Am Neumarkt 😱
254 subscribers
101 photos
3 videos
19 files
635 links
Machine learning and other gibberish
See also: https://sharing.leima.is
Notebooks: https://datumorphism.leima.is
Download Telegram
what the f
ECDC has been like this for at least 2h.
Am I time travelling back to 2000? AND this is about disease prevention and controll not useless blogs...
#neuroscience

Source:
https://science.sciencemag.org/content/370/6523/1410.full
A gatekeeper for learning

> Upon learning a hippocampus-dependent associative task, perirhinal inputs might act as a gate to modulate the excitability of apical dendrites and the impact of the feedback stream on layer 5 pyramidal neurons of the primary somatosensory cortex.

😲 In some sense, perirhinal inputs are like config files for learning.
#intelligence #paper #ML
Superintelligence Cannot be Contained: Lessons from Computability Theory
https://www.jair.org/index.php/jair/article/view/12202

> We argue that total containment is, in principle, impossible, due to fundamental limits inherent to computing itself. Assuming that a superintelligence will contain a program that includes all the programs that can be executed by a universal Turing machine on input potentially as complex as the state of the world, strict containment requires simulations of such a program, something theoretically (and practically) impossible.
Forwarded from The Sociologist
补充一部讨论互联网审查的工作生产、全球影响、社交媒体加剧冲突等议题的德国纪录片 The Cleaners(原名 Im Schatten der Netzwelt,网络阴影之下)。Hans Block 和 Moritz Riesewieck 执导,他们也在 TED 上 讲述了 关于「数字清洁」(digital cleaning)的问题。不过我第一次看这部纪录片,是在 DW 的 YouTube 频道,分为 上 、下 两集播出,目前均已失效,原因不明。一部讲述内容被删除的影片,自身却(或被)删除,不论是因著作权还是其他原因都表现出足够的讽刺。Internet Archive 上仍可找到 播出日 的 存档 回看,也可 在此 下载观看。
#productivity

I have been using Obsidian as my primary note-taking app for a while. It was a rough start. Linking notes was simply not in my workflow. In some sense, I am not familiar with my notes after a while. So I started to work on notes reviews every two weeks. On each notes review, I go through my notes inbox and spend some time connecting them with the the existing ones.

This is how my notes look like now. They are mostly well connected. (The cluster is because I have archived them as they are the notes for my previous position.)

I also borrowed the domain concept from dendron. I created folders with dot delimited domains. For example, I have this folder named inbox.ml which I use as my inbox for machine learning related notes. These notes will be distributed to a corresponding folder during my notes review.

Those notes worth publishing will then be distributed to my websites. For example, https://datumorphism.leima.is/ is for data science related notes.
https://www.chicagobooth.edu/why-booth/stories/in-memoriam-phd-student-yiran-fan

> A 30-year-old Ph.D. student in a joint program of Chicago Booth and the Kenneth C. Griffin Department of Economics, Fan was shot and killed on Jan. 9.

Related news article: https://www.globaltimes.cn/page/202101/1212449.shtml
> Fan was shot and killed in his car in the parking garage at an apartment building at about 1:50 pm Saturday. After shooting and killing Fan, the suspect, identified as 32-year-old Jason Nightengal by police, went on to shoot others across the city, reports said.
#ML

https://alan-turing-institute.github.io/skpro/introduction.html#a-motivating-example

Alan Turing Institute created a package called skpro for probabilistic modeling. Unlike many other probabilistic modeling packages, skpro integrates into sklearn pretty well.
#business
In 2015, there was a company called SixFold. They were one of the first heroes to disrupt an industry that has not changed much for a century, the freight market. They investigated the situation, established their hypothesis, created MVP. They did not succeed. The image is a summary of their post mortem.

There are at least two learning from this story.
- Think in terms of the utility function. Do not just point out blocks of reasons. Write down the utility function for the situation and make assumptions on the parameters.
- Swarm intelligence sometimes works better than one might expect. Improvements in swarm intelligence take a lot of effort if one does not have a smart plan.


Here is the article by their CEO: https://medium.com/@MartKelder/end-of-road-for-trucking-startup-palleter-523a4a906fe9