May 2020 (Links)

The Importance of Saying “Oops”

Source: LessWrong | By: Eliezer Yudkowsky

“Not every change is an improvement, but every improvement is necessarily a change. If we only admit small local errors, we will only make small local changes. The motivation for a big change comes from acknowledging a big mistake.”

Don’t let small adjustments take you in the wrong direction, correct often and sometimes make big corrections.

Pick of the coronavirus papers: Immune system shows abnormal response to COVID-19

Source: Nature (the paper/publisher)

Probably best place to look for news on research surrounding COVID-19.

Lee Kuan Yew’s Rule

Source: Farnham Street

“Does it work?” That is the question that the modernizer of Singapore asked himself.

“I do not work on a theory. Instead, I ask: what will make this work? If, after a series of solutions, I find that a certain approach worked, then I try to find out what was the principle behind the solution. So Plato, Aristotle, Socrates, I am not guided by them…I am interested in what works…”

How uniform is the neocortex?

Source: LessWrong | By: zhukeepa

“The neocortex is the part of the human brain responsible for higher-order functions like sensory perception, cognition, and language, and has been hypothesized to be uniformly composed of general-purpose data-processing modules.”

Based on the current knowledge about the brain, and AI research, the author (and I too) think the idea that the brain is using prediction (and input) and is quite uniform, as quite likely.

Interview with Aubrey de Grey

Source: LessWrong | By: emanuele ascani

Some interesting questions, but not much I didn’t know. We need to solve ageing, and for that to happen need to solve all the aspects/problems.

AI and Efficiency

Source: Open AI

“We’re releasing an analysis showing that since 2012 the amount of compute needed to train a neural net to the same performance on ImageNet1 classification has been decreasing by a factor of 2 every 16 months. Compared to 2012, it now takes 44 times less compute to train a neural network to the level of AlexNet2 (by contrast, Moore’s Law3 would yield an 11x cost improvement over this period). Our results suggest that for AI tasks with high levels of recent investment, algorithmic progress has yielded more gains than classical hardware efficiency.”

Dithering and Open Versus Free

Source: Stratechery | By: Ben Thompson

Open doesn’t equal free exactly, email is open but you can have it be paid (as he does). Subscription are important to have a steady income (vs hoping for hits).

The rest discusses how these dynamics play out and how Spotify is aggregating podcasts (also discussed earlier).

“…Spotify still is not open: they can take down your content or choose not to play it, just as Facebook could not show your page unless you were willing to pay-to-play.”

“…That, by extension, means not agreeing to Spotify’s terms for Exponent, and accepting that leveraging RSS to have per-subscriber feeds makes having the Daily Update Podcast on Spotify literally impossible. More broadly, owning my own destiny as a publisher means avoiding Aggregators and connecting directly with customers.”

Studies on Slack

Source: Slate Star Codex

A great piece on slack (in the system) and how evolution and other systems can be explained by this.

Literature Review For Academic Outsiders: What, How, and Why

Source: LessWrong | By: namespace

Good article on how to do a literature review. Could be useful later to see if I can apply some of these things to my review of psychedelics literature.

Chips and Geopolitics

Source: Stratechery | By: Ben Thompson

When the technology isn’t good enough, you want to be the best (features), when it’s better than needed (e.g. phones now), you want to be the most flexible/quickest.

The rest of the article explores this and TSMC’s geopolitical issues.

The EMH Aten’t Dead

Source: LessWrong | By: Richard Meadows

Interesting article that challenges the view that the efficient market hypothesis (the EMH in the title) isn’t dead. Good arguments and analysis of how it’s working in this Corona-time.

Did actually expect more about government-backing of large companies and some discussion about the listed companies (some of them, tech) not really being affected that much by the situation (read: less personnel that is able to work from home).

OpenAI Finds Machine Learning Efficiency Is Outpacing Moore’s Law

Source: SingularityHub | By: Jason Dorrier

Already read this somewhere else, but interesting to read again. The efficiency of AI is going faster than Moore’s Law. This means both that the big ones can get better, and that we can incorporate AI in smaller things (you know, like the size of our brains). Still not really attacking the correlation vs causation arguments, but still very good.

