goodtothinkwith

@koltregaskes Why in the world is the trending starting at 2020? It clearly needs to start with agents in 2024

I have zero interest in the 50% success rate chart, the 80% success chart is a far better representation of where we are at. People are sharing the linear scale chart, which is great, but I do like looking at the log scale for another perspective. metr.org/time-horizons/

@BernieSanders You need an actual plan. Support the sovereign wealth fund

In the richest country on earth, young people shouldn't have to wait until age 40 to buy a home - 12 years after first-time buyers in 1991. Young people shouldn't have to fear a lower standard of living than their parents. We need an economy that works for all, not just the 1%.

@BernieSanders What’s your plan?

Jeff Bezos is seeking $100 billion to put robots into factories. Millions of manufacturing jobs — GONE. Driverless vehicle companies are expanding rapidly. Millions of transportation jobs — truckers, cab drivers, Uber drivers — GONE. We are not ready for what’s coming.

@DaveShapi I am looking forward to some momentum. You’re right about that

Yeah, it seems like the techno lords don't realize how unpopular they are becoming, and are blithely pushing forward. But I'm all for it. Because then there will be quite a bit of political energy to rectify things.

@LamarMK Not if they make a convertible…

I've never met a single Tesla owner who regrets buying one. Not in person. Not online. Nowhere. In fact, every owner is always excited about the same things. Their car drives itself. It keeps getting better through free software updates. And nothing else on the road even compares. Tesla owners, would you ever go back to a regular car?

@Dr_Singularity Now we’re going to get to some interesting benchmarks…

Pure insanity Claude Mythos Preview broke the METR benchmark. There is no wall. Singularity is near.

@DaveShapi This isn’t a mystery.

@deedydas @StanfordAILab Excellent. I was wondering when someone was going to turn this into a benchmark. Great news

The creators of SWE-Bench just dropped a really simple new benchmark every LLM gets 0% on. ProgramBench asks: can models recreate real executable programs (ffmpeg, SQLite, ripgrep) from scratch with no internet? We are far from saturated on model quality.

@profstonge Not necessarily causal…

Wow

@niccruzpatane They need to make a convertible again

A Tesla will ruin every other car you’ve ever owned.

@ashleytrubin Fundamental research was always enabled by leisure. The next version of that will look quite different, though populated by many of the same people.

I don't think we (academics) realize how vulnerable we are. I'm not sure how much longer tenure will last in an era when a) we've lost the public trust (for a lot of reasons, b) college/academia is a partisan political issue, and c) college-educated white-collar workers are losing their jobs to AI. We are in an incredibly privileged position relative to others, but that privilege also makes us far more precarious than most academics realize bc it makes us a target--and the threat is not just conservative politicians gunning for us but the much larger group of regular people who don't mind if we get put in our place or start losing our jobs like similar others or being asked to do tasks we don't want to do.

@AndrewYang That’s a terrific way of putting it

If we want more jobs, why are we taxing human labor? blog.andrewyang.com/p/tax-the-bots

@dioscuri @willmacaskill This argument may suffer from the same kind of issue as the paperclip problem with AI. It requires an unreasonably constrained and simple reductio

Really interesting stuff! The monoculture problem has bugged me for a long time (see my Elaine case below), and I hadn’t appreciated it could do so much work elsewhere in the axiology. One worry: the similarity metric seems to face a revenge problem. If two lives are phenomenal duplicates except one experiences red slightly more intensely, or if they’re identical for 99.999% of their duration and then differ in a single picosecond-long pleasurable sensation, do they count as distinct for saturation? If yes, infinite ethics issues might sneak back in via arbitrarily fine-grained differentiation. If no, you need a principled threshold, which could be tricky (and vulnerable to sorites-type worries). Connected thought: I’ve argued elsewhere (open.substack.com/pub/eschwitz/p…) that conscious subjects might be better understood as *types* than tokens, which if right would give Saturation a metaphysical rather than purely axiological grounding: qualitatively identical lives would literally fail to add distinct subjects, not just fail to add value. Anyway, curious whether you and Christian see the similarity metric as brute, or whether it’s supposed to track something deeper!

