← 回總覽

应对 AI 能力与可靠性的虚假二分法

📅 2026-04-08 23:43 Gary Marcus 人工智能 3 分鐘 2882 字 評分: 74
AI 安全 虚假二分法 AI 可靠性
📌 一句话摘要 Marcus 驳斥了“AI 不可能既不可靠又危险”的观点,认为这是一种虚假二分法。 📝 详细摘要 针对一位用户质疑 LLM 怎么可能既“不好用”又“能力太强/太危险”,Marcus 指出这是一种虚假二分法。他认为,一个系统可以在推理方面不可靠,但同时仍具备足够的能量(例如在网络攻击或虚假信息传播方面)造成极端危害。 📊 文章信息 AI 评分:74 来源:Gary Marcus(@GaryMarcus) 作者:Gary Marcus 分类:人工智能 语言:英文 阅读时间:1 分钟 字数:91 标签: AI 安全, 虚假二分法, AI 可靠性 阅读推文

Title: Addressing the False Dichotomy of AI Capability vs. Relia...

URL Source: https://www.bestblogs.dev/status/2041904683338625283

Published Time: 2026-04-08 15:43:07

Markdown Content: Skip to main content ![Image 1: LogoBestBlogs](https://www.bestblogs.dev/ "BestBlogs.dev")Toggle navigation menu Toggle navigation menuArticlesPodcastsVideosTweetsSourcesNewsletters

⌘K

Change language Switch ThemeSign In

Narrow Mode

Addressing the False Dichotomy of AI Capability vs. Reliability

Addressing the False Dichotomy of AI Capability vs. Reliability

![Image 2: Gary Marcus](https://www.bestblogs.dev/en/tweets?sourceId=SOURCE_c2fb88) ### Gary Marcus

@GaryMarcus

literally fourteen minutes after my last explanation of why this is a false dichotomy 🤦‍♂️

!Image 3: Michael Foster

#### Michael Foster

@realmfoster · 2h ago @GaryMarcus so do LLM’s not work or are they too capable? Which is it Gary?

3

0

6

4,143

Apr 8, 2026, 3:43 PM View on X

5 Replies

3 Retweets

13 Likes

3,381 Views ![Image 4: Gary Marcus](https://www.bestblogs.dev/en/tweets?sourceid=c2fb88) Gary Marcus @GaryMarcus

One Sentence Summary

Marcus dismisses the argument that AI cannot be both unreliable and dangerous as a false dichotomy.

Summary

Responding to a user who questions how LLMs can be both 'not working' and 'too capable/dangerous,' Marcus points out that this is a false dichotomy. He argues that a system can be unreliable in its reasoning while still possessing enough capability (e.g., in cyberattacks or disinformation) to be extremely hazardous.

AI Score

74

Influence Score 5

Published At Today

Language

English

Tags

AI Safety

False Dichotomy

AI Reliability HomeArticlesPodcastsVideosTweets

Addressing the False Dichotomy of AI Capability vs. Relia...

查看原文 → 發佈: 2026-04-08 23:43:07 收錄: 2026-04-09 02:00:30

🤖 問 AI

針對這篇文章提問,AI 會根據文章內容回答。按 Ctrl+Enter 送出。