Iâm currently working as an independent research contractor, primarily for the Qualia Research Institute. I previously worked as Chief of Staff at the Institute for Law & AI (formerly âLegal Priorities Projectâ) and as COO at the Center on Long-Term Risk (formerly âEffective Altruism Foundationâ). I also co-founded EA Munich in 2015. I have a masterâs and a PhD in Computational Science from TU Munich and a bachelorâs in Engineering Physics from Tec de Monterrey.
đś 10% Pledger
Am I wrong that EAs working in AI (safety, policy, etc.) and who are now earning really well (easily top 1%) are less likely to donate to charity?
At least in my circles, I get the strong impression that this is the case, which I find kind of baffling (and a bit upsetting, honestly). I have some just-so stories for why this might be the case, but Iâd rather hear othersâ impressions, especially if they contradict mine (I might be falling prey to confirmation bias here since the prior should be that salary correlates positively with likelihood of donating among EAs regardless of sector).
Sharing this talk I gave in London last week titled âThe Heavy Tail of Valence: New Strategies to Quantify and Reduce Extreme Sufferingâ covering aspects of these two EA Forum posts:
I welcome feedback! đ
Oh god thatâs a huge preview. OK.
Thanks for writing this! Coincidentally, my talk âThe Heavy Tail of Valence: New Strategies to Quantify and Reduce Extreme Sufferingâ just went online a couple of hours ago. I thought you might like it âşď¸
Thanks, Hasan! :)
Obviously I want to be cautious about making such a recommendation. đ But Iâm confident enough in our data (pointing at chanca piedra being safer than e.g. melatonin) that I myself take 500mg once a week, but mostly because (a) that was the default recommended dosage on the bottle I got, and (b) I thought taking it daily would be an overkill given that I think Iâm at very low risk. But I could see the rationale for taking it daily for one month per year, as you suggest. Or maybe doing both?
Thanks, Henry. Your care, as a doctor, for peopleâs wellbeing shines through! Weâd also hate to learn that this herb causes e.g. delayed, severe side effects. We hope this question can be settled for good asap given the scope and severity of the suffering involved here!
We start our post acknowledging that online reviews are not usually a reliable source of information concerning medical matters. So we asked ourselves: if there was a statistically significant signal, how could we possibly find it? This is what motivated us to:
Compare reviews across three different platforms (WebMD, Amazon, Reddit).
Compare reviews of chanca piedra to reviews of other kidney stone treatments within those same platforms.
Compare reviews of other popular supplements within those same platforms (e.g., ashwagandha, melatonin).
Do sub-analyses of the data, e.g. looking only at the most credible reviews and reading closely any reviews mentioning any severe or rare side effects (which we report).
Look into whatever literature we could find about chanca piedra to see if mechanistic explanations were compatible with our findings.
Would an analysis of Reddit and Amazon reviews of thalidomide have revealed its negative effects? Bex powder? Diethylstilbestrol? Betel nut?
Not necessarily, no. But if we had access to thousands of anecdotes and reviews of these medications dating back to 2008, including by people who have taken them for years, the chances would go up significantly. I guess thatâs our main âselling pointâ.
In the end, our goal is to reduce as much extreme suffering as possible. If it turns out that taking chanca piedra actually increases extreme suffering, weâd want to be the first ones to know! But from what we can tell, that hypothesis is looking very unlikely.
I really appreciate your feedback and your questions! đ
Iâd love to reply in detail but it would take me a while. đ But maybe two quick points:
On observer independence: The main challenge that computational functionalism faces (IMO) is that thereâs no principled way to say âTHIS is the (observer-independent) system I posit to be consciousâ because algorithms, simulations, etc. donât have clearly-defined boundaries. Itâs up to us (as conscious agents) to arbitrarily determine those boundaries, so anything goes! The section âIs simulation an intrinsic property?â in this post sums it up quite neatly, I think. Field topology, as well as, say, entanglement networks, do give us observer-independent boundaries.
