I give full permission for anyone to post part or all of any of my comments/posts to other platforms, with attribution.
Currently open to international job offers and interested in getting a Masters/PhD. Experienced in bioinformatics and summarizing research.
I am. I had pretty low grades in College (~2.0 GDP), and from what I’ve read I would need a lot of work experience and accomplishments to get into a Master’s program. I think I need to convince a professor directly in order to get in. Do you have any recommendations?
If you’re ever actually seriously considering human extinction, you should probably realize that you’re much more likely to be deluded or mentally ill than actually facing the dilemma you think you’re facing. The correct play here would be to check yourself into a mental hospital.
I also apply this principle if Omega ever appears to tell me that I can trust something with p=1. Since Omega doesn’t actually exist, I’m almost certainly on a bad acid trip and should go lie down somewhere.
The band of malicious messages generated by advanced aliens that will both kill us and be detected by a panel of human experts seems extremely narrow. We’re talking about the FDA being able to outsmart an AI running on a Matrioshka brain levels of narrow.
We then have to weigh the potential gains we get from utilizing the message without having to wait for the speed of bureaucracy. How many lives can we save by earlier implementation of alien technology? Plus, confirmation of alien civilizations make existential risk just regular risk. If we can see them, the fate of our light cone depends on their magnamity anyways.
My calculations put the odds of a committee being able to achieve any good so low that we’re better off just not having one.
I assume this is to promote the infohazard policies that a lot of Rat-associated AI alignment teams are using?
Epistemic status: ~30% sophistry
There’s a difference between thinking something won’t work in practice and opposing it. The paper examines opposition. As in, taking steps to make geoengineering less likely.
My personal suspicion is that the more plausible a climate scientist thinks geoengineering is, the more likely they are to oppose it, not the other way around. Just like climate modeling for nuclear famine isn’t actually about accurate climate modeling and finding ways to mitigate deaths from starvation (it’s about opposition to nuclear war), I suspect that a lot of global-warming research is more about opposing capitalism/technology/greed/industrial development than it is about finding practical ways to mitigate the damage.
This is because these fields are about using utilizing how bad their catastrophe is to shape policy. Mitigation efforts hamper that. Therefore, they must prevent mitigation research. There is a difference between thinking about a catastrophe as an ideological tool, in which case you actively avoid talking about mitigation measures and actively sabotage mitigation efforts, and thinking about it as a problem to be solved, in which you absolutely invest in damage mitigation. Most nuclear famine and global warming research absolutely seem like the former, while AI safety still looks like the latter.
AI alignment orgs aren’t trying to sabotage mitigation (prosaic alignment?). The people working on, for example, interpretability and value alignment, might view prosaic alignment as ineffective, but they aren’t trying to prevent it from taking place. Even those who want to slow down AI development aren’t trying to stop prosaic alignment research. Despite the surface similarities, there are fundamental differences between the fields of climate change and AI alignment research.
I vaguely remember reading a comment on Lesswrong about how the anti-geoengineering stance is actually 3d chess to force governments to reduce carbon emissions by removing mitigation as an option. But the most likely result is just that poor nations like Bangladesh suffer humanitarian catastrophes while the developed world twiddles its thumbs. I don’t think there is any equivalent in the field of AI safety.
On one hand, I agree that nothing really special and novel is happening in Cicero. On the other hand, something about it makes me feel like it’s important. I think it’s the plan->communicate->alter plan->repeat cycle partly taking place in English that intuitively makes me think “oh shit, that’s basically how I think. If we scale this up, it’ll do everything I can do”. I don’t know how true this actually is.
I vaguely recall feeling something like this when a general-purpose model learned to play all the Atari games. But I’m feeling it a lot more now. Maybe it’s the fact that if you showed the results of this to pre-GPT2 me, I’d think it’s an AGI, with zero doubt or hesitation.
How did we get here? How did Rationality go from being about winning at life to a Cthullu RPG? I’m pretty interested in AI, and the fact that stuff like this is real (it still doesn’t feel real, it feels like LARP) has really turned me off of alignment research. On a personal level, the literally-having-coworkers-get-driven-insane of alignment research seems a lot worse than merely-slightly-increasing-existential-risk of capabilities research.
Actually, now that I think about it, the most likely scenario is that China shuts down all incoming international travel permanently, even for Chinese citizens. Speaking of which, do we know how new cases are popping up in China? Surely it’s from international travel, right?
I can see zero-covid holding indefinitely. People here are fundamentally OK with daily COVID tests and 1, 2, maybe even 3 months of lockdown a year forever. There are limits to the infectiousness of diseases, and I don’t see any disease spreading in a condition of permanent lockdown.
You’re the one with the novel hypothesis here. The burden of proof is on you. And I am rightfully suspicious of survey results and believe survey-only studies are of almost no use. Why?
Because self-report is fundamentally unreliable. Yes, I am absolutely saying that people who receive money from you are incentivized to lie to you out of a sense of social obligation. Yes, I am saying that this means any study where you provide help and then rely on self-report as the only form of feedback is borderline useless and will convince few people. And no, it’s not an unfalsifiable idea—objective metrics can be used as a comparison. If it is impossible to collect such data ethically, then you should stop doing your studies immediately. If your best piece of evidence in support of your ideas is just “think about it logically”, I would highly recommend that EA charities never fund this project.
While people may feel an obligation to give good answers about drug use, for example, funding has not been determinant on the answers to the questions. Why would anyone go out of their way to lie about an easily checkable questions like: Where are you living at? Do you have a job? Or what is your net worth? People have no incentive to missstate answers to those questions.
