synquid
Re:Show HN: Flash Attention in ~100 lines of CUDA
on Sunday March 17, 2024 @05:11AM (#39733002)
Triton has an AMD backend, although work is still ongoing.
Re:A modern self-referential weight matrix that learns to modify itself
on Wednesday April 13, 2022 @03:25PM (#31019072)
Schmidhuber has written about recursive self-improvement since his diploma thesis in the 80s: "Evolutionary principles in self-referential learning, or on learning how to learn: The meta-meta-... hook".
Your quote sounds like it could just as well have been from that thesis.
Re:Russian forces invade Ukraine after Putin orders attack
on Thursday February 24, 2022 @04:14AM (#30452569)
Poland will definitely accept them, they have publicly stated so.
Re:What the Omicron wave is revealing about human immunity
on Thursday February 03, 2022 @04:51PM (#30199477)
Myocarditis isn't that dangerous, it's acute and can be managed. Infection is much more risky.
Re:Ireland drops most Covid restrictions in wake of 'Omicron storm'
on Saturday January 22, 2022 @08:27AM (#30036412)
I wish that people who say things like "learn to live with it" actually meant it. We do indeed need to learn from the past two years, and that will result in something quite different from pretending that we're still in 2019.
Re:Scientists believed Covid leaked from Wuhan lab, but feared debate ...
on Wednesday January 12, 2022 @03:50AM (#29903924)
Zeynep has been wrong so many times during the pandemic that it hurts. She will not admit to it, though. She claimed that the Delta variant would end the pandemic, in the same way it is claimed for Omicron now. She's very naive about how the immune system works.
Her writing is generally more propaganda than science.
Re:To Understand Language Is to Understand Generalization
on Saturday December 18, 2021 @04:48AM (#29603240)
This seems very similar to the research program led by the late Patrick Henry Winton: https://groups.csail.mit.edu/genesis/index.html
Besides, I wish that causality had been mentioned more than once in passing. Due to the existence of the ladder of causality, many important queries cannot be answered by mere observation, or even by intervention; such queries require counterfactual reasoning, and structural causal models generalize because they describe something that is very invariant in the world.
Re:U.S. Kids Are Getting Nearly 70% of Their Calories from 'Ultra-Proc...
on Tuesday August 10, 2021 @04:22PM (#28133884)
They're processed foods, but not ultra-processed. Generally that means advanced factory processes and chemicals that are only used in industrial settings. Factory made bread is likely to be ultra-processed, however.
Re:Tortured phrases: A dubious writing style emerging in science
on Sunday August 08, 2021 @05:31PM (#28110298)
The smog is just the Chinese cloud.
Re:Where has all the productivity gone?
on Sunday August 01, 2021 @04:10AM (#28024671)
Sounds like a case of Goodhart's law,
"Any observed statistical regularity will tend to collapse once pressure is placed upon it for control purposes."
Re:Software engineering research is a train wreck
on Wednesday July 21, 2021 @05:44AM (#27904175)
Semantic Scholar exists, and does a decent job mostly, in my opinion.
Re:Cancer risks increase with alcohol use, leading to over 740k cases ...
on Saturday July 17, 2021 @05:25AM (#27864594)
If heavy consumers under-reported in an observational study, that would make drinking less seem worse than it really is.
Re:Cancer risks increase with alcohol use, leading to over 740k cases ...
on Saturday July 17, 2021 @04:47AM (#27864456)
But that amount is based on available research, so what is your point? See e.g. https://www.jacc.org/doi/full/10.1016/j.jacc.2017.07.710
Re:Cancer risks increase with alcohol use, leading to over 740k cases ...
on Saturday July 17, 2021 @04:38AM (#27864419)
https://www.jacc.org/doi/10.1016/j.jacc.2017.06.054
What do you make of studies such as this? Their baseline was strictly lifetime abstainers, absolutely no people who stopped drinking for medical reasons. Still, moderate drinking results in lower all-cause mortality.
Re:Cancer risks increase with alcohol use, leading to over 740k cases ...
on Saturday July 17, 2021 @04:36AM (#27864409)
Lowering your expected lifetime probably also lowers your expected time until disease.
Re:Cancer risks increase with alcohol use, leading to over 740k cases ...
on Saturday July 17, 2021 @04:30AM (#27864390)
This is a very uneducated comment, considering available research. While some doubt can be cast on the J-shaped curve, we have a lot of evidence for all-cause mortality either decreasing or at least not increasing with low levels of consumption.
Re:NSA spying row: Denmark helped US gather data on European officials...
on Monday May 31, 2021 @09:37AM (#27343910)
That sounds about right, as I read it. The sources seem to be 9 (former?) employees, confirming a classified internal report from 2015.