On the simulation behaving identically to the brain: Here I think one could reasonably ask: What if, in order for the simulation to behave identically, we had to simulate the brain even at the smallest physical scale? Many people think this isnât necessary and that the âneuron as digital switchesâ abstraction is enough. But say we actually had to simulate EM field phenomena, quantum phenomena, etc. Then I think runtime complexity matters, since maybe some parts of the brain can be simulated easily and others take millions of years. Can one bootstrap a coherent simulation from that? Now imagine trying to simulate multiple brains interacting with each other, running physics experiments, etc. Can one set up the simulation such that e.g. they all measure the speed of light to be the same? Or otherwise always get the same experimental results? I kind of doubt so. But regardless, the previous point about observer dependence would still stand.
Thanks for your questions, Joseph!
Would it be easier to convince lots of people to buy and consume chanca piedra than to convince lots of people to drink an extra liter of water each day? Any thoughts comparing the relative impact and tractability of these two paths? Iâm particularly thinking of patient compliance and of the fad/âpopularity of large water bottles in recent years.
Weâve also been wondering about that. I drink a lot of water myself but I tend to forget how much of a hassle it can be for a lot of people who are not used to drinking much water. Other conditions face similar trade-offs. For example, experts might insist that exercising 30 minutes per day is among the most helpful interventions against depression, yet many people would understandably much rather take a pill. The same goes for weight loss or other interventions that require lifestyle changes. Urologists have been recommending increasing water intake for decades, yet incidence keeps growing. Intuitively, it feels more tractable to test the effectiveness of chanca piedra through a large RCT and (if it works) have urologists adopt it as part of their standard recommendations than running global awareness campaigns to drink more water. But I admit I havenât done a BOTEC.
Have you spoken with a doctor or medical researcher specializing in kidney stones about this?
Not yet! But, as part of our next steps, we plan to use this writeup to reach out to experts.
Considerations regarding the distribution of pain?
We havenât looked into the distribution in detail (and thereâs not much data, unfortunately), but we currently think that the median case is likely still very painful. For example, the American Urological Association states:
The classic presentation of a renal stone is acute, colicky flank pain radiating to the groin or scrotum, often associated with nausea and vomiting. [âŚ] Renal and ureteral colic are often considered the most severe pain ever experienced by patients, and many female stone patients describe the pain as even more intense than that of childbirth.
In this survey, the mean (!) score of kidney stone pain (6.9/â10, n=239) was only slightly lower than that of labor pain (7.2/â10, n=308) with similar error bars. I think that this is more consistent with a story of the median stone being very painful (than with a story where, say, most stones are only mildly painful). In this (much smaller) survey, 9â93 people mentioned kidney stones to be among their top 3 most painful experiences ever, in line with the ~10% lifetime prevalence. âIn other words, there is reason to believe that a large fraction of the people who have had [kidney stones] will rate them as one of their top 3 most painful experiences.â Neither of these two surveys selected for e.g. patients seeking medical treatment for stones (as is usually the case in the kidney stone literature).
Itâs also worth keeping in mind that people who get kidney stones (including small, asymptomatic ones) are at a higher risk of getting stones later on.
I hope this helps! :)
Thank you for your feedback, Henry! :)
Weâd absolutely love there to be RCTs and definitely encourage larger studies to be run (as mentioned in the Final Recommendations section). It seems like such low-hanging fruit. And one of the main motivations behind this project is that it becomes exactly that incentive needed to get those studies running.
At the same time, it could be years until a large RCT is run. Given the rising incidence of kidney stones worldwide, it seems like asking people to drink more water and eat low-oxalate foods isnât cutting it.
We wrote âCautiously consider taking chanca piedra yourselfâ as our very last recommendation. We wanted to strike a balance between recommending something for which large RCTs are still missing and acknowledging that the suffering involved can be so severe that some action could be warranted (especially given that, as we emphasize, chanca piedra seems to cause fewer side effects than e.g. melatonin and ashwagandha).
Thanks, and sorry, I could have been more precise there. I guess I was thinking of the fact that, for example, some quantum systems would take, I donât know, the age of the universe to compute on a digital computer. And as I hinted in my previous response, the runtime complexity matters. I illustrated this point in a previous post, using the example of an optical setup used to compute Fourier transforms at the speed of light, which you might find interesting. Curious if you have any thoughts!
Thank you so much for your questions! :) Some quick thoughts:
If the human brain operates according to the known laws of physics, then in principle we could simulate it on a modern computer, and it would behave identically to the real thing (i.e. would respond in the same way to the same stimuli, and claim to see a purple ball with grandmaâs face on it if given simulated LSD).
