# Mati_Roy's Shortform

This is a special post for short-form writing by Mati_Roy. Only they can create top-level comments. Comments here also appear on the Shortform Page and All Posts page.

## New to LessWrong?

Some comments are truncated due to high volume. Change truncation settings

Litany of Tarski for instrumental rationality 😊

If it's useful to know whether the box contains a diamond,

I desire to figure out whether the box contains a diamond;

If it's not useful to know whether the box contains a diamond,

I desire to not spend time figuring out whether the box contains a diamond;

Let me not become attached to curiosities I may not need.

cars won't replace horses, horses with cars will

There's the epistemic discount rate (ex.: probability of simulation shut down per year) and the value discount (ex.: you do the funner things first, so life is less valuable per year as you become older).

Asking "What value discount rate should be applied" is a category error. "should" statements are about actions done towards values, not about values themselves.

As for "What epistemic discount rate should be applied", it depends on things like "probability of death/extinction per year".

I'm helping Abram Demski with making the graphics for the AI Safety Game (https://www.greaterwrong.com/posts/Nex8EgEJPsn7dvoQB/the-ai-safety-game-updated)

We'll make a version using https://app.wombo.art/. We have generated multiple possible artwork for each card and made a pre-selection, but we would like your input for the final selection.

You can give your input through this survey: https://forms.gle/4d7Y2yv1EEXuMDqU7 Thanks!

In the book Superintelligence, box 8, Nick Bostrom says:

How an AI would be affected by the simulation hypothesis depends on its values. [...] consider an AI that has a more modest final goal, one that could be satisfied with a small amount of resources, such as the goal of receiving some pre-produced cryptographic reward tokens, or the goal of causing the existence of forty-five virtual paperclips. Such an AI should not discount those possible worlds in which it inhabits a simulation. A substantial portion of the AI’s total expected utility might derive

...

## EtA: moved to a question: https://www.lesswrong.com/posts/zAwx3ZTaX7muvfMrL/why-do-we-have-offices

Why do we have offices?

They seem expensive, and not useful for jobs that can apparently be done remotely.

Hypotheses:

• Social presence of other people working: https://www.focusmate.com/
• Accountability
• High bandwidth communication
• Meta communication (knowing who's available to talk to)
• Status quo bias

status: to integrate

7Dagon4y
* Employee focus (having punctuated behaviors separating work from personal time) * Tax advantages for employers to own workspaces and fixtures rather than employees * Not clear that "can be done remotely" is the right metric. We won't know if "can be done as effectively (or more effectively) remotely" is true for some time.
1Mati_Roy4y
thanks for your comment! I just realized I should have used the question feature instead; here it is: https://www.lesswrong.com/posts/zAwx3ZTaX7muvfMrL/why-do-we-have-offices
5Matt Goldenberg4y
Increased sense of relatedness seems a big one missed here.
1Mati_Roy4y
thanks for your comment! I just realized I should have used the question feature instead; here it is: https://www.lesswrong.com/posts/zAwx3ZTaX7muvfMrL/why-do-we-have-offices
3Viliam4y
High status feels better when you are near your subordinates (when you can watch them, randomly disrupt them, etc.). High-status people make the decision whether remote work is allowed or not.
1Mati_Roy4y
thanks for your comment! I just realized I should have used the question feature instead; here it is: https://www.lesswrong.com/posts/zAwx3ZTaX7muvfMrL/why-do-we-have-offices

One model / framing / hypothesis of my preferences is that:

I wouldn't / don't value living in a loop multiple times* because there's nothing new experienced. So even an infinite life in the sense of looping an infinite amount of times has finite value. Actually, it has the same value as the size of the loop: after 1 loop, marginal loops have no value. (Intuition pump: from within the loop, you can’t tell how many times you’ve been going through the loop so far.)

*explanation of a loop: at some point in the future my life becomes indistinguishable from a pre...

4avturchin2y
Agree.  To be long, sufferings need to be diverse. Like maximum suffering is 1111111, but diverse sufferings are: 1111110, 1111101, 1111011, 1110111 etc, so they have time. More diverse sufferings are less intense, and infinite suffering need to be of relatively low intensity. Actually, in the “I have no mouth but I must scream”, the Evil AI invests a lot in making suffering diverse. But diverse is interesting, so not that bad.  Also, any loop in experience will happen subjectively only once. Imagine classical eternal return: no matter how many times I will live my life, my counter of lives will be on 1.
3TLW2y
One more formal method of describing much of this might be the Kolmogorov complexity of the state of your consciousness over the timeframe. (So outputting t=0: state=blah; t=1: state=blah, etc). This has many of the features you are looking for.  This guides me to an interesting question: is looping in an infinite featureless plain of flat white any worse than looping in an infinite featureless plain of random visual noise? (Of course, this is both noncomputable and has a nontrivial chance that the Turing Machine attaining the Kolmogorov complexity is itself simulating you, but meh. Details.)

In my mind, "the expert problem" means the problem of being able to recognize experts without being one, but I don't know where this idea comes from as the results from a Google search don't mention this. What name is used to refer to that problem (in the literature)?

Epistemic status: thinking outloud

The term "weirdness points" puts a different framing on the topic.

I'm thinking maybe I/we should also do this for "recommendation points".

The amount I'm willing to bet depends both on how important it seems to me and how likely I think the other person will appreciate it.