And they built the 5th most powerful computer on Earth.

My Ordinary Life: Improvements Since the 1990s

Source: Gwern

Really cool blog about what has improved in our lives, focus on the daily/small things like VHS tapes, fast internet, group chats (vs telephoning each friend separately, etc).

Fun to read and makes me think about those small things that are now better, e.g. always hot water when you shower (vs only having X minutes, for the whole family).

Post-Prozac Nation

Source: NYT magazine | By: Siddhartha Mukherjee

“The writer Andrew Solomon once evocatively described depression as a “flaw in love” — and certainly, the doctors using Raudixin at Duke had seen that flaw emerge grimly in real time: flaws in self-love (guilt, shame, suicidal thoughts), love for others (blame, aggression, accusation), even the extinction of a desire for love (lethargy, withdrawal, dullness). But these were merely the outer symptoms of a deeper failure of neurotransmitters. The “flaw in love” was a flaw in chemicals.”

About the love, and falling out of love, story of antidepressants. And our understanding of the brain and how it functions.

Studies on serotonin levels showed that higher or lower levels were both found (in depressed populations) and a chemical intervention (lowering serotonin concentration) didn’t affect mood. But maybe the effect is only so in the (chance-of-being) depressed population.

Yet, 75-82% of the effect (depending on the studies included, see article), of antidepressants could be caused by the placebo effect. “But for patients with the most severe forms of depression, the benefit of medications over placebo was substantial. Such patients might have found, as Andrew Solomon did, that they no longer felt “the self slipping out” of their hands. The most severe dips in mood were gradually blunted.”

“Prozac’s positive effects, in other words, depended on the birth of nerve cells in the hippocampi of these mice.” So looking not at serotonin, but the growth of new brain cells.

“Might depression also be a degenerative disease — an Alzheimer’s of emotion, a dementia of mood?”

“If an answer to these questions exists, it may emerge from the work of Helen Mayberg” … “Tracing such sites led her to the subcallosal cingulate, a minuscule bundle of nerve cells that sit near the hippocampus and function as a conduit between the parts of the brain that control conscious thinking and the parts that control emotion.

“A remarkable and novel theory for depression emerges from these studies. Perhaps some forms of depression occur when a stimulus — genetics, environment or stress — causes the death of nerve cells in the hippocampus.”

Very interesting, something I want to learn more about. Start at Helen Mayberg?

Platforms in an Aggregator World

Source: Stratechery | By: Ben Thompson

Interesting (again) analysis of Shopify (that is shop, not spot) and how it competes with other companies (mainly Amazon and Facebook) and the choices they have made. At Queal, we’re on Woocommerce (the open-source version) but I do see where they are at and what they are trying to do.

Why Remote Work Is So Hard—and How It Can Be Fixed

Source: The New Yorker | By: Cal Newport (from Deep Work)

An in-depth analysis of the history of remote work and why it’s hard and what we can do about it.

It’s hard because there is less face-to-face (spontaneous meetings that Apple designed into its new mega donut, uhh office). This relates to both tasks/meetings, and social cues/praise. It’s also hard to do long focussed work at home (you need to separate yourself/your space).

“In many offices, tasks are assigned haphazardly, and there are few systematic ways to track who is working on what or find out how the work is going. In such a chaotic work environment, there are profound advantages to gathering people together in one place.”

Newport makes the analogy with adoption earlier technologies and argues that we get stuck because we try and adapt the new one in the old system. With the current pandemic, we might have such a large shock that we can start fresh.

Personal productivity tips are given (see the book linked above). Blocking time and having office hours (and happy hours?) are also valuable.

Good article overall, and we will see where things take us. Finally also a mention of the ‘tour of duty’ idea from The Alliance.

Effective Altruism and Meaning in Life

Source: EA Forum | By: anon

Good musings on EA and the meaning of life (well, that’s the title). I took away from it that EA isn’t the whole purpose of your life. There is more to it (e.g. love) and more to your career (doing other things).

Note: from June 2020 onward, I’ve been adding interesting links to an Obsidian file, may start this again later if still necessary/useful.