EA Forum Podcast (All audio)

A common framing of the AI alignment problem is that it's a technical hurdle to be overcome. A clever team at DeepMind or Anthropic would publish a paper titled "Alignment is All You Need," everyone would implement it, and we'd all live happily ever after in harmonious coexistence with our artificial friends.

I suspect this perspective constitutes a category mistake on multiple levels. Firstly, it presupposes that the aims, drives, and objectives of both the artificial general intelligence and what we aim to align it with can be simplified into a distinct and finite set of elements, a simplification I believe is unrealistic. Secondly, it treats both the AGI and the alignment target as if they were static systems. This is akin to expecting a single paper titled "The Solution to Geopolitical Stability" or "How to Achieve Permanent Marital Bliss." These are not problems that [...]

---

Outline:

(01:10) The Problem of Aligned To Whom?

(03:27) The Target is Moving

---

First published:

Source:

Linkpost URL:
https://www.lesswrong.com/posts/wgENfqD8HgADgq4rv/why-solving-alignment-is-likely-a-category-mistake

---

Narrated by TYPE III AUDIO.

---

Images from the article:

Apple Podcasts and Spotify do not show images in the episode description. Try Pocket Casts, or another podcast app.

We’re excited to release a new AI governance research agenda from the MIRI Technical Governance Team. With this research agenda, we have two main aims: to describe the strategic landscape of AI development and to catalog important governance research questions. We base the agenda around four high-level scenarios for the geopolitical response to advanced AI development. Our favored scenario involves building the technical, legal, and institutional infrastructure required to internationally restrict dangerous AI development and deployment (which we refer to as an Off Switch), which leads into an internationally coordinated Halt on frontier AI activities at some point in the future. This blog post is a slightly edited version of the executive summary.

We are also looking for someone to lead our team and work on these problems, please reach out here if you think you’d be a good fit.

The default trajectory of AI [...]

---

Outline:

(04:48) Off Switch and Halt

(07:37) US National Project

(09:53) Light-Touch

(12:07) Threat of Sabotage

(14:27) Understanding the World

(15:11) Outlook

---

First published:

Source:

Linkpost URL:
https://techgov.intelligence.org/research/ai-governance-to-avoid-extinction

---

Narrated by TYPE III AUDIO.

---

Images from the article:

Apple Podcasts and Spotify do not show images in the episode description. Try Pocket Casts, or another podcast app.

Introduction

As a community builder, I sometimes get into conversations with EA-skeptics that aren't going to sway the person I'm talking to. The Tree of Questions is a tool I use to be more sure of having effective conversations, faster identifying the crux. Much of this is inspired by Scott Alexander's "tower of assumptions" and Benjamin Todd's ideas of The Core of EA.

The Tree

1. a trunk with core ideas almost all EAs accept - and without which you have to jump through some very specific hoops in order to agree with standard EA stances.
2. Branches for different cause areas or ideas within EA, such as longtermism. If you reject the trunk, there's no point debating branches. 

All too often, I find people cutting off a branch of this tree, and then believing they've cut down the entire thing. "I'm not an EA because I don't believe in [...]

---

Outline:

(00:09) Introduction

(00:34) The Tree

(02:02) Altruism

(02:26) Concede:

(03:02) Debate briefly:

(03:26) Effectiveness

(03:49) Concede:

(04:28) Debate briefly:

(05:09) Comparability

(05:31) Concede:

(06:30) Institutional Trust

(07:19) Further Discussions

---

First published:

Source:

---

Narrated by TYPE III AUDIO.

The basic situation

The world is wild and terrible and wonderful and rushing forwards so so fast.

Modern economies are tremendous things, allowing crazy amounts of coordination. People have got really very good at producing stuff. Long-term trends are towards more affluence, and less violence.

The enlightenment was pretty fantastic not just for bringing us better tech, but also more truthseeking, better values, etc.

People, on the whole, are basically good — they want good things for others, and they want to be liked, and they want the truth to come out. This is some mix of innate and socially conditioned. (It isn’t universal.) But they also often are put in a tight spot and end up looking out for themselves or those they love. The hierarchy of needs bites. Effective altruism often grows from a measure of privilege.

