gwillen

Karma: 2,132

gwillen 20 Jul 2021 19:28 UTC
11 points
on: Working With Monsters
I like it, but 95% seems surprisingly high. Surely there are plenty of other people out there with a similar psychological makeup to Red and Judge, or to the protagonist (who can at least be convinced, with a sufficient threat, to listen before punching.) But I shouldn’t fight the hypothetical too much...

gwillen 29 Jun 2021 21:44 UTC
3 points
on: Why did we wait so long for the threshing machine?
At a slightly different angle: it seems like the reason we couldn’t make the thresher is that, first, we had to create “the art of building things precisely enough to make the thresher”. That is itself a technology, but it’s the sort of technology for which the very concept doesn’t really get reified before you have it, so you don’t even have general understanding that it’s what you’re missing.

From the later parts of the post, it seems like that may not be quite right, since it sounds like other places / industries had more of that technology. But I still like the general concept of “a technology that is so abstract we don’t really understand that we need it yet”, as a prerequisite to some other obviously-desirable thing. “Infrastructure” seems like not quite the thing but having a lot of overlap with it.

gwillen 31 May 2021 17:51 UTC
8 points
on: Electric heat pumps (Mini-Splits) vs Natural gas boilers
If the outside air temperature gets very low (or high?) I’ve heard that ground-source heat pumps might be the way to go: https://en.wikipedia.org/wiki/Geothermal_heat_pump

gwillen 25 May 2021 19:40 UTC
8 points
on: Oracles, Informers, and Controllers
Reminds me of Scott’s story from way back when about the Whispering Earring: http://web.archive.org/web/20121008025245/http://squid314.livejournal.com/332946.html

(I don’t have any specific point, I’m not sure the story has any specific lesson that I endorse, it just seemed interestingly related.)

gwillen 25 May 2021 0:42 UTC
6 points
in reply to: Scott Garrabrant’s comment on: Finite Factored Sets
That sounds like the right choice, but a part of me is incredibly disappointed that you didn’t go for it.

gwillen 25 May 2021 0:37 UTC
6 points
on: We should probably buy ADA?
This is interesting. There’s a lot here, I agree with some but not all of it, and I’m not presently going to comment on most of it (mostly for time reasons.) I am a Bitcoin user and developer, and have looked briefly at Cardano in the past. One thing I’d specifically like to comment on:

“but we already know proof of stake works, this isn’t necessary”

We really don’t, though. We know for sure that the original proof of stake algorithm was badly broken, in multiple ways. There are now a number of different successors that go by the name “proof of stake”. I have no specific reason to believe that Cardano’s is not broken in some way. They appear to use a PoS system called “Ouroboros” that they developed themselves, starting in 2017, according to https://iohk.io/en/research/library/authors/aggelos-kiayias/ . It looks like the original 2017 paper claims that the system is “provably secure”, but judging by the list of papers it’s also undergone a number of iterations since then. (That doesn’t necessarily imply anything bad, but neither does having “provably secure” in a paper title really prove anything good. I haven’t read the paper.) I’ve been meaning to specifically investigate the ETH 2.0 version of PoS, to see what I think about it, since Vitalik has written about its design at some length; I may have to add this one to the reading list.

gwillen 19 May 2021 7:46 UTC
20 points
on: Saving Time

agency is time travel

Since time is the direction of increased entropy, this feels like it has some deep connection to the notion of agents as things that reduce entropy (only locally, obviously) to achieve their preferences: https://www.lesswrong.com/posts/Q4hLMDrFd8fbteeZ8/measuring-optimization-power (I’m not sure this is 100% on-point to what I mean, but it’s the closest thing I could find.)

Since agents can only decrease entropy locally, not globally, I wonder if we could similarly say that they can “reverse the arrow of time” only in some local sense. (Omega can predict me, but I can’t predict Omega. And even Omega can’t predict everything, because something something second law of thermodynamics.)

gwillen 8 May 2021 5:48 UTC
7 points
on: Covid 5/6: Vaccine Patent Suspension
The graph in the CNBC tweet seems pretty misleading without context, and I’ve seen it criticized elsewhere for this reason. Since roughly the start of 2020, NovaVax stock is up 4300%, Moderna is up 730%, and BioNTech is up 460%. (Pfizer is only up 6%, but it’s also about an order of magnitude larger, so a single product line would not have the same effect on the stock price.) Three of the four are still higher than they were in just mid-April; NovaVax is about the same as it was. And I’m trying to be conservative here, and attribute all the recent price drops to the patent issue, but the biggest drops happened in the week before ⁵⁄₅, and are not visible on the graph in the tweet.

