Post by @theodoraward@wandering.shop

oh cool some more guys to look up. who's the crack squad saving the world from the big bad robot. i've talked about HJALMAR WIJK. we've go tom cunningham, a guy with a phd in economics who writes papers about, uh, overcoming bias. we've got parker whitfill, a guy "on leave" from a phd in economics. and we've got this fucking guy

picture on righ: white guy in a green felt hat and a sweater tied around his neck with a white t-shirt and a backpack standing in a field

Nate Rush

CTO

Nate previously was the founder/

CEO of Mito, where he built Python

tooling deployed at the largest /

banks in the US. Before Mito, he

worked as a researcher at the

Ethereum Foundation on consensus

algorithms. He loves code that

makes a real-world impact, bikes, i A

and mushroom foraging. — picture on righ: white guy in a green felt hat and a sweater tied around his neck with a white t-shirt and a backpack standing in a field Nate Rush CTO Nate previously was the founder/ CEO of Mito, where he built Python tooling deployed at the largest / banks in the US. Before Mito, he worked as a researcher at the Ethereum Foundation on consensus algorithms. He loves code that makes a real-world impact, bikes, i A and mushroom foraging.

0 Boosts

0 Likes

Theodora Ward

@theodoraward@wandering.shop 1 hr. ago

first link is to this paper. https://arxiv.org/abs/2609.36054 im sorry thats too many guys to work on a paper. too many annoying guys. geoffrey hinton leaving comments in the google doc fucking kill me man

arXiv.org

What if automating AI R&D triggers an intelligence explosion?

In contrast to even a year ago, AI systems now write most of the code inside the companies that build them. As more of the AI research and development (R&D) pipeline is automated, could AI progress radically accelerate in an "intelligence explosion," where years of advances are compressed into months or less? Preliminary evidence suggests that it could. In this work, we assess this evidence, analyze an intelligence explosion's potential impacts, and propose policy responses. AI systems are on track to automate most AI R\&D work within a few years, and possibly all of it. If this triggers an intelligence explosion, it could dramatically bring forward AI's benefits, but also pose extreme risks: capabilities growth could accelerate far beyond what society can keep up with, humanity could lose control over superhuman AI systems, and checks on power within and between states, companies, and branches of government could be severely eroded. Although there remains much uncertainty about these possibilities, the high stakes warrant serious further attention. Policymakers should urgently obtain more visibility into the automation of AI R&D, develop ways to steer and constrain an intelligence explosion, and prepare society to adapt to an intelligence explosion's impacts.

Theodora Ward

@theodoraward@wandering.shop 1 hr. ago

ok checking in on the "epoch capabilities index," an Internal Version of which is what metr is popping their monocles about

im going to keep reading because im an idio but im sorry if you sincerely deploy the idea of IQ in 2026 you have some real spiritual reckoning to do. because you are a eugenicist

Common questions
What does the ECI represent? -
Similar to an IQ test, ECI is designed to capture a broad, underlying capability useful — Common questions What does the ECI represent? - Similar to an IQ test, ECI is designed to capture a broad, underlying capability useful

Theodora Ward

@theodoraward@wandering.shop 1 hr. ago

i fucked up the thread. i was too tilted to make the thread work.

Theodora Ward

@theodoraward@wandering.shop 1 hr. ago

look im way out of my league when it comes to stats stuff so maybe this isn't actually as ridiculous as it seems but like. lol

ECI scores are abstract values that can’t be interpreted in isolation, but they can be
compared across models to judge meaningful differences in underlying capabilities.
See the next question for more details.

How should I interpret ECI values? -
Absolute ECI values are meaningless by themselves, but meaningful comparisons can
be made between models. The ECI scale is linear; in theory, a 10 point jump should be
equally “impressive” when moving from 100 to 110 as it is going from 140 to 150. That
said, just because ECl is linear doesn’t mean that other things we care about are linear
in ECI units. For instance, at the time of ECI’s launch, a 5 point gain in ECI appeared to
roughly correspond to a doubling of the METR Time Horizon. Similar principles hold
when looking at trends in ECI; absolute slope values are meaningless, but relative
changes in slope indicate faster or slower progress.

The scale of ECI scores is arbitrary. To put values in a convenient range, we have
scaled the raw scores so that Claude 3.5 Sonnet = 130 and GPT-5 = 150. As future
models and benchmarks are incorporated, these calibration points may be revisited.
Note that as with Elo scores, there is no maximum achievable ECI.

Values are not linearly related to benchmark accuracy, since benchmarks differ in both
difficulty and slope. — ECI scores are abstract values that can’t be interpreted in isolation, but they can be compared across models to judge meaningful differences in underlying capabilities. See the next question for more details. How should I interpret ECI values? - Absolute ECI values are meaningless by themselves, but meaningful comparisons can be made between models. The ECI scale is linear; in theory, a 10 point jump should be equally “impressive” when moving from 100 to 110 as it is going from 140 to 150. That said, just because ECl is linear doesn’t mean that other things we care about are linear in ECI units. For instance, at the time of ECI’s launch, a 5 point gain in ECI appeared to roughly correspond to a doubling of the METR Time Horizon. Similar principles hold when looking at trends in ECI; absolute slope values are meaningless, but relative changes in slope indicate faster or slower progress. The scale of ECI scores is arbitrary. To put values in a convenient range, we have scaled the raw scores so that Claude 3.5 Sonnet = 130 and GPT-5 = 150. As future models and benchmarks are incorporated, these calibration points may be revisited. Note that as with Elo scores, there is no maximum achievable ECI. Values are not linearly related to benchmark accuracy, since benchmarks differ in both difficulty and slope.

~1 more replies

Theodora Ward

@theodoraward@wandering.shop 1 hr. ago

im stepping back and trying to figure out how the fuck to even approach the claims that getting faster at improving on internal benchmarks is dangerous because of the p(foom), it probably isn't worth trying. but i will say that the timing of this report given anthropic's flop-ass prospectus makes even clearer than it already was that, though i do think that METR are truly cooked enough to fully believe that they are intervening, their primary actual function is as anthropic's PR wing