Posts
Comments
This looks like a lot of fun!
Very interesting story. One question: you said that the prognosis is "many rounds of heavy-duty antibiotics to hopefully induce periods of remission," but my understanding was that UC was autoimmune, and that the standard treatment was steroids or other immune system modulators? That's certainly what the Mayo Clinic says. Did the specialist mention any of these kinds of medications to you?
AI has a history of surprising us with its capabilities. Throughout the last 50 years, AI and machine learning systems have kept gaining skills that were once thought to be uniquely human, such as playing chess, classifying images, telling stories, and making art. Already, we see the risks associated with these kinds of AI capabilities. We worry about bias in algorithms that guide sentencing decisions or polarization induced by algorithms that curate our social media feeds. But we have every reason to believe that trends in AI progress will continue. AI will likely move from classifying satellite imagery to actually deciding whether to order a drone strike, or from helping AI researchers conduct literature reviews to actually executing AI research. As these AI systems continue to grow more capable, our ability to understand and control them will tend to weaken, with potentially disastrous consequences. it is therefore critical that we build the technological foundation to ensure these systems share our values and the policy and regulatory foundation to ensure these systems are used for good.
Very very interesting, thanks for putting this together! On the initial question (does activated charcoal reduce blood alcohol levels), this seems like a case where a bit of N=1 science could quickly produce some useful results. A few shots of vodka + Gwern-style blinding + a breathalyzer might tell you a lot after a few goes.
Strongly upvoted! I found this to be very informative, if disturbing in places. Thank you for putting this together, and I look forward to any future posts looking at the questions you outlined.
With regards to mandates, something that happened this week is that big federal contractors, including Boeing and Lockheed Martin, formally announced their plans to comply with the contractor mandate.
Minor note: it looks like you misspelled Merck as "Merrick"in the second sentence and executive summary, and as "Merrik" later in the text.
Hmm I think you're right that digital voice protocols aren't allowed on in 462-467 MHz FRS band. But there are a couple regions of the 900-1000 MHz portion of the spectrum that are allocated to commercial two-way radio, which allows digital and might be usable for this application if you were trying to actually manufacture and sell this kind of product.
If one were trying to do this as a homebrew project, either the 902-928 MHz ISM band used by the XBEEs or the 420-450 MHz band used for ham radio could work (with a ham radio license).
I don't think it's an impossible idea! XBEE radios in the form factor of a large watch can give you 200 kbps at 1000 ft in an indoor/urban environment, drawing 229 mA at 3.3 V. That seems to be about twice the data rate needed for clear voice communication. A 400 mAh li-ion battery is about 1x1.5 inch, so in the right ballpark. Add a 1/4 watt for a mini speaker and you'd get you about 25 minutes of talk time, and if you halved the data rate, you'd get 1500 ft of range. It would be pretty unwieldy and probably not quite the performance you'd hope for if you tried to hack it together from hobbyist parts, but a proper manufacturer just might be able to make it work.
The only part I'm not too sure about is antenna design. The 970 MHz XBEEs are usually run with bulky duck antennas, which wouldn't work in a watch form factor. There are definitely designs out there for embedding antennas in wrist straps, but I don't know much about what kind of pattern you'd get.
Wow, I've read a decent amount of GPT-3 content, but this still surprised me with how good it was. I spent the whole time reading it half-expecting that there would be some kind of twist at the end about how you generated it. Very impressive!
Unfortunately for this explanation, COVID cases in the UK appear to be rising once again. At this point I'm at a loss for potential explanations of what could have caused the rapid rise, fall, and rise again in cases.
Ah I see, thanks for the info
It's possible that seasonality is a factor, but temperature doesn't seem like it would be it. July and August are very similar temperatures in London, and both are slightly hotter than June.
I actually don't think the high level of antibodies should be such a surprise. I updated my original comment to clarify, but much of that is from vaccination, not from natural infection. Between high rates of vaccination plus historical infections, it's not surprising to me that such a high fraction of adults in the UK have antibodies.
Sure, the UK had high vaccination rates going into this wave, but the emergence of the delta variant (plus loosening of restrictions) greatly increased COVID transmission. So you'd expect a growth in case numbers. But if there aren't that many fully susceptible people to infect, the case counts wouldn't peak at a high number before turning around because of population immunity.
However, I want to be clear that I think this is just one factor, not the only thing going on. If you play around with SIR model parameters, you can see that you can't get quite as steep a drop in cases as there appears to be in the UK data just by inputting some reasonable values for delta. Changes in the number of tests, weather, behavior, etc. are all potentially playing a role as well, but I don't think we should discount the basic role of immunity.
