I just read commoditize your complement and trying to think where this applies. It’s a common observation that bioinformatics software kind of sucks and is underfunded. Have the companies making sequencing “hardware” not learned the lesson from tech companies to commoditize your complement or is something else going on? If Nvidia is trying to switch to bio as an application for AI are they doing funding to make sequencing a commodity? To which extent have Google, Amazon and Facebook succeeded or failed at making AGI labs work commodities?
Morpheus
Karma: 1,555
It’s evidence of a wider array of knowledge they have not unlocked. Learning more biochemistry has put me on firmer ground when trying to evaluate nutrition facts which was an area where before I was epistemically helpless. Learning the basics takes less time than you might think and you waste less time listening to idiots (on for example nutrition). Language models are good at biochemistry, but they will imitate a dumb nutritionist rather than a nutritionist who knows biochemistry if you don’t use the right vocabulary.
Yeah I didn’t want to imply all chronic pain is not serving a purpose. I meant we should figure out good base rates what the underlying issues are by fixing a bunch of cases exhaustively.
The post Why Not Subagents formalizes an argument for why subagents have an incentive to create contracts to effectively become a single agent. The post frames this as bad news. I think it points to a strategy for avoiding needing to build AI. I currently think of it as “becoming a singleton the hard way”, through contracts, mechanism design, financial engineering and regular engineering of the future. The second ingredient is human intelligence amplification via whatever intelligence enhancement techniques get the job done best at the current technology stack. If you are good at preserving logs and brains you can perhaps even delegate computing fair prices almost fully to the future with ample compute available! No conflict needed! This only works if the agents involved are not myopic enough to destroy the universe for a little fun building companies in the short run (which is where this plan sadly probably fails on Earth?). I am very happy for Earth to surprise me though :). I would argue you can view nationalization as a product of this force (see the first chapter of “Seeing Like a State” for details on how France needed to first unify its units to even tax things). I saw Richard Ngo complain somewhere about how the international cognitive elite is becoming disconnected from everyone else and how we need to go back to ethnonationalism. Speculative: I view the same process as the beginning of the next phase transition that could in principle be the beginning of the earth-wide singleton.
Yeah nocebo is more about suggestibility or wrong causal attribution. (Edit: Feedback loops involving your cognition apply in this case as well as for placebo).
I am confused about the placebo and nocebo effect. Often it is “a fake force”. For example when you try to treat a depressed population, you are picking people at their worst often and there will be regression to the mean. The prototypical story from Wikipedia seems weird:
The mechanism for how placebos could have effects is uncertain. From a sociocognitive perspective, intentional placebo response is attributed to the “ritual effect” that induces anticipation for transition to a better state.[67] A placebo presented as a stimulant may trigger an effect on heart rhythm and blood pressure, but when administered as a depressant, the opposite effect.
I can see how the Placebo effect might be a combination of a whole host of different things where people are just bad at causal attribution. Or a feedback loop gets broken where part of the feedback loop was your own nervous system and cognition. Examples: Chronic pain, RSI, panic attacks (if you believe that your heart exploding in your chest is a sign of something really bad rather than just a panic attack you panic more). This post about the vicious cycle of pain is a good example and has reports of people getting healed from understanding what is going on in the comments. Your pain gets healed by just reading a blogpost on the internet!
Fake treatments can help in such cases if they break that feedback loop. For example by lying to children like in the Mindfield episode: “The power of suggestion”. I am quite confused why it would work this effectively though. We have 3 children. One boy has strong migraines every day for months. One girl keeps picking their arms. One boy is hyperactive/has ADHD (he is 7 or at least very young). They go into a sham MRI machine. They get a watch reminding them that their body can heal them. After that the boy is free of migraines every day or some days weak migraines. The girl stopped picking their arms. The evidence from the ADHD child is absolute weaksauce, because how the fuck do you show that in a video about a 7 year old? What was the team thinking before doing the experiment?
Did Mindfield just get lucky+regression to the mean (seems unlinkely in the migraine case) or does someone here have investigated this further and some people are just really good at this suggestion business and we just need to teach them how to stop lying and how to scale their work in order to heal chronic pain at scale? To be clear I am putting a lot of faith in Mindfield here, though that suggestion generally is a skill that can be improved at seems quite obvious. I have also lots of people reporting in person that other sham treatments are fixing them. For example one teenager reporting the pain in their neck goes away after acupuncture for 4-8 weeks (though usually it only fixes it for a finite amount of time!). I have some trust in Mindfield, where their sample sizes are low, but they actually show you stuff with cameras, which gives you a lot of qualitative information you otherwise don’t get from reading a paper and they don’t shy away from results that go against their narrative. Or is the process just inherently time-intensive? Also has someone ever tried to actually exhaustively fix chronic pain in say 100 cases to classify all the different factors establishing the equilibrium and the kicks that got them out of the attractor. Seems like an obvious thing to try. Also obviously this is the type of thing where you want long-term followup.
(Edit: I got a bit distracted by the immediate application. I think understanding better what is going on in a lot of concrete cases in the first place is quite valuable.)
Funnily enough I am thinking about a post proposing such a change. I am not claiming that this would be a good idea for jail or the legal system specifically. But maybe for community disputes where there is an extremely serious accusation where you know either the accuser or the accused has done something serious (accusation vs. false accusation). I still have to think more about it.
Income-Taxes are a hilarious example where they use a function but the derivative is still a bracket because BRACKET! (doesn’t do that much damage in that case I guess though).
Functions are so underused I agree. You can even still publish a bracket as an approximation for the function if people really need the bracket for communicating the rough payoffs.
