Back
AI Safety - EA Forum Topics
blogCredibility Rating
3/5
Good(3)Good quality. Reputable source with community review or editorial standards, but less rigorous than peer-reviewed venues.
Rating inherited from publication venue: EA Forum
A community aggregator for EA-aligned AI safety discourse; useful for tracking emerging concerns and debates in the field, but not a primary research source. Best used as a discovery tool for specific posts.
Metadata
Importance: 45/100wiki pagehomepage
Summary
The EA Forum's AI safety topic page aggregates community discussions, research posts, and quick takes on reducing existential risks from advanced AI. It serves as a living index of community thinking spanning technical safety, policy, capacity-building, and emerging concerns like superpersuasive AI and evaluation saturation.
Key Points
- •Aggregates thousands of posts (4664+) on AI safety topics including alignment, governance, policy, and community building
- •Quick takes highlight emerging concerns: superpersuasive AI eroding expert epistemic calibration and the 'eval singularity' where capability growth outpaces measurement
- •Features capacity-building discussions, fellowship announcements, and community space fundraising reflecting the field's organizational growth
- •Responsible Scaling Policy v3 and long-term ideological risk posts represent high-engagement policy and governance content
- •Functions as a real-time pulse of EA-adjacent AI safety community priorities rather than a single research contribution
Cited by 1 page
| Page | Type | Quality |
|---|---|---|
| EA Global | Organization | 38.0 |
Cached Content Preview
HTTP 200Fetched Mar 15, 202612 KB
AI safety - EA Forum
This website requires javascript to properly function. Consider activating javascript to get access to all site functionality. AI safety AI safety Studying and reducing the existential risks posed by advanced artificial intelligence Posts Wiki Write new Subscribe New & upvoted 9 What concerns people about AI? spencerg spencerg · 7h ago · 3 m read 0 0 123 The case for AI safety capacity-building work
abergal abergal · 5d ago · 27 m read 3 3 96 New Video: If Anyone Builds It, Everyone Dies ChanaMessinger ChanaMessinger , Aric Floyd + 0 more · 4d ago · 4 m read 8 8 49 We do not live by course alone Joe Rogero Joe Rogero · 3d ago 3 3 69 Mox is the largest AI Safety community space in San Francisco. We're fundraising! Rachel Shu Rachel Shu , Austin + 0 more · 8d ago · 9 m read 5 5 27 Carving up the future Niki Dupuis Niki Dupuis · 3d ago · 5 m read 2 2 68 Announcing: AGI & Animals Debate Week Toby Tremlett🔹 Toby Tremlett🔹 · 9d ago · 4 m read 3 3 8 Futurekind Spring Fellowship 2026 - Applications Now Open Khushbu Sainani Khushbu Sainani , Aditya_Karanam , Electric Sheep | + 0 more · 1d ago · 1 m read 0 0 128 Responsible Scaling Policy v3 Holden Karnofsky Holden Karnofsky · 18d ago 8 8 194 Long-term risks from ideological fanaticism David_Althaus David_Althaus , Jamie_Harris , Vanessa_Sarre , Clare_Diane , Will Aldred + 0 more · 1mo ago · 100 m read 30 30 Load more (10/4664) Quick takes
5 Noah Birnbaum 3d 1 Experts currently treat being persuaded as reasonably good evidence that something is true — their judgment is calibrated enough that when they find an argument convincing, that's correlated with the argument actually being correct. This allows them to update readily in light of new evidence, and is a big part of how intellectual progress happens: lots of innovation and advances in basically every subject come down to experts taking sometimes weird new ideas seriously.
One worry I have about superpersuasive AI is that it could erode this. If a superpersuasive AI can convince experts of things regardless of whether those things are true, experts may cease to see themselves being persuaded as good evidence that something is true — and start treating it the way laypeople do. Laypeople are typically hesitant to take on new, truth-tracking beliefs in light of new information, and (to some degree) rationally so: the fact that someone was able to convince a layperson of something is just not very strong evidence that it is in fact true. Experts might end up in the same position — only updating rarely, and in ways that are often unrelated to the truth.
This would be quite bad. If experts lose their capacity to reliably update on genuine evidence, we could significantly slow the rate of intellectual progress (which could be very important for making AI go well!). This is, I think, an underappreciated argument for caring about AI for epistemics — curious what others think. 27 Ben_West🔸 1mo 2 The AI Eval Sing
... (truncated, 12 KB total)Resource ID:
721b826caa2020b3 | Stable ID: YjdjYWI2Mj