The world is shaped by economics and by incentives and [...]

---

Outline:

(00:11) The basic situation

(01:17) AI enters the picture

(02:50) The crucible

(04:44) Heating up

(05:54) The shape of technology

(07:11) The case for optimism

(09:48) What is needed?

(13:26) Against an overly-narrow focus

The original text contained 1 footnote which was omitted from this narration.

---

First published:

Source:

---

Narrated by TYPE III AUDIO.

---

Images from the article:

Apple Podcasts and Spotify do not show images in the episode description. Try Pocket Casts, or another podcast app.

Vetted Causes is excited to announce our 2025 charity recommendations:

• Animal Legal Defense Fund
• Fish Welfare Initiative
• Shrimp Welfare Project

Each of these recommended charities has received a published review (linked above), and a $1,000 donation in support of their work.

Please join us in recognizing these organizations for their outstanding contributions!

---

First published:

Source:

---

Narrated by TYPE III AUDIO.

I very frequently hear the statement "the best charities are over 1000x more cost effective than the average". This is often alongside the accompanying graph.

Where does this figure come from? Most sources link it to Toby Ord's 2013 paper "The Moral Imperative toward Cost-Effectiveness in Global Health".

The data in this paper comes from the 2006 paper "Disease Control Priorities in Developing Countries".

How should this information affect our claims as EAs?

1. We should not extrapolate this claim to charities unless we have direct evidence (please comment with the best evidence you have seen). We should also not extrapolate this to fields outside of health development- in particular fields that involve creating change in complex systems, for which change-making is less measurable and less linear.

2. We should be transparent about what we mean by 'effective'. Just because some charities use less measurable methods [...]

---

First published:

Source:

---

Narrated by TYPE III AUDIO.

---

Images from the article:

Apple Podcasts and Spotify do not show images in the episode description. Try Pocket Casts, or another podcast app.

(Disclaimer: Post written in a personal capacity. These are personal hot takes and do not in any way represent my employer's views.)

TL;DR:** I do not think we will produce high reliability methods to evaluate or monitor the safety**[1]** of superintelligent systems** via current research paradigms, with interpretability or otherwise[2]. Interpretability seems a valuable tool here and remains worth investing in, as it will hopefully increase the reliability we can achieve. However, interpretability should be viewed as part of an overall portfolio of defences: a layer in a defence-in-depth strategy. It is not the one thing that will save us, and it still won’t be enough for high reliability.

Introduction

There's a common, often implicit, argument made in AI safety discussions: interpretability is presented as the only reliable path forward for detecting deception in advanced AI - notably argued in Dario Amodei's recent “The Urgency of [...]

---

Outline:

(00:58) Introduction

(02:59) High Reliability Seems Unattainable

(05:16) Why Won't Interpretability be Reliable?

(07:50) The Potential of Black-Box Methods

(08:52) The Role of Interpretability

(12:07) Conclusion

The original text contained 5 footnotes which were omitted from this narration.

---

First published:

Source:

---

Narrated by TYPE III AUDIO.

Tl;dr: Over the past few months, the EA Infrastructure Fund's (EAIF) grantmaking has increased significantly, and funding has become more of a constraint. Your donations would help us continue supporting impactful projects.

• EA Infrastructure Fund aims to increase the impact of projects that use the principles of effective altruism, by increasing their access to talent, capital, and knowledge.
• In November 2024, we posted saying that EAIF wasn’t funding constrained at the time.
• We’ve recently increased our grantmaking. In 2024 we made $1.9M in grants. So far in 2025 we have made $0.9M in grants and are only a third of the way through the year.
• Our total available fund balance is currently $2.6M, down from $3.3M in November 2024 when we last posted.
• We project that at our 2024 funding bar (i.e. our sense of how promising an application needs to be for us to fund it), we [...]

---

Outline:

(00:28) The situation

(01:54) Why has this happened?