Okay I see what you're saying here. But do you think that that a substantial number of negative utilitarians would agree with that argument? I don't think they would, because I think integral to many conceptions of negative utilitarianism is the idea that there's a qualitative difference between suffering and lack of pleasure.
One factor I've heard discussed is the UK's very high rate of prior immunity (from both infection and vaccination). About 90% of adults in the UK have COVID antibodies (either from vaccination or infection), meaning the fully susceptible population is very small. Obviously breakthrough cases are possible in people with antibodies, especially if they've only got one dose of the AZ vaccine, but R in that population will be much lower. Given the high initial rates of prior immunity, it would be expected that R will drop pretty quickly as COVID "buns though" the remaining susceptible population.
You might be correct, but I'm not convinced that all negative utilitarians would agree with you. I think that some formulations (e.g. potentially NHU as described here) would describe the person not getting tortured as producing a reduction in pleasure for the sadists, and thus not ascribe any moral value to the sadists' preferences not getting fulfilled.
I'd be curious to read more about your comment that "Positive and negative utilitarianism are equivalent whenever UFs are bounded and there are no births or deaths as a result of the decision." Do you have some resources you could link for me to read?
Ah you’re right, sorry. Edited.
I think negative utilitarianism is the most common ethical framework that would cause someone to choose the torture in the specks vs. torture case and no torture in this case. That's because in the specks vs. torture case involves people being harmed in both cases, whereas this case involves people gaining positive utility vs. someone being harmed. Some formulations of negative utilitarianism, like that advocated for by Brian Tomasik, would say that avoiding extreme suffering is the most important moral principle and would therefore argue in favor of avoiding torture in both cases. But a very simple negative utilitarian calculus might favor torture in the first case but not in the second.
I would guess that few people in the rationalist/EA community (and perhaps in the broader world as well) are likely to think that kind of simplistic negative utilitarian calculation is the morally correct one. My guess is that most people would either think that preventing extreme suffering is the most important or that a more standard utilitarian calculus is correct. For a well-reasoned argument against the negative utilitarian formulation, Toby Ord has a discussion of his point of view that's worth checking out.
I broadly agree with your point here about controlled burning, but I have two comments that I think add worthwhile context to the issue.
First, PG&E has an extremely poor track record when it comes to maintenance on transmission lines. Even with more controlled burning, it's not a good idea to be shooting off sparks in the middle of the dry season. So an optimal balance probably involves both PG&E spending more money on its infrastructure and doing more controlled burning.
Second, a lot of the deaths and injuries related to fires are the result of California's housing crisis pushing more people into urban-wilderness boundary areas. As development pushes out into more areas of chaparral (which as the Pew article you linked noted, consists of fast-growing shrubs that like to burn and can't really be cleared out over the long-run by controlled burning), you're pretty much guaranteed to have more damage from fire. Removing barriers to housing construction in existing urban areas is a key piece of the puzzle in protecting people from fire.
I had two doses of the Pfizer-BioNTech vaccine. I experienced a sore arm that went away after ~1 day after the first dose and moderate aches and chills that went away after ~12 hours after the second dose.
I know around 10 people in my age group (20-35 years old) that have gotten one of the MRNA vaccines and none of them had a significantly more severe reaction than I did.
To expand on this a bit, I think that people with working partners would be the group most likely to be deterred from working at MIRI if it was in either Bellingham or Peekskill. The two-body problem can be a serious constraint, and large metro areas tend to be much easier to find two jobs in. That may be getting better with the rise of remote work, but I do think it's worth keeping in mind.
My experience with elementary and middle school was that small numbers of snow days were not made up at the end of the year, but if more than a week or so was lost due to snow days, then the year would be extended. This appears to be the case in at least some other parts of the country as well. From a quick search: in Michigan "schools can be closed for up to six days before they must make up time to receive full funding from the state."
In defense of the position df fd took, you're playing a very asymmetric game here. Advertisers are investing very large sums of money and lots of person-hours of work to figure out how to change people's preferences with those 15-second ads. There's not a comparable degree of investment in developing techniques for making sure your desires aren't manipulated. I think it's hard to be totally sure that ads aren't subtly creating new associations or preferences that are intended to benefit the advertiser (potentially at the reader's expense).
Taking a bigger look, I think most people would agree that the average person in the United States makes at least a few irrational consumption decisions (such as buying a large expensive car, eating an unhealthy diet, or spending money on mobile games). There are lots of things one could point to in order to try and explain why that is, but I think it's potentially good evidence that people overall are susceptible to having their desires changed by advertising.