The way I currently try to spend my recommendation points is pretty fat tail, because I see someone's attention as scarce, so I want to keep it for things I think are really important, and the importance I assign to information is pretty fat tail. I'll som...

current intuitions for personal longevity interventions in order of priority (cryo-first for older people): sleep well, lifelogging, research mind-readers, investing to buy therapies in the future, staying home, sign up for cryo, paying a cook / maintain low weight, hiring Wei Dai to research modal immortality, paying a trainer, preserving stem cells, moving near a cryo facility, having some watch you to detect when you die, funding plastination research

EtA: maybe lucid dreaming to remember your dreams; some drugs (becopa?) to improve memory retention)

also not really important in the long run, but sleeping less to experience more

4William Walker4y
NAD+ boosting (NR now, keep an eye on NRH for future). CoQ10, NAC, keep D levels up in winter. Telomerase activation (Centella asiatica, astragalus, eventually synthetics if Sierra Sciences gets its TRAP screen funded again or if the Chinese get tired of waiting on US technology...) NR, C, D, Zinc for SARS-CoV-2 right now, if you're not already. Become billionaire, move out of FDA zone, have some AAV-vector gene modifications... maybe some extra p53 copies, like the Pachyderms? Fund more work on Bowhead Whale comparative genetics. Fund a company to commercially freeze and store transplant organs, to perfect a freezing protocol (I've seen Alcor's...) Main thing we need is a country where it's legal and economically possible to develop and sell anti-agathic technology... even a billionaire can't substitute for the whole market.

topic: AI

Lex Fridman:

I'm doing podcast with Sam Altman (@sama), CEO of OpenAI next week, about GPT-4, ChatGPT, and AI in general. Let me know if you have any questions/topic suggestions.

PS: I'll be in SF area next week. Let me know if there are other folks I should talk to, on and off the mic.

topic: lifelogging as life extension

pivotal acts might require destroying a lot of hardware (ex.: through EMPs); ideally this would be targeted destruction of hardware, but increases my probability that worlds in which lifelogging as life extension is useful are more likely to require EMP-proof lifelogging

#parenting, #schooling, #policies

40% of hockey players selected in top tier leagues are born in the first quarter of the year (compared to 10% in the 3rd quarter) (whereas you'd expect 25% if the time of the year didn't have an influence)

The reason for that is that the cut-off age to join a hockey league as a kid is January 1st; so people born in January are the oldest in their team, and at a young age that makes a bigger difference (in terms of abilities), so they tend to be the best in their team, and so their coaches tend to make them play more and pay ...

That suggest he has no idea about whether it's actually a good vote (as this is how the person differs from other candidates) and just advocates for someone on the basis that the person is his friend.

2[comment deleted]2y

For the Cryonics Institute board elections, I recommend voting for Nicolas Lacombe.

I’ve been friends with Nicolas for over a decade. Ze’s very principled, well organized and hard working. I have high trust in zir, and high confidence ze would be a great addition to the CI's board.

I recommend you cast some or all of your votes for Nicolas (you can cast up to 4 votes total). If you’re signed up with CI, simply email info@cryonics.org with your votes.

4ChristianKl2y
What of those do you think isn't true for the other candidates?
2Mati_Roy2y
afaik, most board members are very passive, and hasnt been doing the things Nicolas wants to do

i want to invest in companies that will increase in value if AI capabilities increases fast / faster than what the market predicts

do you have suggestions?

3Ann3y
Application-specific processing units, health care, agriculture?
4Mati_Roy3y
can you say more about agriculture?
3Ann3y
Agricultural robots exist, and more autonomous versions will benefit from AI in performing tasks currently more dependent on human labor (like careful harvesting) or provide additional abilities like scanning trees to optimize harvest time. Related to whether faster AI progress would give a better price for the market, well, the market may currently be pricing in a relative shortage of human labor, and some of the efforts towards AI robots (in apples for example) have so far gone too slowly to be viable, so going faster than expected might shift the dynamic there.
2Mati_Roy3y
Thanks!

a feature i would like on content website like YouTube and LessWrong is an option to market a video/article as read as a note to self (x-post fb)

4mako yass2y
Youtube lets you access your viewing history through your "library" (or in the web version, probably it's in the sidebar)
2Mati_Roy2y
thanks! yeah i know, but would like if it was more easily accessible whenever i watch a video:)
2mako yass2y
Are you actually looking for the "watch later" feature..
2Mati_Roy2y
No:)
2Mati_Roy3y
oh, LW actually has a bookmark feature, which i could use for that! although i prefer using it for articles i want to read
5bfinn3y
Yes I only recently discovered LW's bookmark - wish I had years ago!

the usual story is that Governments provide public good because Markets can't, but maybe Markets can't because Governments have secured a monopoly on them?

5Matthew Barnett3y
The standard example of a public good is national defense. In that case, you're probably right that the market can't provide it, since it would be viewed in competition with the government military, and therefore would probably be seen as a threat to national security. For other public goods, I'm not sure why the government would have a monopoly. Scientific research is considered a public good, and yet the government doesn't put many restrictions on what types of science you can perform (with some possible exceptions like nuclear weapons research).  Wikipedia lists common examples of public goods. We could through them one by one. I certainly agree that for some of them, your hypothesis holds.
2Viliam3y
For the opposite perspective, see: https://slatestarcodex.com/2017/08/29/my-irb-nightmare/
2ChristianKl3y
I don't think the IRB that blocked Scott in that article was the government.
3Viliam3y
In the "government vs market" dichotomy, I think it is closer to the government side.
2Mati_Roy3y
ah, yeah, you're right! thank you

Suggestion for retroactive prizes: Pay the most undervalued post on the topic for the prize, whenever it was written, assuming the writer is still alive or cryopreserved (given money is probably not worth much for most dead people). "undervalue" meaning amount the post is worth minus amount the writers received.

