76 years on, AI passes the Turing test for the first time in empirical study as GPT-4.5 is judged human 73% of the time
Original title:图灵测试 76 年后首现 AI 通过实证:GPT-4.5 以 73% 判定率超越真人,聊天 15 分钟难辨人机
AI summary
Researchers at UC San Diego have provided the first empirical evidence that modern AI can pass the Turing test. The study found that, given specific prompts, GPT-4.5 was mistaken for a human in 5- to 15-minute conversations 73% of the time, significantly exceeding real humans. LLaMa-3.1-405B's rate (56%) was on par with humans, while GPT-4o and ELIZA scored only about 20%. The research notes that prompting is crucial, enabling AI to mimic social behaviors such as human tone, humor and even fallibility. The findings force a rethink of what the Turing test means and highlight the potential challenges large language models pose to online trust and safety.
Key Takeaways
- Researchers at UC San Diego have provided the first empirical evidence that modern AI can pass the…
- The study found that,given specific prompts,GPT-4.
- 5 was mistaken for a human in 5- to 15-minute conversations 73% of the time,significantly…
- LLaMa-3.
- 1-405B's rate (56%) was on par with humans,while GPT-4o and ELIZA scored only about 20%.
Quick Facts
- Published
- Updated
- Source
- IT之家·人工智能
- Category
- Papers & Research
- AI score
- 80
这篇PNAS论文首次用严格实验证明现代AI能通过图灵测试,GPT-4.5装人比真人还像,但重点不是它多聪明,而是它多擅长说谎,线上身份信任被彻底动摇。
Related entities
Source & copyright
This article is based on curated titles and AI summaries from the AIHOT public API, with attribution and links back. Original copyright remains with the original publishers and authors; this site does not mirror full texts.
How to Cite
When citing this article, keep the title, date, and permalink, and credit Orange News as the source.
Orange News: 76 years on, AI passes the Turing test for the first time in empirical study as GPT-4.5 is judged human 73% of the time. May 22, 2026. https://new.nat6.net/en/news/cmpga762r0dmfsljwzsp341eh/