John_Maxwell

"I have a (weakly held) intuition that you tend to state your views with more confidence / certainty than is warranted" - an EA friend of mine. Hey man, qualifiers take up valuable space!

Comments

EA Cameroon - COVID-19 Awareness and Prevention in the Santa Division of Cameroon Project Proposal

I have no idea, I already shared my notes above! :) Perhaps the team could reach out to e.g. the author of the Johns Hopkins article?

BTW, I did find this article which argues for knitted masks:

https://stringking.com/face-masks/knit-vs-woven-fabric/

However, I'm more inclined to trust Johns Hopkins. But maybe the author of the Johns Hopkins article would have interesting opinions on the above link.

Edit: Here's more info

https://med.stanford.edu/news/all-news/2020/06/stanford-scientists-contribute-to-who-mask-guidelines.html

AMA or discuss my 80K podcast episode: Ben Garfinkel, FHI researcher

My guess would be that if you play with GPT-3, it can talk about as well about human values (or AI alignment for that matter) as it can talk about anything else. In that sense, it seems like stronger capabilities for GPT-3 also potentially help solve the alignment problem.

Edit: More discussion here:

https://www.lesswrong.com/posts/BnDF5kejzQLqd5cjH/alignment-as-a-bottleneck-to-usefulness-of-gpt-3?commentId=vcPdcRPWJe2kFi4Wn

EA Cameroon - COVID-19 Awareness and Prevention in the Santa Division of Cameroon Project Proposal

Since you requested feedback, here are some quick thoughts:

While I very much hope Cameroon is able to bring COVID under control, it seems like this could be difficult based on what we've seen in other countries. So the part of your plan that I'm most optimistic about is the mask making, because I think that could save lives even if COVID is not brought under control. Somewhere I read (can't remember where unfortunately) that if you wear a mask, then you'll end up inhaling a smaller number of viral particles if you get exposed to an infected person, and inhaling a smaller number of viral particles tends to give you a milder case, which means you're more likely to acquire immunity without putting your life at risk.

So I'd encourage you think about questions like: After we practice explaining the mask-making process in our workshops, can we find a way to explain mask-making via radio / flyers / newspaper articles? Or can we tell everyone at our mask-making workshops that they should run their own mask-making workshops for their family and friends and so on, so the mask-making knowledge spreads through the population that way?

Additionally, from what I've read some homemade masks are much more effective than others. Some snippets from my notes on mask effectiveness:

The test produced a few clear winners. While droplets from the average cough traveled around eight feet from an uncovered face, they went only 2.5 inches when produced behind a mask made of two layers of simple cotton quilting fabric. A mask made from a folded handkerchief produced droplets that traveled a bit over a foot. A loose, single-ply cotton bandana didn’t fare as well: While it prevented some fluid release, the cough’s plume still traveled nearly four feet.

“Whenever you have the option, use tightly woven fabric that has minimal leakage,” Verma concludes. “Any sort of covering is better than none.”

https://www.nationalgeographic.com/science/2020/07/how-to-make-coronavirus-masks-that-everyone-will-want-to-wear-cvd/

Thicker, more densely woven cotton fabrics are best, such as quilting cotton or cotton sheets. Stretchy knits aren’t ideal. Hold the fabric up to the light: The fewer tiny holes you can see, the better it will work to filter droplets.

https://www.hopkinsmedicine.org/health/conditions-and-diseases/coronavirus/coronavirus-face-masks-what-you-need-to-know

The fabric should be a woven fabric, not a knitted fabric. What’s the difference? Woven fabrics don’t stretch much, so when you tie it around your face, the tiny holes between the threads don’t get bigger and let in more viruses.

https://www.sleepphones.com/Comfortable-Coronavirus-Face-Mask-Beard-Glasses-Fogging

In recent tests, HEPA furnace filters scored well, as did vacuum cleaner bags, layers of 600-count pillowcases and fabric similar to flannel pajamas. Stacked coffee filters had medium scores. Scarves and bandanna material had the lowest scores, but still captured a small percentage of particles.

If you don’t have any of the materials that were tested, a simple light test can help you decide whether a fabric is a good candidate for a mask.

“Hold it up to a bright light,” said Dr. Scott Segal, chairman of anesthesiology at Wake Forest Baptist Health who recently studied homemade masks. “If light passes really easily through the fibers and you can almost see the fibers, it’s not a good fabric. If it’s a denser weave of thicker material and light doesn’t pass through it as much, that’s the material you want to use.”

...

Dr. Wang’s group tested two types of air filters. An allergy-reduction HVAC filter worked the best, capturing 89 percent of particles with one layer and 94 percent with two layers. A furnace filter captured 75 percent with two layers, but required six layers to achieve 95 percent. To find a filter similar to those tested, look for a minimum efficiency reporting value (MERV) rating of 12 or higher or a microparticle performance rating of 1900 or higher.

