pyentropy

You don't need a shared storage if you do SSO the right way. Just redirect with a session token.

And if the SSO server wants to share state with the other services, like forcing a logout on a specific app, it can always do it on the server side (app backend <-> SSO backend with user token).

Just do a redirect with session as a query parameter (like many OAuth apps do it anyway). Each site can then persist the user session for more than a month.

There is very little evidence GeneSight results mean anything for efficacy. Tolerability and dose metabolism in the best case.

Maybe I'm bitter and biased because I like SSC, but based on the author's social media profiles and his new book description, he's purely jealous that he wasn't allowed to be the man to "expose" SSC and how LessWrong and SSC played role in some events in his upcoming AI book "Genius Makers" (jeez what a name...).

That's exactly the point. The selective classes (SNRI/SSRI) are marketed as a complete and safe replacement for MAOIs, on top of the chemical structure irony. Even the new atypicals with innovative chemical structure (and targets other than 5-HT1/SERT) are of significantly worse efficacy MAOIs

I'll edit my parent comment to make it clearer.

Hey, I wrote about monoamine oxidase inhibitors on Scott Aaronson's blog a few months ago: https://www.scottaaronson.com/blog/?p=4933

Basically, there's is an international shortage for one of the most effective treatment for severe depression because of an unfortunate series of events, the drug's long history as a generic and the manufacturers' willingness to replace the drug class, and constantly produce new patented molecules with chemical tricks(ex. just taking one isomer in the citalopram->escitalopram case, or even metabolites in the case of venlafaxine -> desvenlafaxine case) like they are 'annual smartphone model releases', except efficacy even goes down.

You can read more about it the drug itself on the blog of Scott Alexander (the rationalist/Bayesian psychiatrist guy) as well: https://slatestarcodex.com/2015/04/30/prescriptions-paradoxe... And now some years down the line after his observation, the price has exploded, very few new patients get a prescription due to uncertainty and there's no alternative. Your options are either taking taking an atypical antidepressant like vortioxetine or an experimental ketamine.

My project builds 20 binary wheels supporting most platforms except ARM (todo): https://pypi.org/project/elkai/#files It's one line of code using CMake. [1]

Granted, you also need CI that has most Python versions/platforms and the Python CMake extension by scikit team too. Ex. MacOS Travis with brew [2]

[1] https://github.com/filipArena/elkai/blob/ef8ee98d8114240b204...

[2] https://github.com/filipArena/elkai/blob/master/.travis.yml

Lex is bringing CS nerdiness to mainstream culture by mixing poetry/music/fitness/philosophy and I don't get why are you so bitter about it.

Yes, he didn't ask the nitty gritty questions we'd like to hear from Donald Knuth or Jim Keller or dozens of quantum computing, computational biology etc researchers he has interviewed. But if you read their blogs, that's what they like about their appearance on the podcast.

The content the HN bubble is interested in would be understood by ~5-10k people but he has 870k subscribers. If you want research-level interviews, there's https://www.quantamagazine.org/, there are research blogs, journal commentary articles, etc.

My instinct was to close the site immediately but decided to read what Woz is up to.

Navigation didn't work for some stuff on mogile but I dowloaded the PDF whitepaper and I must say this is more shallow than most "tokens".

If I understand correctly, people invest money into a pool that is distributed for companies/industries to spend, in order to increase their energy efficiency. Once they are done ("once" should be "if"), part of their savings goes to the investors. I think it is reasonable to assume that such process doesn't need any crypto proof and ultimately depends on lawyers, salesmen and bankers. Making a company or an industry more efficient requires talent and strategy, not just pouring money and it is a very risky and long process. If these guys feel competent to make a profitable energy efficiency fund, they should start one. And with clear legal protection for the investors, not a token.

The title was confusing to me. This has nothing to do with sharing a core by context switching i.e. giving multiple programs a timeslot on the same core.

It's about HyperThreading/SMT, a feature that allows multiple programs to run in the same instruction cycle by using idle execution units within a shared core.

Will there be a warning if you try to use 'master' as a refname if it doesn't exist but 'main' does? I assume beginners using educational resources that aren't up to date with this change will be really confused.

For readers unfamiliar with the project: it's been in alpha for 24 years, although it now targets Server 2003/XP instead of Win95 which it originally did in 1996.

I'm really angry at Pfizer and co for ditching MAOIs for SSRIs. All research shows that they are much more effective, have no emotional blunting effect and have less impact on libido, at the cost of not being allowed to eat cheese and some insomnia. SlateStarCodex wrote about it in his prescription perversity post. [1]

[1] https://slatestarcodex.com/2015/04/30/prescriptions-paradoxe...

Do you mean container margins? The text is left justified and I don't see any difference between that site and HN.

The article says adaptation can be bypassed.

I think all emotional states can be triggered electrochemichally for arbitrary durations.

The reason all humans aren't like the manic people mentioned in the article is that it's not an evolutionary stable strategy. They hoard, steal, flirt too much and cooperation cannot happen.

Which brings me to a scary question - what stops a future radical extremist from capturing enemies and triggering the most extreme pain that no one has ever felt before with electrodes? Or gene editing millions of babies to be in constant extreme pain for their whole lifetime, without limbs or senses to be able to change anything (inspired by Metallica - One)?

