erostrate

If you try to buy base jumping gear online, some manufacturers' websites will now check who your instructors are and how many jumps you have. That's because people have been known to just buy gear and go base jumping, with no experience or instruction.

It's worse than that. That spreadsheet was created after FTX became insolvent. When Sequoia invested, they probably didn't even have something as good as that terrible spreadsheet.

Or more charitably, Sequoia presented with something more professional looking, but completely incorrect.

Of course now that he's aware that the mistake was on his side, he has apologised and amended his Twitter posts. Because he wants his followers to get accurate information that allows them to make informed decisions based on rational thinking, not emotional outrage or base tribal instincts exacerbated by misleading information propagated by echo chambers. Because he understands that preserving a well functioning democracy and dialog is more important than advancing his own interests. Of course.

So in that theory they do get sufficient USD to cover USDT (as opposed to printing it out of thin air) but are using it to buy questionable CP.

I think many people believe they don't even have enough USD to start with, eg because some of their assets are loans collateralised by crypto (such as the Celsius 1B loan).

I believe both are true - Tether printed USDT against non-cash, and used some of the actual cash to buy dodgy assets for more yield.

Do you know by what mechanism Tether gets CP for USDT? Would they be paying CP issuers with USDT directly? What would they do with it?

I believe you're right just curious about the mechanism.

It's less direct than that. Some crypto fund has say 10M$, they want to buy a lot of BTC, so they borrow 20M$ denominated in USD but settled in USDT from iFinex treasury, and use it to buy BTC. iFinex can now claim that they are backed by USD assets, crypto fund effectively has taken a leveraged BTC long position, pushing BTC up, everybody is happy. Until BTC goes down too much or everybody tries to withdraw USDT, or both, then everything unravels.

You can also replace "crypto fund" by any crypto project that uses leverage to deliver extraordinary returns - as long as everything goes up.

Some points I found interesting:

- Kristian Andersen offered to abuse his power as editor to stealthily edit a preprint. He is the main author of the most notable paper supporting a natural origin, and has been one of the major resources of the natural origins camp. Now his scientific integrity is seriously questioned.

- Peter Daszak appears much more focused on finding new sources of grant money than on doing good science: "What was needed, he exhorted his staff, was a “change in culture” as “part of [a] mentaility [sic] to get money,”

- The Wuhan Institute of Virology is more of a second zone lab than a top lab: "The WIV was also viewed as subpar, especially when compared with the Harbin Veterinary Research Institute. Harbin was China’s Harvard, said the former DARPA official. The WIV was more like a safety school"

- EcoHealth Alliance and Daszak were seen by DARPA as amateurs in virology, not at all the expert frontline organisation that's been suggested before: "EcoHealth Alliance was viewed as a “ragtag group” and a “middle guy,” a backseat collaborator willing to get on an Air China jet, eat terrible food, and stay in bad hotels" "EcoHealth Alliance had “bolted on” a serious scientist, Ralph Baric".

A lot of effort went into these efforts to understand neural networks in terms or kernels or SVMs. To the best of my knowledge, these efforts have not inspired useful new architectures, nor have they made useful experimental predictions on real neural nets, nor have they had any significant impact on important machine learning benchmarks.

I think some researchers are refusing to accept the idea that machine learning is very much an experimental science today, and the (very cool) mathematics of kernels, SVMs, empirical risk minimization, bayesian statistics, etc. are simply no longer useful in the large scale regime.

My prediction is that there will be a useful "deep learning theory" in the future, but it will look a lot more like physics (such as Kaplan's scaling laws) than early 21st century machine learning mathematics/statistics.

Thanks, I like Levine and remember reading that. He basically argues that airlines financial strategy was optimal for shareholders, given covid and guaranteed government support. But firstly, he only considers the two uses of money proposed in the NYT, improving customer service or reducing the debt burden. Secondly, his whole argument is predicated on airlines being bailed out by government - which is true, but I believe shouldn't. I personally believe a lot of value is being destroyed by lack of long term investment and short term incentives, leading to problems such as these ones. And finally, even Levine admits that buybacks might be suboptimal for other stakeholders (eg employees, clients).

Some financial facts that help explain the situation: over the past 10 years US airlines have spent about 95% of their free cash flows on stock buybacks. This is the money they could have used to have a buffer to deal with these situations or to improve their systems. Why do they care so much about their stock price over long term sustainability? In 2020, Southwest lost 3.1 billion dollars and took 6 billions of government money, but executives increased their own compensation by 5% to 14% and the CEO paid himself 9 millions dollars, of which 7 millions were in stock.

https://www.washingtonpost.com/business/2020/04/06/bailout-c...

https://www.dallasnews.com/business/airlines/2021/04/09/pay-...

The articles ends with a suggestion that Mark Zuckerberg should be fired from his CEO role, but he controls 58% of the voting shares, so I'm not sure who this appeal is directed at.

TFA spends a long time belaboring the point that this is not correct. Because many people who click on your URL would have bought the item anyway.

How does your theory explain Rust or Python?

