LarsDu88

32 ms·

601.
by LarsDu88 2y ago
AI progress has been dramatic but the revenue doesn't even come close to the infra spend. Also companies like Meta have started to commoditize the models. Consumers could give less a shit about copilot enabled PCs. The actual revenue i
602.
by LarsDu88 2y ago
Edit I was mistaken: https://www.mandarinmansion.com/glossary/wuke-gong-wukegong These military exam bows went up to 160 lbs. In at least one instance a 240 lbs bow was manufactured which a single elite archer was able
603.
by LarsDu88 2y ago
From the same source above professor Mark Elliot writes: "In regularly held contests, expertness might be demonstrated by accuracy of aim, length of shot, or strength in drawing. It was the last of these that occasioned the most detail
604.
by LarsDu88 2y ago
Also the East India Company was founded in 1600, predating those English Navy exams by 1677. I'm not sure when they started adding standardized exams. Edit: It's possible these EIC exams were only introduced in the 19th century af
605.
by LarsDu88 2y ago
Here are the historical draw weights for Qing dynasty bows: https://www.manchuarchery.org/historical-draw-weights-qing-b... . 80 lbs was considered minimum for military usage and 150 lbs was not unheard of Imperial examinati
606.
by LarsDu88 2y ago
And the manchurian qing dynasty had a test where you had to be able to draw a 300 plus pound composite recurve bow! Which folks were still doing in the 20th century!
607.
by LarsDu88 2y ago
So for all ya'll software companies... one day it might be prudent to update these leetcode problems... With quizzes about tensors, Einstein summation, and recent scientific papers
608.
by LarsDu88 2y ago
Interestingly enough, standardized testing was unknown to Europe until they made contact with China. One company that was inspired by this was the British East India company which introduced an exam for employees. This contributed to the co
609.
by LarsDu88 2y ago
Isn't this a big pseudo religious and speculative? I mean... the statement that it's likely that there's a superintelligent cosmic host is not supported by any sort of evidence whatsoever? Like just the fact that the universe
610.
SwiGLU activation function causes instability in FP8 LLM training(arxiv.org)
10 points by LarsDu88 2y ago|2 comments
611.
by LarsDu88 2y ago
Recent pre-print from Intel Habana labs regarding training instability they encountered when training a Llama model at FP8, traced to the use of the SwiGLU from Noam Shazeer's 2020 paper: https://arxiv.org/abs/2002
612.
by LarsDu88 2y ago
This would not be good for either party. And Qualcomm would only do this to kill x86 prematurely which I'm sure AMD will pick up on.
613.
by LarsDu88 2y ago
Damn this opened my eyes to a whole 'nother world of online marketing
614.
by LarsDu88 2y ago
For all we know, we live inside some large scale invisible jet and our entire understand of physics is really the exception rather than the rule
615.
by LarsDu88 2y ago
If the plasma jet is wildly larger than our entire galaxy, I wonder if some sort of exotic life could evolve inside the jet. Some sort of life that would be totally rare in the universe.
616.
by LarsDu88 2y ago
Mathematically, the chance of a drug going from phase 2 to phase 3 is roughly 30% This can be modeled with a binomial distribution as 1-probability of both failing. That comes out to 1-(.7^2) which is 51% of at least one drug reaching phase
617.
by LarsDu88 2y ago
It's not broken. The board of directors is now a single person. It's like that scene in Star Wars where Vader tells everyone the Senate has been dissolved.
618.
Updated 23andMe Board of Directors Website(investors.23andme.com)
4 points by LarsDu88 2y ago|2 comments
619.
by LarsDu88 2y ago
They use sapling. An in-house clone of mercurial that was open sourced 2 years ago
620.
by LarsDu88 2y ago
It's the ONLY thing they could do. CEO Founder has a direct conflict of interest with shareholders...
621.
by LarsDu88 2y ago
Only the ones who cashed out on the private market before the company went public
622.
by LarsDu88 2y ago
Haha, correct
623.
by LarsDu88 2y ago
OP here. These headlines are simply exaggerated. 23andme sells mostly anonymous aggregated GWAS mapping results to companies like GSK as far as I'm aware
624.
by LarsDu88 2y ago
This is both dystopian and hilarious. 23andMe a subsidiary of Pizza Hut... We know exactly what you crave and it ain't anchovies.
625.
by LarsDu88 2y ago
Does each machine have just 1 gpu per rig or is it multiple? Do you network them to do DDP or is it just 1 inference job per card?
626.
by LarsDu88 2y ago
I had direct knowledge of this. To clarify I believe 23andMe did not give direct access to individual's DNA data. What 23andMe was selling to GSK was the results of GWAS (Genome Wide Association Study) results, which could be used to g
627.
by LarsDu88 2y ago
Funny enough, the entire 23andMe subreddit is purely about ancestry results, but the company marketing is oriented around health.
628.
by LarsDu88 2y ago
Nanopore sequencing fidelity was garbage back in 2021. Not sure what it is now.
629.
by LarsDu88 2y ago
At that burn rate it seems like the only recourse other than a buy out offer from another company at this point is to shutdown everything not related to cancer drug development and self-financing that endeavor.
630.
by LarsDu88 2y ago
Que sera, sera Whatever will be, will be The future's not ours to see Que sera, sera

More ›