You can totally be put to jail 90%. You can make a randomized decision with 90% jail probability.
Yup! That is why I rolled disbelief on “psychosomatics” (if you fall of a clif because you belief you can fly and your clan encourages you that is not prototypical psychosomatics). I asked Claude in a somewhat nonleading way (I should have avoided mentioning thirst) and it mentions a paper that the ill are refused water. Valuable in the desert? A way to take revenge on the annoying guy. Who knows.
Aboriginals experienced nocebo effects strong enough to result in death, even from mild injuries, if the weapon causing the injury was believed to be enchanted
If you refuse to eat (or even drink water?) that doesn’t seem so hard to explain?
I don’t have a great personal sense. I think I read somewhere, but don’t remember exactly where that it is ~20% (could have been Tetlocks book). I guess the few times I did calibration training with people is compatible with that number or an even lower number.
I’d be really curious to hear more about this from people who run calibration exercises. I’ve heard of dramatic overconfidence but not underconfidence. I’m sure it exists.
I did Metaculus predictions for a few months and learned that I was in the category of people who need to increase their confidence slightly. I might also just have overcorrected after reading superforcasting. Generally it is hard to be calibrated on really small probabilities or to give 99% confidence intervals for quantities you know little about, so if you do those with calibration training exercises you will almost definitely come up short unless you already know this is hard in advance. Just don’t use probabilities below 1% if you are just getting started with that calibration stuff etc. because reading comprehension and other mistakes get you.
To be fair this is pretty soon after juggling started to get popular with hobbyists and not just professionals in the 1970s. So low-hanging fruit that got plucked as soon as some nerds got sufficiently invested in the hobby?
Walking and biking to places you need to get to sounds like you are already picking the low hanging fruit. I don’t hate sports, but I find most sports (including all you mentioned) boring. I like most sports that involve chasing some ball (like badminton, tennis, squash). I’d maybe try badminton or something like that in your case and find someone you know to play with you or join some club if you are open to that. I assume you would already know if you like any of those, but you didn’t explicitly mention any of that so seemed worth asking. If you don’t like the social aspect have you considered ball juggling? The skill gain can be fun (managing to do more and more catches with more objects is easy to quantify progress) and equipment is cheap and I am sure there are tutorials online. For the same reason I like bouldering because it includes puzzles. (Not sure if I recommend bouldering as a start if you are not already somewhat fit)
I really appreciated Beren’s post to figure out what I think about this for myself. I think another intuition where I and Beren differ is that it is just fine if 1% of enhanced humans you make is a psychopath or whatever (10% is a bit high though. I don’t expect it to be that high unless you allow the data for you GWAS to have a high % of people cheating on tests or misreporting data (which is a valid concern)). The other enhanced humans can deal with the psychopaths! The human psychopaths would like to solve alignment too and are incentivised to cooperate with the other humans. This is not true with AIs because they are so easily changed and copied. The one project that isn’t careful with their AI’s can then spoil it for everyone else when their Pythia overtakes the universe, while the safe projects work on interp and conceptual foundations. With even mildly superhuman AI (we have superhuman hacking now), you would have to be super paranoid that the AI didn’t poison it’s training data or did anything else mischievous. Meanwhile, as a thought experiment I would feel quite fine retiring to childcare and handing off the future to say ~10 clones of myself that have been enhanced in health and intelligence by ~2-4sd (clones would minimize difference from myself, multiple means less noise in the handover, assuming we did this through editing I could even edit different SNPs for every clone, so there is no consistent misalignment between them and me).
I do think some goal shift from intelligence enhancement is possible and we should be able to get some data on this already from looking at existing humans. One intuition I have is that if higher intelligence leads to weird things when it comes to goal misgeneralization, we should see more difference in value that start emerging in humans during puberty compared to in early childhood. I’d be curious if someone has compared this between twins (monozygous twins being more different typical puberty related emotions like parental defiance and their sexuality compared to their relation to anger?) or has checked for consistent trends in puberty being more weird in smarter people compared to early childhood development.
Rephrasing what you are saying to check if I am understanding: Conceptual progress has slowed down, because most research is bottlenecked on ontology. If we made progress on that, we would see more new notations. As an example you bring up the mental disorders, where people are more concerned about the politics of diagnosis, than understanding the underlying reason why “autism” is a thing (or how many distinct things are behind that label). I feel like the sequences actually are pretty great for the ontology stuff? Or at least I can’t think of anything better I’ve read. Like noticing confusion is a great skill with no skill ceiling in sight. The sequence on words taught me a bunch about semantics that seems like it is important for that type of stuff. I am pretty curious what is up with this Barry Smith guy now though, so if you have some specific reading recommendations from him and maybe you can even pitch what particular skill with regards to ontology you feel better at after reading him that would be great.
The first example that came to my mind for a recent notation that has caught on in the field would be siteswaps in juggling. It was only invented in the 1980s. I am a juggler and can confirm that all the technical juggling nerds know what this is and it is used in crazy tricks. For example see 5551 below, which I heard was the first trick that was found through the notation:
Juggling lab is a software for rendering these.
EDIT: I probably misremembered with 5551, the Wikipedia article mentions 441.
For PacBio and ONT I am not sure. Google I guess would want to make sequencing a commodity. For illumina? Polygenic risk scores in healthcare I guess. All the public polygenic risk score predictors suck. They should change that if they want to make that space a commodity. Fund the personal genome project (the project fizzled out with 400K+ volunteers because they don’t have enough funding to sequence people!). I would also personally be interested in that, so partally motivated reasoning. But when I asked Claude to check online it tells me they make more money in healthcare than in research now.