How was Facemash (Hot or Not) developed (pics of Harvard girls never used the site)? How did they get website traffic data from their like button without visitors using Facebook? Also, how is my grandma part of their database because all her grandchildren have her in their phone contacts?

IMHO, I'm glad Microsoft is not part of this. It's not about profits, it's about data monopoly.

Writing a modern OS is almost impossible from scratch, but there's nothing preventing you from hiring enough people and making Windows or Azure.

But the problem with Facebook and Google is that they started by scraping stuff for free and without limitations, but if you try to do the same thing with them they bite you with Terms of service, paid data graphs and API quotas. How do you build a index of businesses/venues and people? No matter how many people you hire you cannot achieve the same snowball / network effect.

Instead of breaking them up, make them release open data.

When Android started getting big, a huge community of enthusiasts were making cool proof of concepts. Port busybox, VNC, run Ubuntu on your phone, ROM customization.

Beginners started using accelerometers and multitouch to make innovative games. I remember tutorials on how to make algorithms for good gesture detection (gestures were a big thing).

Now app stores are worth almost trillion dollars. You can still find niche communities but just like with the Internet, it got too commercialized.

Marketing dollars work, and bidding for ad placement is competitive. The field is made of clever people that optimize it. However I think it's dangerous to believe that advertising is "numerical" or metrics can fully predict ROI of advertising - the same way subprime mortgages weren't numerical, and the same way a population mortality graph cannot predict coronavirus.

A sudden drop in revenue may adjust the economy of bidding and show a potential bubble. The tiny budgets and one click campaigns (with platforms inventing unique metrics & targeting woo) in social media marketing compared to traditional marketing (huge barrier to entry) feel like penny stocks vs traditional stocks. You can easily mislead individuals and tiny businesses.

Suppose your company is developing a framework for creating mobile apps. You work on optimizing the GUI renderer and you are assigned a task to make a module which finds computationally equivalent render functions (or a company making a programming assignment plagiarism system, or a duplicate config file detector when the config language is accidentally Turing-Complete).

You may try removing whitespace, finding invariants, ordering the variables by type, converting for loops to while, tail call optimization, but your collegue always has a trick to make equivalent functions undetectable by your module. You were solving the halting problem. Instead, you should use CS knowledge and tell them it's impossible but they can have a non-optimal search. Perhaps, the fastest way to complete the task would be to score similarities between the functions' machine code or AST.

AlphaZero has coded all the rules for the respective three games, they do a tree search and their neural network output layer has exactly n neurons for max(n) possible moves. Although it's impressive they don't teach it heuristics and strategies, it's a very specific task.

What about pigeons predicting breast cancer with 99% probability, rats learning to drive cars, monkeys building tools?

Rodents stand a bigger chance at learning Go than AlphaZero spontaneously building stone tools and driving cars.

Progress in AI is due to data and computational power advances. I wonder what kind of advances are needed for AGI.

1. Biological brains are non-differentiable spiking networks much more complicated than backpropagated ANNs.

2. Ion channels may or may not be affected by quantum effects.

3. The search space is huge (but organisms aren't optimal and natural selection is probably local search)

4. If it took ~3.8b years to get from cells to humans, how do we fast-forward:

* brain mapping (replicating the biological "architecture")

* gene editing on animal models to build tissues and/or brains that can be interfaced (and if such interface could exist how do we prevent someone from trying to use human slaves as computers? Using which tissues for computation is torture?)

* simulation with computational models outside of ECT (quantum computers or some new physics phenomenon)

Note: those 3.8b years are from a cell to human. We haven't built anything remotely similar to a cell. And I'm not claiming that an AGI system will need cells or spiking nets, most likely a lot of those are redundant. But the entropy and complexity of biological systems is huge and even rodents can outperform state of the art models at general tasks.

IMHO, the quickest path to AGI would be to focus on climate change and making academia more appealing.

I think his BosonSampling method comes very close to Shor's algorithm, and I don't think "leagues" really exist in academic output - but he's not doing good work - he's doing great work.

Scott Aaronson is probably one of the most underrated scientists of our time. He's very approachable and has responded to comments and emails of undergrads like me in a few hours. Not only is he part of the team that practically invented quantum supremacy, but his blog is the most informative blog out there for ethics, morality, nerdiness, theoretical computer science and the philosophical implications of complexity theory.

Y Combinator had a podcast with him which you can watch here: https://www.youtube.com/watch?v=0jrybODBUpA

So apparently Google rolled out an update where they 'm.domain.com' turn into 'domain.com' in the omnibox. In what world is this acceptable? How can they assume that 'm.' always means mobile and 'www.' is the same as the root domain for all hosts?

Let's assume that you have a blog platform offering subdomains for each user and 'm.blogplatform.com' is available. Now, any user can get that subdomain and impersonate the homepage because Emily from Chromium decided that eliding parts of the URL without any spec is a reasonable decision.

The TSP solver is vanilla C and the Python extension that wraps it only needs "Python.h".

I use scikit-build which is a setup.py alternative and builds and bundles together the solver and the extension. The build system makes a binary version for most platforms.

Correct, but you will have to find/build the shared library. Also, as soon as the build systems support 3.8, I will add 3.8 wheels.

I tried to support all 3.xx platforms that numpy supports: https://pypi.org/project/numpy/#files