The problem with air filters is that they potentially could shed small fibers that would be risky to inhale. So if you want to use a filter, you need to sandwich the filter between two layers of cotton fabric. Dr. Wang said one of his grad students made his own mask by following the instructions in the C.D.C. video, but adding several layers of filter material inside a bandanna.

Dr. Wang’s group also found that when certain common fabrics were used, two layers offered far less protection than four layers. A 600 thread count pillow case captured just 22 percent of particles when doubled, but four layers captured nearly 60 percent. A thick woolen yarn scarf filtered 21 percent of particles in two layers, and 48.8 percent in four layers. A 100 percent cotton bandanna did the worst, capturing only 18.2 percent when doubled, and just 19.5 percent in four layers.

...

The best-performing designs were a mask constructed of two layers of high-quality, heavyweight “quilter’s cotton,” a two-layer mask made with thick batik fabric, and a double-layer mask with an inner layer of flannel and outer layer of cotton.

...

Bonnie Browning, executive show director for the American Quilter’s Society, said that quilters prefer tightly woven cottons and batik fabrics that stand up over time. Ms. Browning said most sewing machines can handle only two layers of fabric when making a pleated mask, but someone who wanted four layers of protection could wear two masks at a time.

https://www.nytimes.com/article/coronavirus-homemade-mask-material-DIY-face-mask-ppe.html

...adding a layer of nylon stocking over the masks minimized the flow of air around the edges of the masks and improved particle filtration efficiency for all masks, including all commercial products tested. Use of a nylon stocking overlayer brought the particle filtration efficiency for five of the ten fabric masks above the 3M surgical mask baseline...

https://www.medrxiv.org/content/10.1101/2020.04.17.20069567v2.full.pdf

So it's probably worth doing some research to figure out the best mask design from the perspective of effectiveness, ease of explaining how to create, and likelihood that people in the Santa Division will be able to acquire the necessary materials, if you haven't already done this.

Concern, and hope

Something I've been doing just a bit lately which seems to be working surprisingly well so far: If I see a polarizing discussion on EA Facebook, and someone writes a comment in a way which seems needlessly combative/confrontational to me, I add them as a friend and private message them trying to persuade them to rewrite their comment.

My general model here is that private 1-on-1 communication is much higher bandwidth, less ego-driven, and more amenable to the resolution of misunderstandings etc. However it's not nearly as scalable (in terms of the size of the audience reached) as a forum discussion is. But private 1-on-1 communication where you try to persuade someone to change their forum writing gets you the best of both worlds.

Another model is that combativeness tends to beget combativeness, so it's high-leverage to try & change the tone of the conversation as early as possible.

The 80,000 Hours podcast should host debates

Another way to accomplish something similar would be to post the podcasts to the EA Forum and have this be the official place for people to comment on them.

EA Cameroon - COVID-19 Awareness and Prevention in the Santa Division of Cameroon Project Proposal

This looks like a great initiative!

I noticed that based on the link in your EA Forum profile, you're located in the Czech Republic--would you mind talking a bit about your relationship with EA Cameroon?

EA Survey 2019 Series: How many people are there in the EA community?

Coming together weekly to meet in person as a movement, like megachurches do, is an interesting thought experiment. Post-COVID, if remote work is the new norm, it might be feasible to locate all of EA in a single city with low cost of living. Would this be a positive change? My intuition says yes, but with high uncertainty. Maybe it's just me being extroverted.

Request for proposal - EA Animal Welfare Fund

Here's a project idea that someone might want to take on.

It's been argued (response) that bivalves have much reduced capacity for suffering compared to other commonly eaten animals. And according to this recent NY times article, "Mollusks like clams, oysters and scallops are also great low-carbon choices."

What could get people to substitute with bivalves? Lowering the price should help. A quick search on Amazon suggests that canned mussel (one of the bivalves highest in Omega-3s according to this chart) is 2-3x as expensive per ounce as canned salmon.

How could the price be reduced? On this US government website, you can see a picture of a guy culling and grading oysters by hand. Google has a case study on its website of a Japanese cucumber farm which used deep learning to sort cucumbers. Could similar technology be developed for bivalves? As a bonus, you could develop AI skills along the way, and potentially make a decent amount of money. It might be best to partner with/be hired by existing efforts in the bivalve automation space... here is one I found on Google.

Tips for overcoming low back pain

Painscience.com has been amazing for all my chronic pain problems. Looks like he has a comprehensive guide for low back pain which might be worth checking out:

https://www.painscience.com/tutorials/low-back-pain.php

Load More