andyljones

8 ms·

31.
by andyljones 6y ago
There's no shortage of cities where central space has become cheaper after their industry has collapsed, and they do not have a great record of turning vibrant. The rust belt and the north of England are the first areas to mind.
32.
by andyljones 6y ago
For anyone else just looking for an outline of the new algorithm: last two paragraphs of p4, first two paragraphs of p5. Coming from a position of a few abstract algebra classes in college many years ago, all the words and notation are fami
33.
Show HN: Megastep, 1M FPS reinforcement learning on a single GPU(andyljones.com)
4 points by andyljones 6y ago|0 comments
34.
by andyljones 6y ago
More generally, the best commentary I've seen on GPT-3 comes from following @gwern and keeping an eye out for whoever else they frequently interact with. https://twitter.com/gwern/with_replies Most other commentar
35.
by andyljones 6y ago
Nic Cammarata on Twitter pointed out that if the prompt gives GPT-3 permission to indicate questions are ridiculous, it'll do so reliably: https://twitter.com/nicklovescode/status/1284050958977130497
36.
by andyljones 6y ago
The link at the bottom of the post to the actual explanation is broken. Here's a web archive version: https://web.archive.org/web/20191106104440/http://cod.ifies.... TL;DR the default Windows random
37.
by andyljones 6y ago
Concrete numbers from the various pullouts: > They saw ratings hover around 60% with their original, in-house tech — this improved by 7-8% with GPT-2 — and is now in the 80-90% range with the API. > The F1 score of its crisis classifi
38.
Cross-Examination for Journalists(crime.scot)
2 points by andyljones 6y ago|0 comments
39.
A Clearer Proof of the Policy Gradient Theorem(andyljones.com)
1 points by andyljones 6y ago|0 comments
40.
by andyljones 6y ago
The reason to use Linux is to follow the crowd. I use Nvidia's docker container because the plurality of devs use it. Over the course of my career I've found that well-trodden paths tend to have a _lot_ fewer bugs thanks to other
41.
by andyljones 6y ago
Here you go: https://cloud.google.com/support#support-plans $250/month/dev is the minimal for phone calls on technical issues, $150k + 4% of GCP spend for 'come running' support. There're more detai
42.
by andyljones 6y ago
SciHub mirror of the paper: https://sci-hub.tw/https://www.sciencedirect.com/science/art... It's a 'brief communication' of three pages, so not many details. Most interesting thing is the
43.
by andyljones 6y ago
Short answer: yes, the underlying black-box fast multipole method will be very accurate with just a small number of Chebyshev coefficients. The main requirement is the kernel is reasonably smooth away from the origin. Here's a toy exam
44.
by andyljones 6y ago
I sent some emails and didn't hear anything back. I'm hoping this post'll get organic attention for it, and I intend on sending more emails and - now you've prompted me! - opening an issue on the tracker too.
45.
by andyljones 6y ago
Oh rats. Thanks for the heads up. I don't have an iPhone and it's not happening on macOS Safari. I'll have a Google around and see if any of the underlying libraries are likely culprits. Which animation resetting seems to tri
46.
by andyljones 6y ago
Funnily enough, this entire project sprung out of my trying to replicate the Imperial model a month ago. I found the slow part was the community transmission kernel, and was surprised to find no epidemiological literature on applying fast m
47.
by andyljones 6y ago
That's a good idea - I'll have a think about how best to implement it. Might fall under the general category of 'add a speed control'.
48.
by andyljones 6y ago
Yup! You can see the animations and their code in isolation here[1] - just click next to the animation to expand it. Fair warning, this is code I expected to write once and read never, and it shows. [1] https://observablehq.com&#
49.
Multipole Methods for the Masses(andyljones.com)
82 points by andyljones 6y ago|28 comments
50.
by andyljones 6y ago
The two big ones in machine learning at least are Google's Gin: https://github.com/google/gin-config and Facebook's Hydra: https://hydra.cc/
51.
The Good Judgement Project's superforecasters expect 250m cases by next April(andyljones.com)
2 points by andyljones 7y ago|0 comments
52.
by andyljones 7y ago
I don't agree with that. There's a spectrum between information being entirely uninformative and perfectly informative. Daily case headlines are pretty close to uninformative on their own, but the underlying case figures, in bulk,
53.
Headlines about coronavirus case counts are worthless(andyljones.com)
7 points by andyljones 7y ago|3 comments
54.
Unlocking Braden's Potential [pdf](demonetized.files.wordpress.com)
1 points by andyljones 7y ago|0 comments
55.
by andyljones 7y ago
Lord, I'm not really qualified; I've only been baking for a few weeks. My justification for writing this was more that I thought I could put it in a style that a new audience would be receptive to, rather than any perceived expert
56.
by andyljones 7y ago
Thanks very much! Yeah, I realised after Loaf #1 my folding technique needed to improve, and being a bit more energetic with the folds has given much better results (today's loaf! https://i.imgur.com/2WQYvcT.jpg ). I&#x
57.
by andyljones 7y ago
It's too small to use a normal gram-accurate kitchen scale; received wisdom is that one quarter-teaspoon of instant yeast is about a gram. So a third of a gram is a twelfth of a teaspoon. Thanks for a prompt - I've now added a not
58.
Quarantine Bread(andyljones.com)
209 points by andyljones 7y ago|150 comments
59.
Coronavirus Contrarianism: The British strategy // Maybe I will go outside(andyljones.com)
4 points by andyljones 7y ago|1 comments
60.
Post-Mortem Python Plotting(andyljones.com)
4 points by andyljones 7y ago|0 comments

More ›