To avoid embarrassment and awkwardness. To make the researchers feel better. This is a well-known problem in psychology called the “self-report bias”, which is why good psychological studies never rely on self-report alone. As verification, you can also interview other family members, and employers, check arrest records, ect. If there is no way for you to do this ethically, then what do you think your studies are achieving?
I’ve never seen a single example of a random in-person check-in discovering that a participant is actually still homeless, having lied about attaining a stable place to live.
You ask someone if it’s OK for you to drop by at their place sometime, and for their address. This is verifiable information, and they aren’t likely to lie to you. If you ask them about things they know you can’t check, they are much more likely to make a socially desirable lie. They know you can’t check how much money they’re spending on drugs, or if they’re employed. People even lie to themselves about how much they exercise all the time. I’m not sure what your position here is. That people don’t make prosocial lies out of a sense of obligation? People lie about how good each others’ cooking is to avoid hurt feelings. Most explanations of the self-report bias effect start with the most important one: Subjects may make the more socially acceptable answer rather than being truthful. This needs to be addressed. If you were ever wondering why people aren’t donating more money to direct cash transfers, this is reason #1.
There is also the problem of reporting bias and problems with subject follow-up. The people who went to jail/hospital/mental hospital/morgue are a lot less likely to be reporting back. Same with the ones who deteriorate mentally. You need to be accounting for all these problems. In order to get people to take direct cash transfers seriously, you need to do a head-to-head comparison with traditional interventions (housing, food stamps), and with objective measurements (days of school attended, arrest record). I don’t recall this being addressed in the studies I’ve examined, but I haven’t been able to go through many of them.
I can’t think of another way to go about collecting more accurate & less biased information.
This is lethal. EA is fundamentally about evidence-based charity. You’re telling us that you can’t get more objective data, and that the best we can get is self-report survey data. And then you ask for money anyways.
I’m not going to pretend to be OK with this. This is not OK. This is upsetting. This is using bad statistics to get charity money. I’m not going to go so far as to suggest you’re doing this intentionally, but you NEED A STATISTICIAN. A good one. Maybe it’s true that you can’t afford one, but you also can’t afford to not have one.
Here’s a review paper of how alcoholism recovery studies inflate their success rates. I notice a disturbing overlap with your practices.
I have a few questions:
What would you do if all of your investments were stuck in the Chinese market? Puts on semi-conductor-reliant industries?
How much benefit can Chinese firms get from ignoring NVIDIA GPU ML training restrictions?
Would it be possible to post a few sources?
The median academic paper is a good hypothesis away from actually making incremental scientific progress. The top 10% of papers written in such a manner, let’s call it “paper milled”, inspire exploration and follow-up in other scientists. They can become highly cited. They can save other people a LOT of time.
Let me give an example of something I’m working on. My team was looking at Clonal Hematopoiesis of Indeterminate Origin (CHIP), non-disease-causing mutations in cancer-driving genes that appear in white blood cells as people age. They were causing interference in our cancer panels. We were trying to disentangle the effects of selection pressures and mutation hotspots on the appearance of CHIP. How much was due to higher replication rates, and how much was due to the higher likelihood of mutation in specific locations? We were going to do a time-consuming analysis of our data before I found several twin studies, which found no significant difference between identical and fraternal twins (same sex) in terms of CHIP correlation. Turns out, mutation hotspots didn’t play a major role (there are a lot of caveats here, obviously; I’m oversimplifying).
These papers, which were simple analyses done on an existing data set, probably saved us a hundred man-hours. They may have been a formulaic analysis of other peoples’ data, but are still fundamentally good and useful papers that advance science forward. Soon, I’ll be using AI to sort through the enormity of papers to find useful ones like these. I’ll also be using better hypotheses with the aid of AI.
AI is going to matter. It’s already mattering. This is important for the average scientist.
Think about what happens in the dataset of human games where such conversations take place. It probably adds more uncertainty to the predicted actions of players who say these things.
I mean, what would you do if you saw such messages in a game you’re playing? Probably assume they’re mentally unstable and adjust accordingly.
How bad are things, really? I’m not part of EA/Rat/AI risk IRL, so I dont have first-hand experience. Are people actually having mental breakdowns over the control problem? Some of the comments here seem to imply that people are actually experiencing depersonalization and anxiety so bad it’s affecting their work performance specifically because of AI concerns. And not just them, but multiple people they work with. Is the culture at AI alignment orgs really that bad?
Hun, I didn’t know that. I’d expected that fundraisers would be more common in the Western intellectual sphere if that were the case. Can you recommend a charity for those who want to donate? Also, are the ones where they write a custom message on a grenade before it gets dropped on Russians legit?
What about strikes into staging/rear areas though? You can just set grid coordinates a few km deep into enemy territory, minimizing risk of friendly fire. Are there currently a shortage of UAV operators? While I’m sure that for now human control can give higher performance, the idea with automation is massive scaling—instead of a few hundred UAVs operating at a time, each controlled by an operator, you can have tens of thousands of autonomous UAVs operating at the same time, needing human interaction only during resupply.
A full explanation of the replication crisis and how to evaluate scientific papers would be really nice. Most smart science-adjacent (and a lot of scientists!) still believe pop-science conceptions of studies like Dunning-Kreuger and Rat Utopia. Encouraging critical thinking, even of results from high-status scientists, would really improve the smart-normie memespace. It’s pretty crazy how much my understanding of science has changed since I learned how the sausage was made. As things are, there aren’t many ways for someone who isn’t a scientist themselves to learn of the replication crisis and its consequences.
I’d be willing to write the script myself if prizes for the script contest are still being given.