The brain operating according to the known laws of physics doesnât imply we can simulate it on a modern computer (assuming you mean a digital computer). A trivial example is certain quantum phenomena. Digital hardware doesnât cut it. And even if you do manage to simulate certain parts of the system, the only way to get it to behave identically to the real thing is to use the real thing as the substrate. For example, sure, you can crudely simulate the propagation of light on a digital computer, but in order for it to behave identically to the real thing, youâd have to ensure e.g. that all âobserversâ within your simulation measure its propagation speed to be c. I donât believe you can do that given the costs of embodiment of computers.
Would such a brain simulation have qualia according to your view? Yes, no, or you donât think the brain operates according to known laws of physics?
It would be trivial âqualia dust,â like most electromagnetic phenomena (which are not globally bound). (I do think that the brain operates according to the laws of physics.)
If (1) is answered no, what would happen if you gradually replaced a biological brain with a simulated brain bit by bit, replacing sections of the cells one at a time with a machine running a simulation of its counterpart? What would that feel like for the person? Their consciousness would slowly be disappearing but they would not outwardly behave any differently, which seems very odd.
I think âgradually replacing the biological brain with a simulated brain bit by bitâ begs the question. For example, what would it mean to replace a laser beam âbit by bitâ?
If (1) is answered yes, does that mean that whatever this strange property of the EM field is, it will necessarily be possessed by the inner workings of the computer as well, when this simulation is run?
Just to be clear, Iâm not claiming that the EM field has some additional strange property, but rather that the EM field as it is is conscious (cf. dual-aspect monism). Also consider: When you talk about âthe simulation being run,â where exactly is the simulation? In the chips? In sub-elements of the chips? On the computer screen? Simulations, algorithms, etc. donât have clearly-delineated boundaries, unlike our conscious experience. This is a problem.
If all the fields of physics are made of qualia, then everything is made of qualia, including the electron field, the quark fields, etc?
I believe that to be the most parsimonious and consistent view, yes.
@Lucius Caviola has also written about this topic, e.g. Will disagreement about AI rights lead to societal conflict?
And my two cents on why I donât think we should worry about digital sentience (plus the slicing problem). :)
Qualia Research Institute, plus some pro bono time for a project at the intersection of law and AI.
I get what you mean! Here are a few very quick thoughtsâhope they make sense!
Maybe one challenge around introspecting on binding is that one typically has to be in altered or ill states of consciousness to experience what itâs like for binding to break down. For instance, under some psychedelics, the color of an object can âbleed outâ of its boundaries.
In normal waking consciousness, we just take binding for granted. We have no problem recognizing that the eyes, nose, mouth, etc. of the person in front of us are âgluedâ together into a unified gestalt that we can make sense of (and e.g. recognize as our friend). Some people with integrative agnosia canât do that.
As David Pearce says, no binding = no mind. Without binding, there would only be âqualia dustâ floating around. It makes sense that evolution would recruit bound experiences to make sense of the external world. Ethically, we should care about bound experiences.
Binding is also the reason why you canât do your taxes on a high dose of LSD (though maybe altered states of consciousness can be used for other types of computation).
Happy to chat about this more in person one of these days! I just moved to London and plan to visit Oxford every now and then. âşď¸
Thanks for sharing your thoughts!
Iâm unsure whether we can in principle ascertain whether a digital mind is conscious.
I used to believe this too, but Iâve actually become quite optimistic that this is a tractable question. If, as David Pearce argues, the laws of physics donât break down inside the brain, and we define the constraints that a satisfactory theory of consciousness should fulfill, then I believe we can make very good progress. For instance, one such constraint is that the theory should solve the binding problem (see also here). So if we cannot observe a binding mechanism in digital computers, or explain how digital computers could in principle give rise to binding (as I and other at QRI argue), then that should be very strong evidence against digital computers being conscious. More in: Digital Sentience: Can Digital Computers Ever âWake Upâ?
(Note that this doesnât rule out other forms of artificial, non-digital consciousness!)
Thanks for sharing this! You might be interested in this: Consciousness Isnât Substrate-Neutral: From Dancing Qualia & Epiphenomena to Topology & Accelerators
Yes, though I thought maybe among EAs there would be some correlation. đ¤ˇ
Yeah, me neither (which, again, is probably true; just not in my circles).