Granted, the stock prices have recovered a bit since you wrote this, which is information you didn’t have, but even still—the price drop depicted in that tweet is a pretty minor fluctuation in the context of even a week or two of price movements, and even considering the total drop from most recent peak to now (which has different timing for each stock, so presumably they aren’t 100% from a common cause) you’re looking at drops of 3-25%, which is obviously not tiny, but it hasn’t come anywhere near wiping out the tremendous growth in the value that the market has placed on these companies since the start of COVID. The market doesn’t love this news, but it still likes all four companies pretty well regardless.

gwillen 5 May 2021 19:34 UTC
17 points
on: [link] If something seems unusually hard for you, see if you’re missing a minor insight
Meta: The way this post works in the LW interface feels weird to me. The email notification doesn’t mention that it’s a linkpost at all, and doesn’t include the link. And linkposts on LW usually seem to mirror the content of the link, but this one is a summary instead—which is fine, except that it doesn’t feel very obvious. The link is small, in italics, and easy to read right past, especially when it’s usually just a link to another copy of the same content.

gwillen 7 Apr 2021 5:49 UTC
9 points
in reply to: HDMI Cable’s comment on: Don’t Sell Your Soul
This seems like a classic Pascal’s Wager, and as such probably not a great strategy to follow—see “Pascal’s Mugging” for a discussion of what happens if you start letting people push you around by declaring things to have infinite (or incredibly large) utility in some unlikely scenario.

… but I can’t figure out how this comment got down to −9 points on that basis alone. I’m wondering if your username caused a bunch of people to assume—as I did initially—that you were a spambot, and thus downvote you extra-harshly after a bare skim of your actual comment.

gwillen 2 Apr 2021 17:33 UTC
4 points
on: 2012 Robin Hanson comment on “Intelligence Explosion: Evidence and Import”
I think my immediate objection to Robin’s take can be summarized with a classic “shit rationalists say” quote: “You have no idea how BIG mindspace is.” Sure, we quibble over how much it’s okay to impose our values on our descendants, but those value differences are about (warning: tongue firmly in cheek) relatively trivial things like “what fictional deity to believe in” or “whether it’s okay to kill our enemies, and who they are”. Granted, Robin talks about acceleration of value change, but value differences like “whether to convert the Earth and its occupants into computronium” seem like a significant discontinuity with previous value differences among generations of humans.

Humans may have different values from their ancestors, but it is not typical for them to have both the capability and the desire (or lack of inhibition) to exterminate those ancestors. If it was, presumably “value totalitarianism” would be a lot more popular. Perhaps Robin doesn’t believe that AGI would have such a large value difference with humans; but in that case we’ve come back into the realm of a factual dispute, rather than a philosophical one.

(It does seem like a very Robin Hanson take to end by saying: Although my opponents advocate committing what I clearly suggest should be considered an atrocity against their descendants, “[e]ven so, I must admit that [it] deserves to be among the range of responses considered to future intergenerational conflicts.”)

gwillen 1 Apr 2021 5:05 UTC
4 points
on: Core Pathways of Aging

Main upshot of all this: since aging involves changes on a timescale of decades, there must be some component which is out-of-equilibrium on a timescale of decades or longer (i.e. does not turn over significantly across a full human lifespan). These are the components which we’ll call “root causes”. Everything else which changes with age, changes only in response to the root causes.

A quibble: Just because some component turns over frequently, doesn’t mean that higher-level structures made from that component aren’t degraded in the process. For example, if I accidentally cut off the tip of my finger, the relevant cells will all grow back, but the finger will not; the larger-scale pattern remains degraded for life.

In the case of my fingertip, obviously we would consider that an injury, not an aspect of aging. But it seems hard to be sure that there aren’t any aspects of aging that work this way?

gwillen 17 Mar 2021 0:49 UTC
4 points
in reply to: Gerald Monroe’s comment on: Nitric Oxide Spray… a cure for COVID19??
The linked site appears to be down, which is definitely not making this seem more trustworthy.

EDIT: I guess that’s just one random site with the press release.

Interestingly, the clinical trial seems to be pre-registered: https://clinicaltrials.gov/ct2/show/NCT04443868 Not sure how much I should infer from that about how good it was.

gwillen 16 Mar 2021 22:42 UTC
2 points
in reply to: cistran’s comment on: Covid 3/12: New CDC Guidelines Available

One objection to that is that one must not just compare between vaccinating and getting Covid. One must also strive to pick the safest vaccine.

