apsec112

17 ms·

211.
Things I learned talking to the new breed of scientific institution(lesswrong.com)
1 points by apsec112 2y ago|0 comments
212.
Alternate RNA decoding results in stable and abundant proteins in mammals(biorxiv.org)
3 points by apsec112 2y ago|0 comments
213.
Half of AGI safety researchers have left OpenAI(archive.is)
9 points by apsec112 2y ago|0 comments
214.
Bioaccumulation of Microplastics in Decedent Human Brains(ncbi.nlm.nih.gov)
131 points by apsec112 2y ago|153 comments
215.
by apsec112 2y ago
Sure, but getting a bunch of apparently real people to edit their LinkedIn profiles to list this as their fake full-time job (in place of whatever real job they presumably have) seems like too much work for a prank
216.
by apsec112 2y ago
This seems like a real company? They have a LinkedIn page with several dozen real-seeming employees: https://www.linkedin.com/company/black-mesa-technology/
217.
by apsec112 2y ago
It doesn't show this! It shows that there's a correlation between low magnesium, and biomarkers which are themselves correlated with disease. But that could mean anything. Maybe foods that are high in magnesium are also more satia
218.
by apsec112 2y ago
Weird that they don't mention Triton? I only skimmed it, but I'm not sure what the pros and cons would be vs. Triton, which is the tool I'd use if I wanted custom quantized inference kernels.
219.
by apsec112 2y ago
I mean, the ultimate reason why the US doesn't have China's electric vehicle policy is that the US is a democracy, and it would be massively unpopular. The median US voter is, like, a 40-something mother in a random suburb of Penn
220.
Artificial Wombs When(asteriskmag.com)
1 points by apsec112 2y ago|0 comments
221.
by apsec112 2y ago
In 2014, 1.7% of workers were paid the minimum, and the 10th percentile hourly wage was $8.62, which you could reasonably say was barely more than the minimum. So there were a lot of people making minimum wage or fairly close to it: https:
222.
by apsec112 2y ago
Nominal wages have increased, but in 2022 they increased slower than inflation, so people became poorer in real terms: https://www.stlouisfed.org/on-the-economy/2023/feb/nominal-w... The number of people ma
223.
by apsec112 2y ago
What observation, if you saw it, do you think would falsify that hypothesis?
224.
by apsec112 2y ago
That's not the training code, just the inference code. The training code, running on thousands of high-end H100 servers, is surely much more complex. They also don't open-source the dataset, or the code they used for data scrapi
225.
by apsec112 2y ago
Trump supported a plan to reduce legal immigration by half: https://www.nytimes.com/2017/08/02/us/politics/trump-immigra...
226.
by apsec112 2y ago
Trump supported a plan to reduce legal immigration by half: https://www.nytimes.com/2017/08/02/us/politics/trump-immigra...
227.
10% of Cubans left Cuba between 2022 and 2023(miamiherald.com)
546 points by apsec112 2y ago|1127 comments
228.
by apsec112 2y ago
FWIW the version of Claude they use is quite old, v1.3. I'd be curious to see how the new Claude 3.5 performs on these
229.
by apsec112 2y ago
On GPU, or on CPU/Metal? For the latter I'm not surprised, but that's because they have a totally different memory/cache hierarchy.
230.
by apsec112 2y ago
1) Pretty much, it's mathematically equivalent. The only software issues are things like managing dependency versions and data formats in-memory, but Flash Attention 2 is already built into HuggingFace and other popular libraries. Flas
231.
by apsec112 2y ago
It's Hopper-specific, the improvements are closely tied to Hopper features like warp groups and TMA. For 4090s, you might get a speedup by using the Triton implementation of FP8 attention: https://triton-lang.org/main&#
232.
by apsec112 2y ago
"About 90% of the diverted immediate-release stimulants are used by a narrow demographic: white male college students" I'll eat my hat if this is actually true. US college students are about half white right now. The other ha
233.
Gacha Game(en.wikipedia.org)
1 points by apsec112 2y ago|0 comments
234.
Conspiracy Theorists Aren't Ignorant. They're Bad at Epistemology(benthams.substack.com)
7 points by apsec112 2y ago|1 comments
235.
Liveblog of that Situational Awareness piece(threadreaderapp.com)
1 points by apsec112 2y ago|0 comments
236.
200 people charged in $2.7B health care fraud crackdown(apnews.com)
425 points by apsec112 2y ago|225 comments
237.
Trans Youth (1931)(weimar-project.page)
3 points by apsec112 2y ago|0 comments
238.
Building a Pause Button for Biology(cradle.xyz)
9 points by apsec112 2y ago|3 comments
239.
by apsec112 2y ago
Excellent question. There was actually zero difference between the treatment group (given $1,000 a month) and the control group (given $50 a month). https://www.denverbasicincomeproject.org/research
240.
by apsec112 2y ago
There was no difference between the treatment group, and a control group given $50 per month. https://www.denverbasicincomeproject.org/research

More ›