To add a bit more detail to your comment, this form of housing used to exist in the from of single room occupancy (SRO) buildings, where people would rent a single room and share bathroom and kitchen spaces. Reformers and planners started efforts to ban this form of housing starting around the early 20th century. From Wikipedia:
By the 1880s, urban reformers began working on modernizing cities; their efforts to create "uniformity within areas, less mixture of social classes, maximum privacy for each family, much lower density for many activities, buildings set back from the street, and a permanently built order" all meant that SRO hotels had to be cut back.[10] By the 1890s, SRO hotels became "forbidden housing; their residents, forbidden citizens."[10] New York City police inspector Thomas Byrnes stated that rather than give SRO hotels "palliative" care, they should be dealt with using a "knife, the blister, the amputating instruments."[12]
Reformers used moral codes, building codes, fire codes, zoning, planning committees and inspections to limit or remove SRO hotels.[12] An example of moral critiques is Simon Lubin's claims that "unregulated hotels" were "spreading venereal diseases among the soldiers".[12] Other reformers tried to ban men and boys from rooming in the same hotels, due to concerns about homosexuality.[12] The building and safety codes criticized SRO hotel problems such as "firetraps, dark rooms, inadequate plumbing, an insufficient ventilation."[12] In San Francisco, building code inspections and restrictions were often used to racially harass Chinese labourers and the places they lived.[12]
In 1917, California passed a new hotel act that prevented the building of new hotels with small cubicle rooms.[12] In addition to banning or restricting SRO hotels, land use reformers also passed zoning rules that indirectly reduced SROs: banning mixed residential and commercial use in neighbourhoods, an approach which meant that any remaining SRO hotel's residents would find it hard to eat at a local cafe or walk to a nearby corner grocery to buy food.[12] Non-residential uses such as religious institutions (churches) and professional offices (doctors, lawyers) were still permitted under these new zoning rules, but working class people (plumbers, mechanics) were not allowed to operate businesses such as garages or plumbing businesses.[12]
This fits into a set of ideas about urban planning that were popular in the 20th century but have (at least in my opinion) contributed to housing unaffordability and reduced the diversity and vitality of many of America's cities.
Yes! We had a nice discussion in the comments of your "Fun with +12 OOMs of Compute" post.
This is an interesting idea and I appreciate you putting this together. A few comments:
- I'm a bit skeptical of your comment in part 3 that "none of the biological/medical community has started doing something like this," as that just seems unlikely to me. I'm less deeply involved in biology than it sounds like you are, but how confident are you that these kinds of modeling efforts don't already exist?
- You say "I also (without a mechanism) believe that Aβ plaques (and smaller soluble aggregates) have some feedback effect which further damages glucose metabolism in the brain." How confident are you that this kind of feedback loop is really needed to make the model work? If the rest of your model holds true, could disease progression just be due to further mitochondrial damage occurring as people continue to age? And could AD tend to occur in people with excessive Aβ aggregation simply because this predisposes people to develop AD if they start having hyperphosphorylation of Tau proteins?
- There's some evidence that misfolded Aβ and Tau can have prion-like properties that can convert other Aβ and Tau into misfolded forms (https://sci-hub.se/10.1001/jamaneurol.2013.5847). Have you thought at all about how that might fit into the model?
This is a lovely companion writeup to Julia's. I especially liked your section on respect, I think that it's critical to not become Machiavellian when using this kind of method to shape the world.
I also wanted to add some personal thoughts related to your comments on fun. Part of my work involves being a mission controller for spacecraft. Training for that is long and daunting process involving going from being a trainee during simulated rehearsals, then being in a non-lead role during actual operations, then being in a lead role in simulations, and finally getting to lead actual operations. It can take a year or more between steps in that chain, so bridging stimuli are absolutely essential to making the time between advances not feel endless. I would describe the emotion that the process is headed towards more as "satisfaction" than as "fun," and bridging stimuli are things that let you immediately experience some of the eventual satisfaction of advancement. The kinds of bridging stimuli I can think of might be a thumbs up from a team member, a compliment for something well done, or being temporarily allowed to do something like make a call over the comm net when it's not normally your role. These are a little more personal than the clicker, but I think it's really getting at the same basic idea.
I don't want to get too into the weeds here. But I think that someone in the top few percent of their school would be smart. The kind of kid who might be feeling without intellectual peers and posting here about it could be the smartest in their school or their town (or they could not). But I don't think that really changes the conclusions.