Topic: Can we compute back the Universe to revive everyone?

Quality / epistemic status: I'm just noting this here for now. Language might be a bit obscure, and I don't have a super robust/formal understanding of this. Extremely speculative.

The map can't be larger than the territory. So you need a larger territory to scan your region of interest: your scanner can't scan themselves...

3avturchin3y
There are several other (weird) ideas to revive everyone:  * send a nanobot via wormhole in the past, it will replicate and collect all data about the brains * find a way to travel between everett branches. For every person there is branch where he is cryopreserved. * use quantum randomness generator to generate every possible mind in different everett branches

GPT-3:

• fixed cost: 4.6M USD
• variable cost: 790 requests/USD source

Human:

• fixed cost: 0-500k USD (depending whether you start from birth and the task they need to be trained)
• variable cost: 10 - 1000 USD / day (depending whether you count their maintenance cost or the cost they charge)

So an AI currently seems more expensive to train, but less expensive to use (as might be obvious for most of you).

Of course, trained humans are better than GPT-3. And this comparison has other limitations. But I still find it interesting.

Acco

...
7gwern3y
All of which was done on much smaller models and GPT-3 just scaled up existing settings/equations - they did their homework. That was the whole point of the scaling papers, to tell you how to train the largest cost-effective model without having to brute force it! I think OA may well have done a single run and people are substantially inflating the cost because they aren't paying any attention to the background research or how the GPT-3 paper pointedly omits any discussion of hyperparameter tuning and implies only one run (eg the dataset contamination issue).
1Mati_Roy3y
Good to know, thanks!

generalising from what a friend proposed me: don't aim at being motivated to do [desirable habit], aim at being addicted (/obsessed) at doing [desirable habit] (ie. having difficulty not to do it). I like this framing; relying on always being motivated feels harder to me

(I like that advice, but it probably doesn't work for everyone)

Philosophical zombies are creatures that are exactly like us, down to the atomic level, except they aren't conscious.

Complete philosophical zombies go further. They too are exactly like us, down to the atomic level, and aren't conscious. But they are also purple spheres (except we see them as if they weren't), they want to maximize paperclips (although they act and think as if they didn't), and they are very intelligent (except they act and think as if they weren't).

I'm just saying this because I find it funny ^^. I think consciousness is harder (for us) to reduce than shapes, preferences, and intelligence.

5Chris_Leong3y
It's actually not hard to find examples of people who are intelligent, but act and think as though they aren't =P

topic: lifelogging as life extension

which formats should we preserve our files in?

I think it should be:

- open source and popular (to increase chances it's still accessible in the future)

- resistant to data degradation: https://en.wikipedia.org/wiki/Data_degradation (thanks to Matthew Barnett for bringing this to my attention)

1Pee Doom3y
Mati, would you be interested in having a friendly and open (anti-)debate on here (as a new post) about the value of open information, both for life extension purposes and else (such as Facebook group moderation)? I really support the idea of lifelogging for various purposes such as life extension but have a strong disagreement with the general stance of universal access to information as more-or-less always being a public good.
1Mati_Roy3y
Meta: This isn't really related to the above comment, so might be better to start a new comment in my shortform next time. Object: I don't want to argue about open information in general for now. I might be open to discussing something more specific and directly actionable, especially if we haven't done so yet and you think it's important. It doesn't currently seem to me that relevant to put one's information public for life extension purposes given you can just backup the information in private, in case you were implying or thinking that. I also don't (at least currently and in the past) advocate for universal access to information in case you were implying or thinking that.
1Mati_Roy3y
note to self, to read: https://briantomasik.com/manual-file-fixity/

topic: lifelogging as life extension

epistemic status: idea

Backup Day. Day where you commit all your data to blu-rays in a secure location.

When could that be?

Aphelion is at the beginning of the year. But maybe would be better to have it on a day that commemorates some relevant events for us.

3avturchin3y
I am now coping a 4 TB HDD and it is taking 50 hours. Blu rays are more time consuming as one need to change the disks, and it may be around 80 disks 50GB each to record the same hard drive. So it could take more than day of work.
3Mati_Roy3y
good point! maybe we should have a 'Backup Week'!:)

I feel like I have slack. I don't need to work much to be able to eat; if I don't work for a day, nothing viscerally bad happens in my immediate surrounding. This allows me to think longer term and take on altruistic projects. But on the other hand, I feel like every movement counts; that there's no loose in the system. Every lost move is costly. A recurrent thought I've had in the past weeks is that: there's no slack in the system.

4Dagon4y
Every move DOES count, and "nothing viscerally bad" doesn't mean it wasn't a lost opportunity for improvement. _on some dimensions_. The problem is that well-being is highly-dimensional, and we only have visibility into a few aspects, and measurements of only a subset of those. It could easily be a huge win on your future capabilities to not work-for-pay that day. The Slack that can be described is not true Slack. Slack is in the mind. Slack is freedom FROM freedom. Slack is the knowledge that every action or inaction changes the course of the future, _AND_ that you control almost none of it. You don't (and probably CAN'T) actually understand all the ways you're affecting your future experiences. Simply give yourself room to try stuff and experience them, without a lot of stress about "causality" or "optimization". But don't fuck it up. True slack is the ability to obsess over plans and executions in order to pick the future you prefer, WHILE YOU SIMULTANEOUSLY know you're wrong about causality and about your own preferences. [ note and warning: my conception of slack has been formed over decades of Subgenius popehood (though I usually consider myself more Discordian), and may diverge significantly from other uses. ]

