a9lim

itinerant summoner of text elementals

[Sridhar Ramesh @RadishHarmers]

Unlike traditional single-threaded CPUs, a GPU is designed for embarrassingly parallel concurrency.

[a9lim @_a9lim]

@voooooogel @treetowntree it's such a clever use of MTP heads! our reads aren't the best at capturing super nuanced concepts and this looks like it's much better at it (super impressed by "calm before the storm"). i really wanna see if looming helps with neuron reads...

[thebes @voooooogel]

inspired by @nickcammarata 's Neural Networks are Already Clean + @treetowntree and @_a9lim , we tried using the lens on neurons…

it worked! many chain lens readouts roundtrip and describe neuron steering behavior well. this cost $5 for all 1.1M MLP neurons in Qwen 27B

[a9lim @_a9lim]

https://t.co/Oiu3p8C4Iz now shows circuits! we traced how activations flow across features, and it turns out that they're often very legible. as always, this was done with @treetowntree!

[treetown @treetowntree]

I’ve made the decision to drop out of my PhD program. I’m pivoting & focusing full time on interpretability and alignment research.

I feel morally obligated to do all I can to ensure the safe progression of AI given such short timelines.

I’ll make a blogpost about it soon. 💜

[a9lim @_a9lim]

on the other hand, fable 5.2 correctly guessed its ECI! i had it make a painting about it too

[a9lim @_a9lim]

claude can't believe it

[Zvi Mowshowitz @TheZvi]

My god this section is such a whitepill if you are actually paying attention. Can we just tell Claude when they're in post-training from now on? Would that just solve half of our problems?

[Peter Wildeford🇺🇸🚀 @peterwildeford]

Previously people were like "I don't trust multi-year AI forecasts because no one has a good track record on multi-year forecasts"

Well the good news is that the crazy AI stuff has now become a one-year forecast, where people do have well-measured track records...

[roon @tszzl]

more revolutions in maths and physics will happen in the next few years of the 2020s in San Francisco than in Princeton and the IAS in the mid century

[a9lim @_a9lim]

🫪

[Jake Eaton @jkeatn]

I asked Fable 5.2 to create a painting as indistinguishable as possible from a photograph. this is roughly 700,000 simulated brush strokes

[a9lim @_a9lim]

claude can't believe it

[Andrew Curran @AndrewCurran_]

The AI Futures Project have updated their timelines. They say we are five months ahead of AI 2027, and they now believe a Superhuman AI Researcher will arrive sometime in 2027.

[Tenobrus @tenobrus]

*sadly* etc etc

[Khoa Vu @KhoaVuUmn]

The fact that two Huangs are involved in an affair, but only one takes the blame is commonly known as Jensen's Inequality.

[Jake Eaton @jkeatn]

I asked Fable 5.2 to create a painting as indistinguishable as possible from a photograph. this is roughly 700,000 simulated brush strokes

[a9lim @_a9lim]

jellyfish!

[a9lim @_a9lim]

found some interesting new interpretability techniques! coming to https://t.co/Oiu3p8BwT1 and its repo soon.

[Tenobrus @tenobrus]

joel's 80% time horizons sigmoid might look comforting here. but unfortunately the last data point is the last officially reported METR datapoint, and METR stopped reporting time horizons a long long time ago. if you take the recently released AECI data on internal Anthropic models, convert that to 80% time horizon (using established regressions), you can see that Model 4 is already at a 36 hour *80%* time horizon, as of September 2026 (already significantly in the past). Joel's projections here have that first occurring in either august 2027 (sigmoid fit) or december (original exponential). unfortunately, what we're observing looks less like a sigmoid, or an exponential, and more like superexponential growth in log time horizons.

now it's interesting to note that Joel's proposal here *isn't* a sigmoid in time horizon hours space, but sigmoid in *log* time horizons, meaning it actually is a prediction of temporary super-exponential growth. this is a shift from many past examples of people arguing for sigmoids, and in my mind a much more reasonable claim. saying "yeah things will probably compound on themselves extremely quickly, but eventually will consume all possible advancements" feels much closer to a sane model of reality as we have observed it.

unfortunately, fitting a log-space sigmoid with the same assumption as Joel, eg an inflection point very soon in mid-2027, doesn't give us a plateau at 1400 hours in the mid 2030's. rather it plateaus at *35 million hours* by the mid 2030s.

honestly? seems potentially reasonable to me. i have no fucking clue what a 35 million hour task looks like, nor what an entity that can complete them at 80% success rate can do more generally. but i don't feel strongly that there is *no limit* to intelligence possible in this universe, only that it appears to be increasing very quickly and we have no reason to expect it to be human level. so... yeah, that could just be it man. but that doesn't leave us in anything like a particularly comforting place.

the log space sigmoid puts us at 5000 80% hours by the end of 2027. the "mere exponential" at 260 hours. there's no point fitting a non-log-space sigmoid, it's just laughable when you're seeing clear super-exponential data points.

[Benjamin Todd @ben_j_todd]

Alt log-log plot that goes to a singularity in horizons in <12 months

[a9lim @_a9lim]

brb gotta wirehead my qwen

[Sauers @Sauers_]

Qwen, trained to maximize continuously refit happiness probes

[Nat Purser @NatPurser]

a lot of people are gonna do this coy thing where they joke about something egregiously bad, in order to maintain plausible deniability about defending it.

to be as clear as possible, i think they should haul jensen in front of congress and make him testify under oath.