(01:57) 1. Growth in demand from established organisations

(02:38) 2. Our active grantmaking has worked

(03:33) 3. We continue to fill an important role in the EA funding ecosystem

(04:25) Room for additional funding

The original text contained 4 footnotes which were omitted from this narration.

---

First published:

Source:

---

Narrated by TYPE III AUDIO.

Introduction

This post is intended to be read after reading our previous post outlining Shrimp Welfare Project's 2030 Vision & Absorbency Plans. This post therefore assumes some baseline knowledge of our Humane Slaughter Initiative, which can be found in that post.

Problem(s)

Market Incentives

How do we create the incentives for the market to shift to pre-slaughter stunning?

The suffering shrimps experience at the end of their lives can be reduced by rendering them unconscious prior to slaughter through the use of electrical stunning technology. However, producers currently have very little incentive to implement pre-slaughter stunning as it's expensive and few buyers require it. Even relatively forward-thinking retailers often can’t push an initiative through internally without some leverage (as they need to be able to convince their higher-ups that this is something worth doing and to persuade their suppliers to invest in the technology).

To solve this, we buy [...]

---

Outline:

(00:12) Introduction

(00:30) Problem(s)

(00:33) Market Incentives

(01:51) Market Transition

(02:54) Strategy

(02:57) Existing Policy Tools

(04:23) Cage-Free Impact Incentives

(06:38) Shrimp Stunning Credits

(07:31) Buying Credits

(08:21) Precision Aquaculture

(10:17) How it works

(12:16) Further thoughts

(12:20) Cost-Effectiveness

(13:25) Expanding Credits

(15:28) How You Can Help

---

First published:

Source:

---

Narrated by TYPE III AUDIO.

---

Images from the article:

Apple Podcasts and Spotify do not show images in the episode description. Try Pocket Casts, or another podcast app.

Introduction

~440 billion shrimps are farmed each year [1]. This is over 5x the total number of all farmed land animals put together [2].

Many farmed shrimps suffer from conditions that can and should be addressed, such as poor water quality, high stocking densities, inhumane slaughter methods, and avoidable mutilations (such as eyestalk ablation) [3].

Shrimp Welfare Project is an organisation of people who believe that shrimps are capable of suffering and deserve our moral consideration [4].

We aim to cost-effectively reduce the suffering of billions of farmed shrimps.

This post is essentially an expanded version of our 2025 Funding Proposal.
If you want the TL;DR version of this post, I'd recommend reading that.

(Shr)Impact and Vision

Shrimp Welfare Project has four workstreams, two of which we consider our Core or Foundational workstreams - those are Corporate Engagement and Farmer Support. Two more are relatively new, but we [...]

---

Outline:

(00:12) Introduction

(01:01) (Shr)Impact and Vision

(01:39) Core: Corporate Engagement

(02:05) Problem (and Context)

(03:12) Strategy

(06:01) Achievements

(06:31) 2030 Vision

(08:24) Core: Farmer Engagement

(08:45) Problem (and Context)

(09:36) Strategy

(11:42) Achievements

(13:51) 2030 Vision

(16:52) New: Research & Policy

(17:16) Problem (and Context)

(18:26) Strategy

(19:35) 2030 Vision

(22:17) New: Precision Welfare

(22:42) Problem (and Context)

(23:50) Strategy

(25:15) 20(26) Vision

(26:49) Absorbency Plans

(27:26) ~$50k-250k+

(27:31) Scaling the Humane Slaughter Initiative (High)

(28:37) Traceability (High)

(29:11) Shrimp Welfare industry Marketing Campaign, Conference, or Webinars (Low)

(29:52) ~$250k-1M

(29:57) Large Stunner commitment (Medium)

(30:40) New Stunners (Medium)

(31:00) ~$1M+

(31:04) Creating Market Incentives or, Shrimp Stunning Credits (High)

(31:48) Supporting Us

The original text contained 2 footnotes which were omitted from this narration.

---

First published:

Source:

---

Narrated by TYPE III AUDIO.

---

Images from the article:

Apple Podcasts and Spotify do not show images in the episode description. Try Pocket Casts, or another podcast app.