Well, as to the vaccines available right now, I’m not aware of any evidence for one of them being more or less safe than another, so your choices seem limited to “vaccine” or “not”. But even if they were different—getting the safest one is only a usable strategy if it’s available, i.e. there is not a vaccine shortage, which there currently is. If your choices are “get whatever vaccine is on offer” and “nothing”, you should get whatever vaccine is on offer as long as the expected value of doing so is better than the risk of COVID exposure if you do nothing.

Another objection is that long term effects of any covid vaccine are not yet well-studied and there could be plausible mechanisms by which some of the vaccines could cause long term damage for example due to cumulative effects.

This is definitely the scariest hypothetical, IMO, but I’m not aware of any evidence for it, only a lack of long-term data. How you weigh “unknown unknowns that are hard to measure” against the risks of COVID seems like a very personal choice. (I’m not aware of any past vaccines having hidden long-term side effects that didn’t appear at all in trials or early use.)

gwillen 14 Mar 2021 21:57 UTC
2 points
in reply to: cistran’s comment on: Covid 3/12: New CDC Guidelines Available
Supposing hypothetically that those 1400 deaths were all caused by the vaccines—wouldn’t the math still be on the side of getting one? (Of course, most likely the relationship is not causal, as discussed by the CDC link below, as well as Zvi’s reply, but hypothetically assuming it was.)

from https://www.cdc.gov/coronavirus/2019-ncov/vaccines/safety/adverse-events.html:

“Over 92 million doses of COVID-19 vaccines were administered in the United States from December 14, 2020, through March 8, 2021. During this time, VAERS received 1,637 reports of death (0.0018%) among people who received a COVID-19 vaccine.”

Chance of death from COVID for young healthy people is probably somewhere between 0.05% and 0.1%. (CFR seems to be around 0.2%, but IFR is lower. I’m eyeballing from https://link.springer.com/article/10.1007/s10654-020-00698-1 .) For older people it can be as high as 10-20%. Even if you are in the youngest, healthiest group, making the most optimistic assumptions about the danger of COVID, and the most pessimistic assumptions about the danger of the vaccine, that it caused every single death reported after receiving it—it seems to me that COVID would still be at least 20x more dangerous. And realistically the difference is much higher than that, and for older people it’s going to be orders of magnitude higher.

Of course, if you think you can perfectly avoid exposure to COVID, then you could declare any amount of risk from the vaccine to be too much. But I think most people are both unable and unwilling to do that, and I suspect that most people who ARE able and willing to completely avoid exposure are also more likely to be getting the vaccine, not less.

gwillen 7 Mar 2021 13:43 UTC
5 points
in reply to: ChristianKl’s comment on: A whirlwind tour of Ethereum finance
China could just pass a law that says: “It’s illegal to sign [certain blocks]. Without a fork that would mean that the Chinese miners are able to have the longest blockchain and thus the official one. I don’t see an easy way to fork and invalidate the ability for the Chinese miners to do that either. You would likely need to do things like changing the underlying signing algorithms to invalidate the ASICs.

I agree that you’re right about what would initially happen in this scenario (miner majority conspires to censor transactions).

It’s worth noting that you can only tighten restrictions this way, not loosen them (e.g. China can’t pass a law sending themselves all the coins, miner majority or not, without automatically forcing everyone else into a separate fork.)

The difficulty level of responding to this would depend on the policy goals of the Chinese government, and the level of coordination of the Chinese miners. If they were content to have their own China-Bitcoin fork, it would be easy enough to coordinate to let them go, and nothing as drastic as a PoW change would required. If not, it would certainly get messy, although I think we have some reasonable approaches prepared for this contingency, short of nuclear options.

If a substantial amount of the value of cryptocurrency comes out of the DeFi stuff that will likely mean that a crypto network that allows that stuff will move to the top.

I agree. Whether it’s worth the complexity is an interesting open question.

I don’t see much difference between Binance losing 40$ million worth of Bitcoin in a hack of their software to a smart contract on the Ethereum network being hacked.