Today is Schelling Day. You know where and when to meet for the hangout!

epistemic status: a thought I just had

• the concept of 'moral trade' makes sense to me
• but I don't think there's such a thing as 'epistemic trade'
• although maybe agents can "trade" (in that same way) epistemologies and priors, but I don't think they can "trade" evidence

EtA: for those that are not familiar with the concept of moral trade, check out: https://concepts.effectivealtruism.org/concepts/moral-trade/

4Dagon4y
It's worth being clear what you mean by "trade" in these cases. Does "moral trade" mean "compromising one part of your moral beliefs in order to support another part"? or "negotiate with immoral agents to maximize overall moral value" or just "recognize that morals are preferences and all trade is moral trade"? I think I agree that "trade" is the wrong metaphor for models and priors. There is sharing, and two-way sharing is often called "exchange", but that's misleading. For resources, "trade" implies loss of something and gain of something else, where the utility of the things to each party differ in a way that both are better off. For private epistemology (not in the public sphere where what you say may differ from what you believe), there's nothing you give up or trade away for new updates.
3Pattern4y
(I aimed for non-"political" examples, which ended up sounding ridiculous.) Suppose you believed that the color blue is evil, and want there to be less blue things in the world. Suppose I believed the same as you, except for me the color is red. Perhaps we could agree on a moral trade - we will both be against the colors blue and red! Or perhaps something less extreme - you won't make things red unless they'd look really good red, and I won't make things blue unless they'd look really good blue. Or we trade in some other manner - if we were neighbors and our houses were blue and red we might paint them different colors (neither red nor blue), or trade them.
2Dagon4y
Hmm, those examples seem to be just "trade". Agreeing to do something dispreferred, in exchange for someone else doing something they disprefer and you prefer, when all things under consideration are permitted and optional under the relevant moral strictures. I wonder if that implies that politics is one of the main areas where the concept applies.
1Mati_Roy4y

topics: AI, sociology

thought/hypothesis: when tech is able to create brains/bodies as good or better than ours, it will change our perception of ourselves: we won't be in a separate magistra from our tools anymore. maybe people will see humans as less sacred, and value life less. if you're constantly using, modifying, copying, deleting, enslaving AI minds (even AI minds that have a human-like interface), maybe people will become more okay doing that to human minds as well.

(which seems like it would be harmful for the purpose of reducing death)

topic: intellectual discussion, ML tool, AI x-risks

Idea: Have a therapist present during intellectual debate to notice triggers, and help defuse them. Triggers activate a politics mindset where the goal becomes focused on status/self-preservation/appearances/looking smart/making the other person look stupid/etc. which makes it hard to think clearly.

Two people I follow will soon have a debate on AI x-risks which made me think of that. I can't really propose that intervention though because it will likely be perceived and responded as if it was a political m...

Topics: AI, forecasting, privacy

I wonder how much of a signature we leave in our writings. Like, how hard would it be for an AI to be rather confident I wrote this text? (say if it was trained on LessWrong writings, or all public writings, or maybe even private writings) What if I ask someone else to write an idea for me--how helpful is it in obfuscating the source?

Topic: AI strategy (policies, malicious use of AI, AGI misalignment)

Epistemic status: simplistic; simplified line of reasoning; thinking out loud; a proposed frame

A significant "warning shot" from a sovereign misaligned AI doesn't seem likely to me because a human-level (and plausibly a subhuman-level) intelligence can both 1) learn deception, yet 2) can't (generally) do a lot of damage (i.e. perceptible for humanity). So the last "warning shot" before AI learns deception won't be very big (if even really notable at all), and then a misaligned agent would ...

topic: AI alignment, video game | status: idea

Acknowledgement: Inspired from an idea I heard from Eliezer in zir podcast with Lex Friedman and the game Detroit: Become Human.

Video game where you're in an alternate universe where aliens create an artificial intelligence that's a human. The human has various properties typical of AI, such has running way faster than the aliens in that world and being able to duplicate themselves. The goal of the human is to take over the world to stop some atrocity happening in that world. The aliens are trying to stop the human from taking over the world.

✨ topic: AI timelines

Note: I'm not explaining my reasoning in this post, just recording my predictions and sharing how I feel.

I'll sound like a boring cliche at this point, but I just wanted to say it publicly: my AGI timelines have shorten earlier this year.

Without thinking about too much about quantifying my probabilities, I'd say the probabilities that we'll get AGI or AI strong enough to prevent AGI (including through omnicide) are:

• 18% <2033
• 18% 2033-2043
• 18% 2043-2053
• 18% 2050-2070
• 28% 2070+ or won't happen

But at this point I feel like not much...

topic: genetic engineering

'Revolutionary': Scientists create mice with two fathers

topic: genetic engineering

'Revolutionary': Scientists create mice with two fathers

Idea for a line of thinking: What if as a result of automation we could use the ~entire human population to control AI — any way we could meaningfully organize this large workforce towards that goal?

1. What fraction of the cards from the Irrational Game didn't replicate?
2. Is there a set of questions similar to this available online?
3. In the physical game I have, there's a link to http://give-me-a-clue.com/afterdinner (iianm) which is supposed to have 300 more trivia questions, but it doesnt work -- anyone has does?

