EA Forum Podcast (All audio‪)‬ EA Forum Team

(This came up for this post, but seemed worth pointing out separately in its own post, so this post is mostly copied from there. I don’t consider the referenced sources here highly reliable for how exactly prey die when swallowed live, but I wasn't able to find anything better when I checked.)
Predation is one of the most common ways for wild aquatic animals to die, perhaps the most common way (Dall et al., 1991 for penaeid shrimp, Hurst, 2007 for fish during the winter, Pauly & Palomares, 1989 for Peruvian anchoveta, Vollset et al., 2023). (Predation is also the largest cause of death among terrestrial vertebrates (Hill et al., 2019)).
Predatory/carnivorous fish typically swallow their prey whole (Lindsay, 1984, Meekan et al., 2018, Luiz, 2019, Yasuda, 1960, Amundsen & Sánchez‐Hernández, 2019, St John, 1999, Gill, 2003, Lundstedt et al., 2004), and so without tearing or chewing. This is [...]
---

First published:

June 9th, 2024


Source:

https://forum.effectivealtruism.org/posts/i4KbJubBJfsaNgAJP/being-swallowed-live-is-a-common-way-for-wild-aquatic

---
Narrated by TYPE III AUDIO.

Original by Leopold Aschenbrenner, this summary is not commissioned or endorsed by him.
Short Summary
Extrapolating existing trends in compute, spending, algorithmic progress, and energy needs implies AGI (remote jobs being completely automatable) by ~2027. AGI will greatly accelerate AI research itself, leading to vastly superhuman intelligences being created ~1 year after AGI. Superintelligence will confer a decisive strategic advantage militarily by massively accelerating all spheres of science and technology. Electricity use will be a bigger bottleneck on scaling datacentres than investment, but is still doable domestically in the US by using natural gas. AI safety efforts in the US will be mostly irrelevant if other actors steal the model weights of an AGI. US AGI research must employ vastly better cybersecurity, to protect both model weights and algorithmic secrets. Aligning superhuman AI systems is a difficult technical challenge, but probably doable, and we must devote lots of [...] ---
Outline:
(00:12) Short Summary
(02:16) 1. From GPT-4 to AGI: Counting the OOMs
(02:23) Past AI progress
(05:37) Training data limitations
(06:41) Trend extrapolations
(07:57) The modal year of AGI is soon
(09:29) 2. From AGI to Superintelligence: the Intelligence Explosion
(09:36) The basic intelligence explosion case
(10:46) Objections and responses
(14:06) The power of superintelligence
(16:28) III The Challenges
(16:31) IIIa. Racing to the Trillion-Dollar Cluster
(20:58) IIIb. Lock Down the Labs: Security for AGI
(21:05) The power of espionage
(22:09) Securing model weights
(23:46) Protecting algorithmic insights
(24:41) Necessary steps for improved security
(26:35) IIIc. Superalignment
(29:15) IIId. The Free World Must Prevail
(32:15) 4. The Project
(34:47) 5. Parting Thoughts
(35:51) Responses to Situational Awareness
The original text contained 1 footnote which was omitted from this narration.
---

First published:

June 8th, 2024


Source:

https://forum.effectivealtruism.org/posts/zmRTWsYZ4ifQKrX26/summary-of-situational-awareness-the-decade-ahead

---
Narrated by TYPE III AUDIO.

This is a link post. This is an article of mine from Recursive Adaptation. We are always looking for new collaborators as we expand our research and policy development!
Excerpt:
Suzetrigine is a novel non-opioid, sodium channel painkiller with no addictive potential, from mid-size pharma company Vertex. It has shown positive results in Phase 3 and is being submitted to the FDA for approval by the end of 2024 or early 2025. Suzetrigine is roughly equivalent in pain reduction efficacy to Vicodin (hydrocodone / APAP), which is one of the most commonly prescribed opioids. We recently reviewed the efficacy and promise of suzetrigine in this article. There are other broad non-opioid painkillers in the pipelines of other companies that may arrive over the next few years.
This is great news—can we start removing opioids from medical treatment to prevent patients from developing addictions? Unfortunately, there's a huge obstacle: like other [...]
---

First published:

June 5th, 2024


Source:

https://forum.effectivealtruism.org/posts/wkp8Xhco7NfdpgReC/the-market-failure-that-will-force-doctors-to-keep-writing

---
Narrated by TYPE III AUDIO.

