#evidence
9 discoveries · Hand-picked discoveries from the open web, newest first.
-
We need more decorrelated evidence
AI agents do a growing share of work on each new model. Then, through some process, the results are combined into an overall safety assessment, which labs use to estimate the probability that the next model is not catastrophically misaligned. I feel that, several months later, we appreciate further the foresight of Automated Alignment is Harder Than You Think (Bowkis et al. 2026): much is handed off to the machines because of the urgency of fast timelines, and we miss many things due to the problem of undetected systematic errors . MS Paint drawing by Fable 5.2. In this post, I’ll develop a si
LessWrong · EN
-
The Mistake This Filmmaker Made After His Film Was Pirated
Ryan McCoy's first feature found an audience worldwide, but he's never made a dollar from it. It's a filmmaker's worst nightmare as well as a paradox. You want people to know your work, and you want people to watch it. But if someone steals your work from you and takes on a life of its own separate from you, you don't benefit at all, monetarily or otherwise. McCoy wrote, produced, and starred in Evidence , a found-footage horror film directed by Howie Askins and shot on McCoy's days off from his regular job. It played Glasgow FrightFest in 2012, per Eye for Film . Its first sale was in the U.K
Nofilmschool · EN
-
Evidence that the ‘talker’ has introspective access to the ‘doer’
TL;DR: We train models to predict their own behaviour in a chat setting. This transfers to making predictions about cheating behaviour in a coding setting. We provide evidence that self-prediction utilises introspective-access to internal states, and that this introspective access is shared across contexts between different parts of the model. Introduction Frontier AIs can exhibit different alignment propensities in different contexts. For instance, recent egregious misalignment incidents have typically happened in a coding or cyber context, or on impossible evaluation tasks, wherein the model
LessWrong · EN
-
Israel’s Shin Bet finds no evidence of foreign election interference
Al Jazeera English
-
After 40 years, CERN finds evidence of a one-in-10-billion particle decay
CERN scientists have finally found evidence of a particle decay so rare that it happens only once in roughly 10 billion decays, ending a search spanning more than 40 years. The breakthrough could help physicists uncover hidden particles and forces that lie beyond the reach of today's most powerful accelerators.
ScienceDaily · EN
-
China should share the evidence behind animal-testing alternatives openly
Nature, Published online: 03 November 2026; doi:10.1038/d41586-026-03561-7 China should share the evidence behind animal-testing alternatives openly
Nature News · EN
-
Large Hadron Collider reveals evidence of Big Bang matter in surprisingly small collisions
Scientists at CERN's Large Hadron Collider have uncovered compelling new evidence that even relatively small atomic collisions can recreate the extraordinary state of matter that filled the universe moments after the Big Bang. Known as quark-gluon plasma, this exotic substance forms at temperatures more than 100,000 times hotter than the center of the Sun, causing ordinary particles to break apart into their fundamental building blocks.
ScienceDaily · EN
-
There’s Still Too Much Uncertainty About Conflict of Interest Bias
A paper published this month got me frustrated, yet again, about this situation: There really is still too much uncertainty about conflict… The post There’s Still Too Much Uncertainty About Conflict of Interest Bias appeared first on Absolutely Maybe .
Absolutely Maybe · EN
-
Beyond Belief: How Evidence Shows What Really Works (Book Review)
There are several reasons to buy Helen Pearson’s excellent new book, Beyond Belief: How Evidence Shows What Really Works—whether you want to… The post Beyond Belief: How Evidence Shows What Really Works (Book Review) appeared first on Absolutely Maybe .
Absolutely Maybe · EN