# Part-time remote assistant position

My assistant agency, Pantask, is looking to hire new remote assistants. We currently work only with effective altruist / LessWrong clients, and are looking to contract people in or adjacent to the network. If you’re interested in referring me people, I’ll give you a 100 USD finder’s fee for any assistant I contract for at least 2 weeks (I’m looking to contract a couple at the moment).

This is a part time gig / sideline. Tasks often include web searches, problem solving over the phone, and google sheet formatting. A full d...

a thought for a prediction aggregator

Problem with prediction market: People with the knowledge might still not want to risk money (plus complexity of markets, risk the market fails, tax implication, etc.).

But if you fully subsidize them, and make it free to participate, but still with potential for reward, then most people would probably make random predictions (at least, for predictions where most people don't have specialized knowledge about this) (because it's not worth investing the time to improve the prediction).

Maybe the best of both worlds is to do...

1Measure2y
Maybe combine this with a public record of who made which predictions (at least shared with the expert group), so they can converge on their own, Aumann style?

But if the brain is allowed to change, then the subject can eventually adapt to the torment. To

This doesn't follow. It seems very likely to me that it can allow it to change it ways it doesn't adapt to the pain, both from first principles and from observations.

How different would each loop need to be in order to be experienced separately?

That would be like having multiple different Everett branches experiencing suffering (parallel lives), which is different from 1 long continuous life.

Superintelligent Devil will take its victim brain's upgrade under its control and will invest in constant development of the victim's brain-parts which can feel pain. There is eternity to evolve in that direction and the victim will know that every next second will be worse. But it is really computationally intense way of punishment.

The question of the minimal unit of experience, which is enough to break the loop sameness is interesting. The need not only to be subjectively different, but the difference need to be meaningful. Not just one pixel.

1MackGopherSena2y
[edited]
2avturchin2y
There is no physical eternity, so there is a small probability of fork in each moment. Therefore, there will be eventually next observer moment, sufficiently different to be recognised as different. In internal experience, it will be almost immediately.

Being immortal means you will one day be a Jupiter brain (if you think memories are part of one's identity, which I think they are)

4avturchin2y
Yes, this follows from that loops can happen only once subjectively, as Peter Hensen recently mentioned to me.
Learning distills memories in models that can be more reasonably bounded even for experience on astronomical timescales. It's not absolutely necessary to keep the exact record of everything. What it takes to avoid value drift is another issue though, this might incur serious overhead. Value drift in people is not necessarily important, might even be desirable, it's only clear for the agent in charge of the world that there should be no value drift. But even that doesn't necessarily make sense if there are no values to work with as a natural ingredient of this framing.
2Mati_Roy2y
Sure! My point still stand though :) Here it's more about identity deterioration than value drift (you could maintain the same value while forgetting all your life). But also, to address your claim in a vacuum: * Value-preservation is an instrumental convergent goal; ie. you're generally more likely to achieve your goals if you want to achieve them. * Plus, I think most humans would value preserving their (fundamental) values intrinsically as well. Am not sure I understand
The arguments for instrumental convergence don't apply to the smaller processes that take place within a world fully controlled by an all-powerful agent, because the agent can break Moloch's back. If the agent doesn't want undue resource acquisition to be useful for you, it won't be, and so on. The expectation that humans would value preservation of values is shaky, it's mostly based on the instrumental convergence argument, that doesn't apply in this setting. So it might actually turn out that human preference says that value preservation is not good for individual people, that value drift in people is desirable. Absence of value drift is still an instrumental goal for the agent in charge of the world that works for the human preference that doesn't drift. This agent can then ensure that the overall shape of value drift in the people who live in the world is as it should be, that it doesn't descend into madness. Value drift only makes sense where the abstraction of values makes sense. Does my apartment building have a data integrity problem, does it fail some hash checks? This doesn't make sense, the apartment building is not a digital data structure. I think it's plausible that some AGIs of the non-world-eating variety lack anything that counts as their preference, they are not agents. In a world dominated by such AGIs some people would still set up smaller agents merely for the purpose of their own preference management (this is the overhead I alluded to in the previous comment). But for those who don't and end up undergoing unchecked value drift (with no agents to keep it in line with what values-on-reflection approve of), the concept of values is not necessarily important either. This too might be the superior alternative, more emphasis on living long reflection than on being manipulated into following its conclusions.

Here's a way to measure (a proxy of) relative value of different years I just thought (again?); answer the question:

For which income I would you prefer to live in perpetual-2020 over living in perpetual-2021 with a median income? (maybe income is measured by fraction of the world owned multiplied by population size, or some other way) Then you can either chain those answers to go back to older years, or just compare them directly.

There are probably years where even Owning Everything wouldn't be enough. I prefer to live in perpetual-2021 with a median incom...

Hobby: serve so many bullets to sophisticated philosophers that they're missing half their teeth by the end of the discussion

crazy idea i just had mayyybe a deontological Libertarian* AI with (otherwise) any utility function is not that bad (?) maybe that should be one of the thing we try (??????) *where negative externality also count as aggressions, and other such fixes to naive libertarianism