Summary. Around 1.2 to 1.9 trillion fish fry have been produced by artificial propagation (breeding) annually in China, and it seems those farmed for direct human consumption and pre-harvest mortality can only account for at most around 460 billion of them (more). I estimate that 170 billion to 700 billion animals (my 80% credible interval) — probably almost all artificially propagated fish — are fed live to mandarin fish (or Chinese perch), Siniperca chuatsi, in China annually, with 9 to 55 billion of them (my 80% credible interval) alive at any time (more, Guesstimate model). By contrast, the number of farmed fish produced globally for direct human consumption is around 111 billion per year, and 103 billion alive at a time, with another 35 to 150 billion raised and stocked per year (Šimčikas, 2020). It's unclear how bad their deaths are as live feed to mandarin fish, but I’d guess they die by suffocation [...] ---
Outline:
(01:55) Acknowledgements
(02:06) Fish fry production in China
(06:30) Fish raised to become feed for farmed animals
(14:18) Mandarin fish farming
(14:30) Mandarin fish diets
(17:25) Scale of mandarin fish farming and feed
(20:45) How bad is being live fed to mandarin fish?
(24:41) Intervention ideas
(25:03) Feed substitution
(28:47) Alternatives for maintaining water quality
(29:54) Reducing mandarin fish production
(30:59) Welfare reforms for artificially propagated fish
The original text contained 9 footnotes which were omitted from this narration.
---

First published:

June 8th, 2024


Source:

https://forum.effectivealtruism.org/posts/j5XTuJqbE3CK7fTCM/170-to-700-billion-fish-raised-and-fed-live-to-mandarin-fish

---
Narrated by TYPE III AUDIO.

A superhuman combination of human and AI I believe that:
AI-enhanced organization governance could be a potentially huge win in the next few decades. AI-enhanced governance could allow organizations to reach superhuman standards, like having an expected "99.99" reliability rate of not being corrupt or not telling lies. While there are clear risks to AI-enhancement at the top levels of organizations, it's likely that most of these can be managed, assuming that the implementers are reasonable.  AI-enhanced governance could synchronize well with AI company regulation. These companies would be well-placed to develop innovations and could hypothetically be incentivized to do much of the work. AI-enhanced governance might be necessary to ensure that these organizations are aligned with public interests. More thorough investigation here could be promising for the effective altruism community. Within effective altruism now, there's a lot of work on governance and AI, but not much on using [...]
---
Outline:
(02:10) What is “AI-Assisted” Governance?
(03:08) Arguments for Governance Improvements, Generally
(04:42) Arguments for AI in Governance Improvements
(10:44) Strategy: Bootstrapping Effective AI Governance
(13:01) Objection 1: Would AI-Assisted Governance Increase AI Risk?
(15:41) Objection 2: Will AI Governance Dangerously Increase Complexity?
(17:41) Objection 3: Could AI Governance Tools Backfire by Causing Overconfidence?
(18:59) What Should Effective Altruists Do?
---

First published:

June 7th, 2024


Source:

https://forum.effectivealtruism.org/posts/nyiXej8EtoFwigrsw/a-case-for-superhuman-governance-using-ai

---
Narrated by TYPE III AUDIO.

It'd be great if anyone landing on the forum could have the opportunity to seamlessly translate its content into their language with the click of a button.
A bit like Wikipedia, but for all languages, and powered by AI; a bit like copy-pasting the URL into Google Translate, but without copy-pasting, and with better translation.
I don't know to what extent it's easy to implement into the forum, but I would have expected it wouldn't be much more complicated than the AI-generated readings of the posts. But I could be wrong about this, of course.

Image from Freepik
---

First published:

June 6th, 2024


Source:

https://forum.effectivealtruism.org/posts/h2kB7fu56fhkHvJTh/in-favor-of-an-ai-powered-translation-button-on-the-ea-forum

---
Narrated by TYPE III AUDIO.