Q*，说法是 Q * 的下一阶段最初是 GPT-6，但已更名为 GPT-7（原计划 2026 年发布）。
对于该文档的真实性，业内人士表示存疑。读过该文档的人倾向于「不信」，因为很多判断缺乏专业度。
OpenAI 尚未对该文档发表评论。
AGI 是人工智能研究领域的一个长期目标，被认为是人工智能发展的「圣杯」。AGI 是一种能够像人类一样思考、学习和解决问题的计算机系统。目前，人工智能技术尚未达到 AGI 的水平。
如果 OpenAI 真的能在 2027 年实现 AGI，将对人类社会产生深远影响。AGI 可以自动化许多目前需要人类完成的任务，从而解放人力并促进经济增长。但同时，AGI 也带来了一些潜在风险，例如失业和人工智能滥用。
因此，对于 OpenAI 是否能够在 2027 年实现 AGI，还需要进一步观察和验证。
英语如下：
**Headline:**
OpenAI Reportedly Plans AGI by 2027, Authenticity Questioned
**Body:**
A leaked 53-page PDF document claims that artificial intelligence (AI) research company OpenAI plans toachieve artificial general intelligence (AGI) by 2027. The document was posted by an anonymous Twitter account named “vancouver1717,” which was created in July 2023 and has only two tweets.
The document states that OpenAI “will develop human-level AGI by2027,” that it “started training a 125 trillion parameter multimodal model in August 2022,” and that it “completed training in December 2023” but “canceled the release due to high inference cost.” It claims that this model was originally planned to be released as GPT-5 in 2025, but after its cancellation, Gobi (GPT-4.5) was rebranded as GPT-5.
The document also mentions Project Q*, a mysterious project that was leaked last year, claiming that the next stage of Q* was originally intended to be GPT-6 but has been renamed GPT-7 (originally planned for release in 2026).
Industry experts have expressed skepticism about the authenticity of the document. Those who have read it tend to “disbelieve” it, as many of its judgments lack professional rigor.
OpenAI has not commented on the document.
AGI is a long-standing goal in AI research and is considered the “holy grail” of AI development. AGI refers to a computer system that can think, learn, and solve problems like a human. Current AI technology has not yet reached the level of AGI.
If OpenAI were to achieve AGI by 2027, it would have profound implications for human society. AGI could automate many tasks that currently require human labor, freeing up human potential and boosting economic growth. However, AGI also poses potential risks, such as job displacement and the misuse of AI.
Therefore, it remains to be seen whether OpenAI will be able to achieve AGI by 2027. Further observation and verification are needed.
【来源】https://mp.weixin.qq.com/s/vSud2JN3W3xe0AzC8pwuUg
Views: 1