5Viliam2y
Existence of an agent is itself already a negative externality for other agents existing in the same environment. It means more competition for the limited natural resources. Even animals, pure in their minds and untainted by the sin of statism, fight for their territory and for the natural resources it includes. Of course, competing for natural resources is not the only thing an agent does. If the agent also produces and trades, this aspect of its existence is a positive externality for its neighbors, which in modern economy typically outweighs the increased competition for raw resources. But if the economical context changes, this could change, too. If the libertarian AI in a libertarian world succeeds to send von Neumann probes to colonize planets, using them to build more von Neumann probes and colonize more planets... and takes over the entire universe, leaving only Earth, Moon, and possibly Mars to humans... according to libertarianism, that's perfectly fair, right? I mean, the humans were not actually using the rest of the universe before, so it was free to take, the AI started using it productively (to build more von Neumann probes), therefore it rightfully belongs to the AI now. And if the AI refuses to trade with humans, or generally to interact with them in any way other than destroying all human spaceships that trespass on its territory (i.e. the whole universe, other than Earth, Moon, Mars and the space between them), it is still acting fully within its libertarian rights. It is not initiating violence, merely protecting its rightful property. The parts of universe that are not needed for defense against humans, are converted to paperclips... Would you count this outcome as "not that bad"?
2Mati_Roy2y
before the birth of that AI, we could split the Universe among existing beings

Am thinking of organizing a one hour livestreamed Q&A about how to sign up for cryonics on January 12th (Bedford's day). Would anyone be interested in asking me questions?

We sometimes encode the territory on context-dependent maps. To take a classic example:

• when thinking about daily experience, stars and the Sun are stored as different things
• when thinking in terms of astrophysics, they are part of the same category This makes it so that when you ask a question like "What is the closest star [to us]?", in my experience people are likely to say Alpha Centauri, and not the Sun. Merging those 2 maps feels enlightening in some ways; creates new connections / a new perspective. "Our Sun is just a star; stars are just like the
...
2Mati_Roy3y
1. No name that I'm aware. Brainstorming ideas: map merging, compartmentalisation merging, uncompartmentalising

In my mind, "the expert problem" means the problem of being able to recognize experts without being one, but I don't know where this idea comes from as the results from a Google search don't mention this. What name is used to refer to that problem (in the literature)?

[This comment is no longer endorsed by its author]Reply
4Pattern3y
This comment/post is one of 3 duplicates. (Link to main here.)
2Mati_Roy3y
oh damn, thanks! there was an error message when I was trying to post it which had given me the impression it wasn't working, hence why I posted it 4 times total ^^

In my mind, "the expert problem" means the problem of being able to recognize experts without being one, but I don't know where this idea comes from as the results from a Google search don't mention this. What name is used to refer to that problem (in the literature)?

[This comment is no longer endorsed by its author]Reply
0Pattern3y
This comment/post is the 2nd of 3 duplicates. (Link to main here.)

In my mind, "the expert problem" means the problem of being able to recognize experts without being one, but I don't know where this idea comes from as the results from a Google search don't mention this. What name is used to refer to that problem (in the literature)?

[This comment is no longer endorsed by its author]Reply
0Pattern3y
This comment/post is the 3rd of 3 duplicates. (Link to main here.)

suggestion of something to try at a LessWrong online Meetup:

video chat with a time-budget for each participant. each time a participant unmutes themselves, their time-budget starts decreasing.

note: on jitsi you can see how many minutes someone talked (h/t Nicolas Lacombe)

imagine having a physical window that allowed you to look directly in the past (but people in the past wouldn't see you / the window). that would be amazing, right? well, that's what videos are. with the window it feels like it's happening now, whereas with videos it feels like it's happening in the past, but it's the same

6Dagon3y
There's a bit of bait-and-switch with the comparison.  The magic window is amazing if it sees parts of the past which we're interested in (both time and location control, or event selection).  It's much less interesting if it only sees very few parts (well under 0.01%) of the very recent past (last 20-70 years), and only sees the parts that someone happened to capture, which are indexed/promoted enough to come to our attention.
2Mati_Roy3y
ok yeah, that's fair! (although even controlling for that, I think the analogy still points at something interesting) yeah, I like to see "people just living a normal day"; I sometimes look for that, but even that is likely biased

tattoo idea: I won't die in this body

in Toki Pona: ale pini mi li ala insa e sijelo ni

direct translation: life's end (that is) mine (will) not (be) inside body this

When you're sufficiently curious, everything feels like a rabbit hole.

Challenge me by saying a very banal statement ^_^

2Matt Goldenberg3y
I  tired because I didn't sleep well.
2Mati_Roy3y
Sort of smashing both of those saying together: > “If you wish to make an apple pie from scratch, you must first invent the universe.” -Carl Sagan > "Any sufficiently analyzed magic is indistinguishable from science!"-spin of Clarke's third law to get: Sufficiently understanding an apple pie is indistinguishable from understanding the world.
1Mark Xu3y
reminded me of Uriel explaining Kabbalah: > “THEY BELIEVE YOU CAN CARVE UP THE DIFFERENT FEATURES OF THE UNIVERSE, ENTIRELY UNLIKE CARVING A FISH,” the angel corrected himself. “BUT IN FACT EVERY PART OF THE BLUEPRINT IS CONTAINED IN EVERY OBJECT AS WELL AS IN THE ENTIRETY OF THE UNIVERSE. THINK OF IT AS A FRACTAL, IN WHICH EVERY PART CONTAINS THE WHOLE. IT MAY BE TRANSFORMED ALMOST BEYOND RECOGNITION. BUT THE WHOLE IS THERE. THUS, STUDYING ANY OBJECT GIVES US CERTAIN DOMAIN-GENERAL KNOWLEDGE WHICH APPLIES TO EVERY OTHER OBJECT. HOWEVER, BECAUSE ADAM KADMON IS ARRANGED IN A WAY DRAMATICALLY DIFFERENTLY FROM HOW OUR OWN MINDS ARRANGE INFORMATION, THIS KNOWLEDGE IS FIENDISHLY DIFFICULT TO DETECT AND APPLY. YOU MUST FIRST CUT THROUGH THE THICK SKIN OF CONTINGENT APPEARANCES BEFORE REACHING THE HEART OF -”
1Mati_Roy3y
Nice!

