Optimal Word Combination

the new state of things does not encourage deep work or craft except in people who have a really Benedictine self discipline

@charlesmurray Falls a bit short of high dimensional taunting

I’m pleased.

New @CGDev paper! What can you actually do about lead poisoning? Nutritional supplements probably help a bit, are very cheap, but far from solve the problem. A systematic review/meta-analysis of a very thin literature (more research needed!)

@VCBrags That is not persuasive though (no idea how he would be, given the considerations)

🚨Drake mentions venture capitalist Peter Thiel on his new album Iceman "Zero to one, but I did it twice. Thiel told me monopoly, I took his advice. Palantir in my pocket, Founders Fund on ice"

Cunningham & Whitfill are really onto something here, showing how two obvious measures of speed-up from AI aren't right. But their own measure has issues too. Let me explain, introducing a 4th measure too.

Sent another 64 ETH to the Animal Welfare Fund. I encourage others to think and act more in support of our non-human cousins too! The extreme suffering we're imposing on them in the billions is not something we talk about often, but it continues to be one of the larger blights on humanity. And I'm getting optimistic that this century we can finally end it. Farming practices are improving, synthetic alternatives are improving. Also, in my recent experience, good old low-tech vegetarian and vegan food has improved massively worldwide over the last ten years; I encourage anyone who has tried it long before and given up to take second look; there are far more healthier and tastier options today than the "pasta and salad" you would often get ten years ago.

If you could save one person you love or reduce suffering for 100 strangers… what would you choose? In this episode of More Than Good, Peter Singer, Grace Adams, Krystian Seibert, and Yip Fai Tse tackle questions about wealth, obligation, and what it actually means to live a good life. They get into why most charitable giving doesn't go where suffering is greatest, why lower-income people often give more proportionally than the wealthy, and whether you can enjoy comfort without looking away from the world. More Than Good is a new podcast fromEffective Altruism Australia @EA_Aus youtube.com/watch?v=fF8Y9v…

@albrgr The problem is that managers only do this if they anticipate capability to hide from higher ups or larger people, determined by the topology including each individual case, similar to potemkin village dynamics in general like in Do Androids Dream Of Electric Sheep.

@robertwiblin I think I really liked your podcast with ocb on defense layers but also the one with aktipis, redundancy here is a systemic risk not a diminishing returns problem.

Reading this excellent new piece from the Economist it's clear everything being done to tackle AIxBio is hopelessly insufficient: 1. Guardrails - reasonably effective for closed models, won't cover open source or leaked weights 2. Pre-training data filtering - AIs can infer missing info, won't cover open source 3. DNA synthesis screening - would buy significant time if done well, but only a few countries clearly moving in this direction (US/UK/NZ) 4. New tech / 'defensive acceleration' - might work but we're starting a decade behind where we need to be and barely trying to catch up even now So how are we not just toast? Seems like we are mostly stuck relying on 'nobody competent will really try' as our defence. If that fails having everyone perpetually wear serious PPE when they leave the home might become the fallback position to keep society running for an extended period. Pretty grim.

@KelseyTuoc Everyone who has truly grokked high dimensional epistasis variance endorses this, which makes the current situation strange and difficult to understand. Some don't want good results, hide this, persist with novel strategies, including obfuscation, focusing assets, odd angles etc.

In the Argument today, Nicholas Bagley and Robert Gordon argue that Democrats have a public sector unions problem: if you want good results, you have to be able to hire the best people and fire the worst ones.

@deanwball Setting aside which is more desirable, permanent authoritarianism seems more likely than the state withering away die to no longer being necessary. So twilight century of the state feels optimistic to me!

"Sherwin Williams, the largest global United States-based paint company, continues to sell lead paints in Mexico."

Anton is on point here, as always. Policymakers and other strategists tend to implicitly assume a level of frontier AI abundance that I do not expect to materialize over the next few years. Scarce frontier AI profoundly changes the political economy of AI.

In life, everything is a wager. Whether you realize it or not, you are constantly making implicit and explicit predictions about the future state of reality. To live is to predict. So when you are faced with something like Mythos, and you say, “this is just ‘doomer hype’!,” what you are really doing is making a bet against model capabilities growth, and thus ultimately you are making a broad directional bet against deep learning, which has usually been a pretty bad bet to make. I am surprised that so many people—people who are otherwise AI optimists!—continue to make these bets against deep learning. They keep being wrong, and the less humble among them have torched their credibility with anyone paying attention. So ask yourself, when you make claims about AI and its future: “am I making an implicit bet against deep learning in a broad directional way?”

@albrgr @cayimby Hard to verify anything through the campaign hardware because there they're talking heads that can just be spending the whole time doing the opposite: building houses in the middle of nowhere, in service of nimbys running the hardware, such that they goodhart supply metrics.

Pretty cool: @cayimby declined to endorse anyone for California governor because all the top candidates have strong housing plans

@MATSprogram Mats was based at lighthaven for a while which was pretty concerning independence-wise (other stuff is great but not that). How is it going now?

1/ 🚨 MATS Autumn 2026 applications are now open. 10-week fully-funded fellowship for aspiring AI alignment, security & governance researchers and field-builders. 📍 Berkeley + London 📅 Sep 28 – Dec 4, 2026 💰 $5000/month stipend + $8,000/month compute Apply by June 7 AoE ↓

alignment theory: we need fifty years worth of shard theory progress in five years alignment practice: lets make sure to tell it no goblins twice so we're absolutely sure there's no goblins

New paper: research agenda for secret loyalties Imagine a frontier model that has been trained to covertly advance a specific actor's interests (a nation-state, a CEO, an adversary). @joemkwon argues this is an urgent, neglected, and addressable problem. 🧵