I think of "smart" as (at least approximately) referring to g.
As I read this post, I couldn't help but feel like it crossed over from reasonable advice into elitism. I wouldn't argue with the basic idea of surrounding yourself with smart, interesting people who will help stimulate you intellectually and push you to grow. But spending your time worrying about how many standard deviations above the mean everyone's IQ is seems like it's crossing a line from reasonable to excessive, particularly when you describe normal life as "living among the monkeys."
So what I'd say to this is that yes, finding and befriending people who are smart is great. Same with people who are kind, people who are sincere, people who have a deep understanding of their emotions, people who make art, people who can build things with their hands, people who care for others, people who support their communities...
If you're a smart kid in high school it can be especially frustrating to not feel like you have intellectual peers. I absolutely get that, but I want to caution against advice that deemphasizes all the other worthwhile things about people in response. People are good and valuable and can teach you in myriad ways.
Thanks for this writeup. Could you share a bit more about how you got into using Vim and why you've found it to improve speed so much? I occasionally need to use vi when there's nothing else installed on a system, but the clunkiness and high barrier to entry has never made me tempted to use Vim as my primary editor.
An update to this comment: there is now some evidence to suggest the rates of cerebral venous sinus thrombosis are substantially different in UK recipients of the two vaccines. It is a very low rate (30 in 28 million), but there does seem to be a real difference there.
From Twitter, it looks like the rates of clotting-related issues in UK recipients of the AstraZeneca and Pfizer/BioNTech vaccines are essentially indistinguishable.
The new Apple M1-based mac mini appears to be able to do 2.6 teraflops on a power consumption of 39 W. That comes out to 0.000066 W/petaflop, or ~4x the efficiency of Fugaku.
Your comment about Crystal Nights makes sense. I guess humans have evolved in a word based on one set of physical laws, but we're general purpose intelligences that can do things like play videogames really well even when the game's physics don't match the real world's.
Really interesting post, I appreciate the thought experiment. I have one comment on it related to the Crystal Nights and Skunkworks sections, based on my own experience in the aerospace world. There are lots of problems that I deal with today where the limiting factor is the existence of high-quality experimental data (for example, propellant slosh dynamics in zero-g). This has two implications:
- For the "Crystal Nights" example, I think that our current ability to build virtual worlds that are useful for evolutionarily creating truly transformative AIs may be more limited than you might think. A standard physics simulation-based environment is likely to not be that good a map of the real world. And a truly "bottom-up" simulation environment that recreated physics by simulating down at the molecular level would require a few orders of magnitude more computing power (and may run into similar issues with fidelity training data for modeling molecular interactions, though alphafold is evidence that this is not as great a limitation).
- For the "Skunkworks" example, I think that you may run into similar problems where the returns to more computing power are greatly limited by the fidelity of the training data.
Now if fidelity of training data was the only thing holding Google et al. from making trillions off of AI in this world, there would be a very strong push to gather the necessary data. But that kind of work in the physical world tends to move more slowly and could well push the timelines required for these two applications past the 4-year mark. I couldn't find similar objections to the other three.
I finally got around to making these! I was very pleased with the result, they were tasty and distinct from anything I've had before. While I thought they were about as delicious as most homemade cookies, my partner who is not generally a huge fan of cookies liked them much more than previous cookie attempts and kept coming back for more.
I agree with the other commenters who've suggested that like with most homemade cookies, they're better than store-bought cookies because they don't have to last for months on a store shelf. But I am surprised it's not a more popular home recipe in the US as it's about the easiest recipe for tasty cookies I've come across.
This comment (and the whole discussion) really resonated with me. I think a hard part of this is that if I try and totally remove the activities that allow for opting out of being (video games, mindless reddit scrolling etc.), it tends to only work for a short time before I relapse all at once into them. It seems like this is a case where moderation might be the answer for me personally rather than abstinence.
One unexpected positive of Hammertime is that I've noticed my desire to play video games gradually decreasing over the last month. This might be an interesting case where the solution to the problem is to solve other life problems, at which point the desire to cease to exist simply fades away.
Overestimation: Interacting with external reviewers/customers at work. I thought I had useful things to contribute to discussions with external folks starting maybe 4-5 months into my job. I didn't understand how to handle those interactions tactfully (and overestimated the chillness of by bosses) and got slapped down pretty hard.
Underestimation: Research ability as an undergrad. I kept thinking I was a fraud and doing terrible work right up until the day I won the research top prize in my department.
Done well: I really like the daily prompts to comment, I think they've done a lot to encourage me to stick with it. They've also been nice because I get to see everyone else's responses.