In collaboration with Christian Tarsney, I’ve developed a new theory of population ethics, which I call the Saturation View. I think that, from a purely intellectual perspective, it’s probably the best idea I’ve ever had. It was certainly great fun to work on. The motivation is that many views of population ethics, like the total view, suffer from some major problems. Some are already widely discussed: The Repugnant Conclusion: For any utopian outcome, there’s always another outcome containing an enormous number of barely-positive lives that is better. Fanaticism: For any guaranteed utopian outcome, there’s always some gamble with a vanishingly small probability of an even better outcome that has higher expected value. Infinitarian Paralysis: Given that the universe contains an infinite number of both positive and negative lives, no finite or infinite change to the world makes any difference to overall value. These are pretty bad! But there’s another less-discussed problem, too: The Monoculture Problem: Given fixed resources, the best-possible future consists essentially only of qualitatively identical replicas of a small number of lives. Essentially all extant impartial accounts of population ethics suffer from the monoculture problem. It follows from Pareto and Anonymity alone — you don't need totalism. And perfectly-replicable digital minds mean this is a real issue that future generations will face. But a monoculture seems far from ideal. Endless galaxies containing nothing but the same blissful experience, repeated and repeated, seem impoverished; like a song with only one note. The Saturation view deals with all these problems at once, using broadly the same machinery for all of them. The core idea is that the realisation value of a type of life (or experience) is determined by both the wellbeing of that life, and by how many very similar lives there are in the world. Endlessly creating replicas of the same identical life becomes progressively less valuable, tending to an upper bound. The total value of a world is given by the integral of realisation value over the space of types. Think of types of life as forming a landscape. Adding different types of life lights up different parts of the landscape. The value of the world is given by how fully illuminated the landscape is. Why does this help? In brief: Monoculture: Because there are diminishing returns to increasing wellbeing of very similar types, there’s greater value in having a diversity of lives. Repugnant Conclusion: The classic path to the Repugnant Conclusion requires trading a utopian world for an enormous population of barely-positive lives. But, on the Saturation view, barely-positive lives can only illuminate a tiny corner of the landscape. The path to the Repugnant Conclusion is blocked. Fanaticism: Total achievable value is bounded above. That means no tiny-probability gamble can have arbitrarily high expected value. Infinite ethics: In any infinite universe, the value of a world is finite and well-defined — even if some locations have infinite wellbeing. Unlike other approaches, this does not depend on spatiotemporal structure or choice of ultrafilter. Separability: Like nearly all non-totalist views, Saturationism is non-separable — background populations can affect how we rank options. But the violations are tame: populations with sufficiently different populations simply add, and at small scales the view behaves just like totalism. If the Saturation View is right, then the best future isn't the one where we've found the optimal experience and copy-pasted it across the cosmos. The best future is the one where we've gone exploring, and we've fully lit up the landscape of possible experiences.

@dioscuri So many pedestrian details like this I’d like to pass off to AI… more time for philosophy

Very cool. I have lots of items that I no longer need, but feel a bit too valuable to just throw away or donate. Meanwhile my available time for doing stuff like eBay sales is increasingly limited. An AI sales bot could be a gamechanger!

@imPenny2x Game theory

This is diabolical.

@Tesla isn’t it time to make a convertible again?

@elonmusk You get what you incentivize

Many

@dioscuri Entirely agree. Wooldridge seriously oversteps there

I am once again begging people to read David Marr’s Vision and properly internalise the fact that informational systems are subject to multiple levels of analysis. “Doing statistics” doesn’t mean you’re not doing other things.

@elonmusk You know the real cause, right?

This collective suicide of humanity needs to turn around!

@AlpacaAurelius Control for income and show the same data…

americans in the southeast of the US have a 20 year shorter life expectancy than those elsewhere insane difference why?