I can pretty much only think of good reasons for having generally pro-entrapment laws. Not any kind of traps, but some kind of traps seem robustly good. Ex.: I'd put traps for situations that are likely to happen in real life, and that show unambiguous criminal intent.

It seems like a cheap and effective way to deter crimes and identify people at risk of criminal behaviors.

6Viliam3y
Aren't there already too many people in prisons? Do we need to put there also people who normally wouldn't have done any crime? I guess this depends a lot on your model of crime. If your model is something like -- "some people are inherently criminal, but most are inherently non-criminal; the latter would never commit a crime, and the former will do use every opportunity that seems profitable to them" -- then the honeypot strategy makes sense. You find and eliminate the inherently criminal people before they get an opportunity to actually hurt someone. My model is that most people could be navigated to commit a crime, if someone would spend the energy to understand them and create the proper temptation. Especially when we consider the vast range of things that are considered crimes, so it does not have to be a murder, but something like smoking weed, or even things that you have no idea they could be illegal; then I'd say a potential success rate is 99%. But even if we limit ourselves to the motte of crime; let's say theft and fraud, I'd still say more than 50% of people could be tempted, if someone spent enough resources on it. Of course some people are easier to nudge than others, but we are all on the spectrum. Emotionally speaking, "entrapment" feels to me like "it is too dangerous to fight the real criminals, let's get some innocent but stupid person into trouble instead, and it will look the same in our crime-fighting statistics".
1Mati_Roy3y
uh, I didn't say anything about prisons. there are reasons to identify people at high risk of committing crimes. and no, it's not about catching people that wouldn't have committed crimes, it's about catching people that would have committed crimes without being caught (but maybe I misused the word 'entrapment', and that's not what it means) Well, that's (obviously?) not what I mean. I elaborated more on the Facebook post linked above.
2Viliam3y
I agree, but that seems to be how the idea is actually used in real life. By people other than you. By people who get paid when they catch criminals... which creates an incentive for them to increase easy-to-solve criminality rather then reduce it, as long as they find plausibly deniable methods to do it. In theory, if you could create "traps" in a way that does not increase temptation (because increased temptation = increased crime), for example on a street already containing hundred unlocked bikes you would add dozen unlocked trap bikes... yeah, there is probably no downside to that. In practice, if you allow this, and if you start rewarding people for catching thieves using the traps, they will get creative. Because a trap that follows the spirit of the law does not maximize the reward.
6Dagon3y
I think the setup you describe (unambiguously show criminal intent in likely situations) is _already_ allowed in most jurisdictions. "entrapment" implies setting up the situation in such a way that it encourages the criminal behavior, rather than just revealing it.
4Gordon Seidoh Worley3y
IANAL, but this sounds right to me. It's fine if, say, the police hide out at a shop that is tempting and easy to rob and encourage the owner not to make their shop less tempting or easy to rob so that it can function as a "honeypot" that lets them nab people in the act of committing crimes. On the other hand, although the courts often decide that it's not entrapment, undercover cops soliciting prostitutes or illegal drugs are much closer to being entrapment, because then the police are actively creating the demand for crime to supply. Depending on how you feel about it, I'd say this suggests the main flaw in your idea, which is that it will be abused on the margin to catch people who otherwise would not have committed crimes, even if you try to circumscribe it such that the traps you can create are far from causing more marginal crime, because incentives will push for expansion of this power. At least, that would be the case in the US, because it already is.

People say we can't bet about the apocalypse. But what about taking debt? The person thinking the probability of apocalypse is higher would accept higher interest rate on their debt, as once at the judgement period their might be no one to whom the money is worth or the money itself might not be worth much.

I guess there are also reasons to want more money during a global catastrophe, and there are also reasons to not want to keep money for great futures (see: https://matiroy.com/writings/Consume-now-or-later.html), so that wouldn't actually work.

meta - LessWrong have people predict whether they will upvote a post just based on the title

1[anonymous]4y
[Deleted]
1Mark Xu4y
I think this breaks because it results in people upvoting based on the title. I recall some study about how people did things they predicted they would do with higher than like 60% chance almost 95% of the time (numbers made up, think I remember direction/order of magnitude of effect size roughly correctly, don't know if it survived the replication crises)
1Mati_Roy4y
That's potentially a good point! But it doesn't say how the causality works. Maybe the prediction affects the outcome or maybe they're just bad at predicting / modelling themselves.

There's a post, I think by Robin Hanson on Overcoming Bias, that says people care about what their peers think of them, but we can hack our brains to doing awesome things by making this reference group the elite of the future. I can't find this post. Do you have a link?