Done badly: I wish more days had direct connections to the bug list (i.e. more challenges directly of the form "pick a bug from the bug list and apply today's technique to it"). It's harder to motivate myself to tackle challenges on the bug list when it's implicit that today's technique can be applied to them than it is when it's explicit.
I applied the method of exhaustion to my course final project this semester, breaking it into 9 steps. It was a fun exercise, and I appreciated it!
It's interesting, you definitely see failure rates that look logarithmic in marriages and bankruptcies, but I do think that some of that is what tcheasdfjkl said - some of that is just from the fact that things that fail early don't get a chance to fail later. In my personal experience, I think there are two big places where my plans fail: before they start and at the first major setback. I think that usually if I can get started on something and keep going past the first time there's a problem, I can usually overcome future problems. But sometimes the first setback is enough to make me set something aside, and I just never end up coming back.
Interestingly enough, five minutes wasn't enough for me to get any improvement in typeracer or the arithmetic game. I started at 80 WPM and got results both above and below that on my subsequent tries. Similarly I got 20 on my first attempt at the arithmetic game, 23 on the second attempt, and 17 on the third. I don't think that makes speed impossible to train, it just suggests that it'll take longer than 5 minutes!
Here are some of my proudest speed records:
- Performance on exams when time is tight. For example, I had an undergraduate materials science exam where the professor had grossly underestimated the required time, making it a pure game of speed. I got a 93 when the average was in the low 60s.
- Finishing an anomaly investigation at work. I was assigned to lead the investigation and resolution efforts to a significant issue found in one of our systems. I found the root cause, came up with a corrective action, and got agreement from our customer and independent reviewers in less than 1/2 the time it had taken on an earlier similar issue that been assigned to another engineer.
- Cooking breakfast! I'm a master at all-out speed maximization on making breakfast burritos, avocado toast, eggs etc. I've gotten really good at prepping absolutely no more than is needed before I start the longest-cooking item going, then using the time when things are toasting and cooking to get everything together in just a few minutes.
I do think I've gotten better at achieving my values over the course of Hammertime. the biggest way that's manifested has been in how I'm spending my time. I'm spending more time reading books and talking to friends rather than aimlessly browsing the internet or playing videogames. Interestingly enough, I've actually been spending less time working, which I wasn't consciously trying to do but I think is positive, as I generally work somewhat too hard.
Reflecting back on the bug list for the second time, I only came up with a modest number of additional bugs. The list currently stands at 132 bugs, of which I've solved 38. My favorite bug solve was to finally get around to hanging some art in the empty space near my desk.
This was a great post, and I fully agree with the idea that working to make our interactions with others positive is an important part of making life good. I found that a good additional exercise to go with this post was to set a Yoda timer and reach out to friends who I haven't talked to in a bit.
I like your weakened version of TDT, it feels like it does really capture something salient about human decision-making. I recognize that the exact number isn't really important, but I think I'd describe it as close to a one-percent shift than a ten-percent shift for myself. I feel like I personally have taken a very long time to go from the first few times I do something to that thing feeling natural. I wonder if that's something that tends to differ a lot between people or different kinds of actions.
I like Noah Smith and enjoyed that post, but I'm not sure I agree with the conclusions. I think I agree more with the critique from Applied Divinity Studies than I do with Noah. There are definitely areas for optimism, but I haven't seen anything that looks like we're actually getting increases in productivity growth in the US or similar countries. Moreover, I have seen no indications that cost disease has slowed (see healthcare costs for example, other than the decrease in care due to COVID lockdowns, there's no real evidence of a slowdown). And beyond all that, the ineffective response to COVID in the US has shown just how badly our state capacity for effectively confronting problems has deteriorated.
I would really, really like stagnation to be over (and I do my best to support to efforts to make that happen!), but I don't yet feel very hopeful.
Listing all of the times I've improved rapidly in the past was a very interesting exercise. Many of the times where I made a really big improvement stemmed from making a decision that was hard to make in the moment, but locked me into a path toward something good. That's very much in line with the basic principle is "take advantage of willpower now to put yourself on a course in which you don't need willpower to do the right thing later."
Another idea that seems common to a number of the examples I can come up with is an idea that I might call "seize the moment." It's very similar to the first one but perhaps just slightly distinct. The basic idea is if you do one hard thing and it works, you may find that lots of other hard things suddenly seem easier, at least momentarily. I've had times where I was able to get on a roll of compounding improvements that way, and I think it can be surprisingly successful at allowing you to make big changes.