"Crypto AG was a Swiss company specialising in communications and information security. It was secretly jointly owned by the American Central Intelligence Agency (CIA) and West German Federal Intelligence Service (BND) from 1970 until about 1993, with CIA continuing as sole owner until about 2018."

https://en.wikipedia.org/wiki/Crypto_AG

I checked out of curiosity and the SuperMicro article was written by the same author, Jordan Robertson. Make of that what you will.

https://www.bloomberg.com/news/features/2018-10-04/the-big-h...

The cord is designed to get rid of an incorrectly opened main parachute. Not an unopened one. This is to allow safe deployment of the reserve parachute.

On modern parachutes it won't actually get rid of an unopened main parachute. A main parachute that refuses to open at all is actually scarier than a poorly opened one, because it could get open anytime. Including after you have opened the reserve one. Then you have two parachutes open, and they can get into surprising unstable aerodynamic configurations (a downplane). And you can't easily get rid of an unopened main parachute.

Source: I learned skydiving this summer

Do you have a source for that?

Estonia is ranked 5th in reading, 8th in maths, and 4th in science, despite having a budget per student lower than OECD average! It's the first non Asian country. What's happening there?

The most positive thing about this leak is that it includes government officials. The more they realize that the mass surveillance they are pushing on us will also apply to them, the more chances we'll have that they push back against it.

Combine that with the sick lab researchers, with the work performed at BSL-2 instead of BSL-4, with the main researcher being a world leading expert on gain of function research, with a WIV database of genomes surprisingly taken offline, with inconsistent sequencing dates reported by WIV researchers, with huge conflicts of interest from several of the main researchers supporting the natural origin theory (Daszak), etc, etc, etc.

I'm about 80/20 on a lab leak.

I'd encourage you to read what she says before you decide. Her background should make you take her more seriously that if she was, say, a software engineer, but less seriously than if she was a virologist. But for me, her background is only one factor in my decision of how seriously I take her. A bigger factor is how serious, informed and reasonable what she says is.

Thanks for pointing that out. I agree with your first point about the book deal.

On your second point, she has stated repeatedly and explicitly that her position is that she doesn't know what the origin of the virus is, natural origin is not ruled out either, and that all she is asking for is that both theories are credibly investigated, for example here [1]. Of course she talks a lot more about the lab leak, but that's largely because she feels this side has not been discussed sufficiently, which was certainly true until 2/3 months ago.

On your third point, you're right that she truncated the quote, but I think her interpretation is the one many readers would come to: "lab leak=conspiracy theory". The quote is dishonestly structured to start with, lumping together lab leak with 5G/Bill Gates silliness. It's the same mistake that the Andersen paper made lumping lab leak with bioweapon. I wouldn't count that as a win on either side.

[1] https://twitter.com/ayjchan/status/1411049136762327043?s=21

An interesting aspect of the evolution of the lab leak theory is the unhealthy relation of mainstream media to scientific research. When the Andersen paper came out conflating the lab leak theory with the crazy bioweapon theory and lumping both as conspiracy theories, all mainstream medias jumped on this and used it for several month to quash discussion of the lab leak theory. The most vocal were the "science correspondents" and "decoders". I suspect this is because it's the kind of people that are sufficiently interested in science to have a healthy respect for it, but not sufficiently involved in science to have the healthy skepticism towards published papers. People that do publish and review scientific papers are well aware that published papers are just additional evidence for a theory, not a golden source of truth, but that seems largely lost on mainstream media. Especially when they can use it as anti-Trump material. For anyone interested in a balanced scientific view on the lab leak theory, Alina Chan on twitter is amazing.

Where does the potential energy that the water gets when moving up come from?

Vision models need the pixel ordering to match the one they have been trained on, in order to work.

They won't generalize after training to transformations of the data that they haven't been trained on, even simple ones such as rotations, whereas humans will.

So I would argue that vision model do use the "underlying structure", and even that one of their problems is that they make use of some of the "underlying structures" that are not actually important, such as image luminosity, rotations etc. I think people usually augments the data with these transformations beforehand during preprocessing to enforce invariance.

Sorry if that's a naive question but why do you need several signed messages? If you have a quantum computer and a quantum period finding function don't you get immediately the discrete log? Assuming you have one public key (not hashed) doesn't that give you the private key immediately?

This was as fascinating and engaging than the best documentaries I've seen. Thank you!

In the long run it is preferable to lead the world by example, through moral authority, which implies respect of international conventions. Authority derived from economic and military might can vanish quickly once economic and military rivals appear.

In related news: https://www.cnbc.com/2021/02/01/new-chart-shows-china-gdp-co...

What they are wrong about is that backdoors would only be "exceptionally used". It might be that way for a couple of years, then there will be a terrorist attack or a major child abuse scandal, new laws will be passed eroding legal protections, and before you know it any random official will be happily looking into your private conversations.

Think I'm exaggerating? This has already happened in the UK, RIPA bill was passed to "fight terrorism and child abuse" and ended up used by local officials to identify people who let their dog shit on the sidewalk.

https://en.wikipedia.org/wiki/Regulation_of_Investigatory_Po...

It's so satisfying when the good guys significantly outskill the bad guys and humiliate them. This made my day.