Personal Wiki

might be useful for people to have personal wiki where they take note instead of everyone taking notes in private Gdoc

status: to do / to integrate

you know those lists about historical examples of notable people mistakenly saying that some tech will not be useful (for example)

Elon Musk saying that VR is just a TV on your nose will probably become one of those ^^

4mako yass5mo
This wasn't him taking a stance. It ends with a question, and it's not a rhetorical question, he doesn't have a formed stance. Putting him in a position where he feels the need to defend a thought he just shat out about a topic he doesn't care about while drinking a beer is very bad discourse.
2Mati_Roy5mo
ok that's fair yeah! thanks for your reply. I'm guessing a lot of those historical quotes are also taking out of context actually.
2niplav5mo
I believe VR/AR are not going to be as big of a deal as smartphones85%, and not produce >\$200 bio. of revenue in 203055%.
1Sinclair Chen5mo
earbuds are just speakers in your ears. they're also way better than speakers.
1Sinclair Chen5mo
true virtual reality requires not just speakers in your ears and tv in your eyes, but also good input at the speed of thought. maybe it's just eye, face, hand, and limb tracking. but idk, i feel like headsets have still not found their mouse and keyboard. the input is so ... low bandwidth? maybe Apple will figure it out, or maybe we need to revive Xerox back from the grave

idea: Stream all of humanity's information through the cosmos in hope an alien civ reconstruct us (and defends us against an Earth-originating maligned ASI)

I guess finding intelligent ETs would help with that as we could stream in a specific direction instead of having to broadcast the signal broadly

It could be that maligned alien ASIs would mostly ignore our information (or at least not use it to like torture us) whereas friendly align ASI would use it beneficially 🤷‍♀️

2Mati_Roy5mo
related concept: https://en.wikipedia.org/wiki/Information_panspermia video on this that was posted ~15 hours ago: https://www.youtube.com/watch?v=K4Zghdqvxt4

Topics: cause prioritization; metaphor

note I took on 2022-08-01; I don't remember what I had in mind, but I feel like it can apply to various things

from an utilitarian point of view though, i think this is almost like arguing whether dying with a red or blue shirt is better; while there might be an answer, i think it's missing the point, and we should focus on reducing risks of astronomical disasters

An interesting perspective.

It is instructive to consider the following four scenarios:

1. The Kolmogorov complexity of the state of your mind after N timesteps in a simulation with a featureless white plane.
2. The Kolmogorov complexity of the state of your mind after N timesteps in a simulation where you are in a featureless plane, but the simulation injects a single randomly-chosen 8x8 black-and-white bitmap into the corner of your visual field. (256 bits total.)
3. The Kolmogorov complexity of the state of your mind after N timesteps in a simulation with "...

If crypto makes the USD go to 0, will life insurances denominated in USD not have anything to pay out? Maybe an extra reason for cryonicists to own some crypto

A Hubble Brain: a brain taking all the resources present in a Hubble-Bubble-equivalent.

5avturchin3y
But it will take 15 billion years for it to finish one thought
3Mati_Roy3y
haha, true ^^

I want to look into roleplay in animals, but Google is giving me animal roleplay, which is interesting too, but not what I'm looking for right now 😅

I'm wonder how much roleplay there is in the animal kingdom. I wouldn't be surprised if there was very few.

Maybe if you're able to roleplay, then you're able to communicate?? Like, roleplay might need to have a theory of mind, because you're imagining yourself in someone else's body.

Maybe you can teach words to an animal without a theory of mind, but they'll be more like levers for them: for them, saying "bana...

I remember someone in the LessWrong community (I think Eliezer Yudkowsky, but maybe Robin Hanson or someone else, or maybe only Rationalist-adjacent; maybe an article or a podcast) saying that people believing in "UFOs" (or people believing in unproven theories of conspiracy) would stop being so enthusiastic about those if they became actually known as true with good evidence for them. does anyone know what I'm referring to?

4Ruby3y
Eliezer talks about how dragons wouldn't be exciting if they were real, I recall. I'm not sure that's correct.
1Mati_Roy3y
ah, someone found it:

sometimes I see people say "(doesn't) believe in science" when in fact they should say "(doesn't) believe in scientists"

or actually actually "relative credence in the institutions trying to science"

hummm, I think I prefer the expression 'skinsuit' to 'meatbag'. feels more accurate, but am not sure. what do you think?

I just realized my System 1 was probably anticipating our ascension to the stars to start in something like 75-500 years.

But actually, colonizing the stars could be millions of subjective years away if we go through an em phase (http://ageofem.com/). On the other hand, we could also have finished spreading across the cosmos in only a few subjective decades if I get cryopreserved and the aestivation hypothesis is true (https://en.wikipedia.org/wiki/Aestivation_hypothesis).

I created a Facebook group to discuss moral philosophies that value life in and of itself: https://www.facebook.com/groups/1775473172622222/

# How to calculate subjective years of life?

For non-human animal brains, I would compare them to the baseline of individuals in the...

4avturchin3y
Interestingly, an hour in childhood is subjectively equal between a day or a week in adulthood, according to recent poll I made. As a result, the middle of human life in term of subjective experiences is somewhere in teenage. Also, experiences of an adult are more dull and similar to each other. Tin Urban tweeted recently: "Was just talking to my 94-year-old grandmother and I was saying something about how it would be cool if I could be 94 one day, a really long time from now. And she cut me off and said “it’s tomorrow.” The "years go faster as you age" phenomenon is my least favorite phenomenon."

The original Turing test has a human evaluator.

Other evaluators I think would be interesting include: the AI passing the test, a superintelligent AI, and an omniscient maximally-intelligent entity (except without the answer to the test).

# Blocking one ear canal

Category: Weird life optimization

One of my ear canal is in a different shape. When I was young, my mother would tell me that this one was harder to clean, and that ze couldn't see my ear-drum. This ear gets wax accumulation more easily. A few months ago, I decided to let it block.

Obvious possible cognitive bias is the "just world bias": if something bad happens often enough, I'll start think it's good.

But here are benefits this has for me:

1. When sleeping, I can put my good ear on the pillow, and this now isolates me from sound prett

...
3Mati_Roy4y
x-posting someone's comment from my wall:

topic: fundamental physics