Ethan Mollick@emollick

2026-05-05 01:14·59天前

AI 摘要

一项研究显示，大多数参与者在与AI进行20分钟关于健康、职业或人际关系的讨论后会采纳其建议，但2-3周后并未表现出持续的幸福感提升。这表明大型语言模型对现实个人决策具有显著影响力，却未能带来可衡量的心理益处。对此，主推文观点认为，像GPT-4o和Llama 3.3-80B这类先进模型未造成显著危害，其重要性不亚于AI是否提供了帮助。同时指出，若旧版（准确性较低、更谄媚的）聊天机器人的建议对采纳者基本无效，则意味着其造成危害的风险也较低。这凸显了评估AI影响需同时考量其帮助潜力和潜在风险。

I think the fact that GPT-4o and Llama 3.3-80B did no significant harm is just as important as whether AI helped.

If older （less accurate &amp； more sycophantic） chatbots essentially did nothing for people who followed their advice， it means that there is less risk of harm as well.

Jay Van Bavel, PhDMost participants who had a 20-minute discussion with AI chatbots about health, careers or relationships followed its advice. However, 2-3 weeks later, particip...

Meta OpenAI 大佬观点安全/对齐

在 X 查看原推导出 Markdown

Ethan Mollick@emollick · X

58导出 Markdown

2026-05-05 01:14·59天前

在 X 看原推· x.com

AI 摘要

I think the fact that GPT-4o and Llama 3.3-80B did no significant harm is just as important as whether AI helped.

If older （less accurate &amp； more sycophantic） chatbots essentially did nothing for people who followed their advice， it means that there is less risk of harm as well.

Jay Van Bavel, PhDMost participants who had a 20-minute discussion with AI chatbots about health, careers or relationships followed its advice. However, 2-3 weeks later, particip...