Here I do see a tremendously large difference:
- First of all, Binance is an exchange which trades both Bitcoin and Ethereum, and holds balances of both. The fact that they lost Bitcoin in the hack, and not Ether, is only because Bitcoin is more popular than Ethereum. There is zero structural advantage for Ethereum here.
- Binance is a third-party website. I don’t know what standard practice in Ethereum is—hopefully the same—but it is strongly recommended in the Bitcoin community never to store coins with a third party. By contrast, Parity is one of the standard Ethereum local wallet apps, and the affected contract was the built-in multisig option in that app. “I gave my coins to a third party and they got lost” is very different from “I stored my coins in the official wallet app on my local computer and they got lost.”
- Also, the amount in question was apparently ~7 times as large, in USD terms (or ~4x from another source): https://www.theguardian.com/technology/2017/nov/08/cryptocurrency-300m-dollars-stolen-bug-ether . And worse in market cap terms: this single bug destroyed about 1% of all Ether in existence (or 0.5%, according to a different source). The Binance hack represented about 0.03% of the Bitcoin in existence at that time.
- In point of fact, Binance users did not lose funds—as far as I can tell, Binance covered all the losses. By contrast, in the Parity multisig incident, all affected user funds were—as far as I know—lost irrecoverably.
- It’s also worth noting, if only for amusement value, that the Parity multisig issue was not technically a “hack”; the bug was triggered by accident, and the coins were lost (not taken). (Although the previous incident in the same contract, just 4 months prior was a theft.)
Obviously these points are applicable to these specific examples. There were other Ethereum contract losses where users did realistically know they were putting funds at risk, unlike the Parity incident. And there were other Bitcoin incidents where user funds did get lost, and were not returned (although of course they all involved storage of funds with third parties, not bugs in the standard client software, that I’m aware of.) But the sheer volume, severity, and general nature of the issues seems incomparable to me.

gwillen 6 Mar 2021 21:14 UTC
15 points
in reply to: ChristianKl’s comment on: A whirlwind tour of Ethereum finance
Hm, I guess I’m conflating two different things.

One of them is really more about governance than engineering, although it mixes both. The Bitcoin model is “this thing exists, it has to work, we’re not going to screw with it unless there is overwhelming consensus to do so.” And in fact it is structurally pretty hard to screw with it, absent overwhelming consensus to do so; if you screw it up, you fork the chain. (Realistically, the core devs could probably force something through once, at great reputational cost; but both the core devs themselves, and the user community, are philosophically opposed to that kind of thing.)

Meanwhile, the Ethereum model seems to be “you run what the devs tell you to run.” If we want to change the rules, we change the rules. Most people don’t even run full nodes; they use a small number of major service providers. And incidents like this are scary: https://www.coindesk.com/ethereum-service-providers-experiencing-issues-after-reported-blockchain-split . The devs released new software that changed the rules; major service providers were not given advance notice and did not update; this broke a ton of stuff. Since the devs define the protocol, those service providers were then forced to update.

The last accidental hard fork of the Bitcoin chain, and the only one I’m aware of, was in 2013. https://github.com/bitcoin/bips/blob/master/bip-0050.mediawiki . It was due to a bug, and unlike the Ethereum scenario, the clear Bitcoin community consensus was to roll back to the version before the accidental break. Nobody in Bitcoin has the mandate to roll out a breaking change and force everybody to adopt it. (You can definitely argue about the various community splits on breaking change proposals in the past, but IMO this underscores that the community’s view of a change IS the critical determining factor; the devs can’t just release a breaking change and expect everybody to go along, whereas in Ethereum that’s the norm.)

From a more engineering-oriented and less governance-oriented view, I guess my main concern is that Ethereum’s scripting model is very complicated compared to Bitcoin’s. I’m not aware of a Bitcoin script ever being hacked, granted that this is because Bitcoin scripts are much less capable than Ethereum scripts! Meanwhile, here is a list of Ethereum smart contract postmortems: https://forum.openzeppelin.com/t/list-of-ethereum-smart-contracts-post-mortems/1191 . It’s a long list.

The worst one I’m aware of is https://www.parity.io/security-alert-2/ . Parity was/is one of the first-party, supported, officially-sanctioned Ethereum nodes/wallets. (It’s hard for me to tell whether it’s currently the recommended full node to run yourself, because the Ethereum website does not recommend ANY full node software to run yourself, that I can see; only light wallets and third-party hosted wallets. One of the recommended open-source Bitcoin wallets, Bitcoin Core, is a full node.)

In the linked incident, ALL PARITY MULTISIG WALLETS AND THEIR CONTENTS WERE DESTROYED. (And I had forgotten, but that site reminds me, that this was actually the SECOND issue with Parity multisig wallets, and was introduced in the fix to a previous issue. The link to the previous issue is dead, so I don’t know whether that also resulted in loss of all funds.)

This isn’t a detailed or exhaustive investigation, and of course not all of this can be blamed on Ethereum itself. But generally I would say that my impression is that Ethereum has lower engineering standards than Bitcoin does.

For an example of good engineering on the Bitcoin side: Some Bitcoin devs did exhaustive testing of some of Bitcoin’s cryptograpy against OpenSSL, in order to check for bugs in the Bitcoin code. Instead, they found a bug in OpenSSL: https://www.openssl.org/news/secadv/20150108.txt (search-in-page for “squaring”.) Obviously there could be stuff like this in the Ethereum camp, too, and I wouldn’t necessarily know since I mostly only deal with Bitcoin.

gwillen 6 Mar 2021 19:16 UTC
13 points
in reply to: ChristianKl’s comment on: A whirlwind tour of Ethereum finance
Meta: did you delete and repost this comment or something? The link to it from the email notification gives me a 404.

For the benefit of the admins, the original link I got was:

http://email.mg.lesserwrong.com/c/eJyFj0FrxCAYRH-NuRkSTVw9eFi6u9BCc-3Z6md0m-hWDdL—tpLr4WBgcfwYIxUloiJdl6ykQgy6wEDvAtsxQSYaS2wIXY0VmlFtEDTsEHOkGqKYe113Dsn6UwGY82omWBkMhOn5sRPnOgBBNiZdZt0pTwyomdEbi211v5X8ydp7P_FI-aSW4fXxXPm1uWuluC4PxpTuDqftuqDwSUeCUeLoThIcOzY-qCCBkRvTbRDKM8G0Qu_vrlve8-cfOqXr_PHU5fkWv22QWgvAyQToPQxrT8j4l_r

which serves a redirect to:

https://www.lesswrong.com/https://www.lesswrong.com/posts/nMNi86hgNjaNnh8iu/a-whirlwind-tour-of-ethereum-finance?commentId=8EWhzfjs82qcJyAkC

EDIT: Oh, I see the problem. It looks like they broke the comment link redirects. See the doubled first half of the URL.

FURTHER EDIT: Reported as https://github.com/LessWrong2/Lesswrong2/issues/3957 .

gwillen 6 Mar 2021 19:13 UTC
6 points
in reply to: ChristianKl’s comment on: A whirlwind tour of Ethereum finance
This is a really good response and it made me think, thanks.

I should disclose (I wasn’t meaning to hide it before, but in retrospect it’s clearly relevant) that I do have a position in ETH, much smaller than my position in BTC, as a hedge against a scenario like this.

Historically, the correlation between the two has been pretty strong, as has the correlation among all the cryptocurrencies. But if they successfully switch to Ethereum 2.0, that could have a major effect here.

Overall, my sense is that Ethereum has a weaker foundation, engineering-wise, than Bitcoin; but it also obviously has a LOT more features (some of which can be footguns.) And there’s no inherent reason that Ethereum can’t gradually improve the engineering while retaining the features, and it’s definitely been working on doing that. On the other hand, Bitcoin has a fundamental philosophical conservatism that I think makes it a better long-term bet in some ways. Assuming it survives, you can bet that it will still be recognizably Bitcoin well into the future. Who really knows what Ethereum will be—the governance model allows for a lot more change, which can be good or bad.

gwillen 3 Mar 2021 23:17 UTC
21 points
on: A whirlwind tour of Ethereum finance
I work in Bitcoin professionally, have not spent a lot of time looking at Ethereum/DeFi stuff, and found this surprisingly educational.

I will make one broken-record comment, which I always repeat when the topic of DeFi comes up. These things typically require you to lock up some collateral, in a way that subjects you to the price fluctuations of some cryptocurrency on top of whatever DeFi thing you’re doing. Your ultimate return may have a lot more more to do with those price fluctuations, than with the DeFi stuff itself.

That may be fine, but make sure you account for it. In the easiest case, you’re already holding whatever cryptocurrency you’d be exposed to. In the hardest case, you are using money that was previously in the stock market / your bank account, and exposing it to changes in the market value of some obscure token that nobody’s ever heard of. Making 25% APR on your investment in that token is probably meaningless.

(This is VERY not investment advice, but: if you want to go long on “crypto-mania”, my first suggestion would be that you (1) buy some Bitcoin, and (2) don’t sell it. It’s a lot less effort than DeFi, and so far the return has been pretty good. Future returns may not be so good! But the correlations between cryptocurrencies are pretty high, so if Bitcoin does badly, whatever token your DeFi thing uses probably will also do badly, over the medium term at least.)