Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[silverspawn]

I mean probably not

[faust]

Yet for some reason, [Bill Gates is] still doing global poverty stuff instead of the thing that actually matters.

I take enjoyment in the fact that I can interpret this sentence in a way that was certainly not intended.

[silverspawn]

Finally continued the vaccine dialogue with my sister. Still put quite a bit of effort into it.

Here is what Paul Christiano (something like a level 100 titan in the AI safety sphere and used to be an important figure in OpenAI, probably before it scaled back its safety team) wrote in response to the article that I believe people gushed over in the second video my sister sent to me this time

I read the start of Kirsch's article here since I'm slightly interested in vaccine skepticism and have never dug into it before. (I'm not sure if this is the article that Bret is gushing about.)

The first claim in the article is that there are many more deaths reported by VAERS. I ignored this because without some other context it seemed unsurprising and uninformative that VAERS is used more for the covid vaccine than for the flu vaccine (and the absolute numbers of deaths reported are negligible given the number of people who have taken the vaccine). The inference of "at least 20,000 deaths due to the vaccine" looks like it's probably bullshit but it would take a few minutes to establish that and so I moved on.

The next claim about harms was "82% miscarriage rate in first 20 weeks", which links here. But this number seems to be computed as "Amongst pregnancies that ended, how many were miscarriages?" The table was published in April 21 about people who were vaccinated December 14 - February 28, and it looks like there shouldn't have bee enough time for anyone to have a healthy live birth if they were vaccinated in the first 20 weeks.

The authors of the letter-to-the-editor acknowledge this as saying "We acknowledge this rate will likely decrease as the pregnancies of women who were vaccinated <20 weeks complete but believe the rate will be higher than 12.5%," but it's unclear why they believe this (the study reports 96 spontaneous abortions amongst something like 1000 people who were vaccinated in the first trimester, which seems like it may be right on track).  And of course that wouldn't make the 82% number less dubious.

So as far as I can tell this datapoint is very misleading, and this should have been obvious to anyone who took a cursory look before including it in an article. It's also such an extreme and bizarre claim that it should certainly have warranted a cursory look.

I didn't look into any of the other claims because it would probably be more valuable to instead read some more credible vaccine skeptic.

The reason why this sticks out to me is because it's probably roughly what a fair assessment of the evidence looks like. Which is translates to "no reason to spend any time on this".

Which makes my response way way kind and patient. Basically I'm taking it for granted that the whole thing deserves thorough investigation, when in reality there was probably never good reason to take it seriously in the first place, or if so never good reason to keep taking seriously after the first two claims don't check out.

The miscarriage rate was a repeat; this claim is already made in the first video, where I picked it apart in detail.

Anyway I was being a bit more aggressive this time, though still very cordial. Let's see what happens.

[silverspawn]

Woah just saw that Paul Christiano did something Dominion related. Cool.

i don't understand this on first glance and wanted to go back to work now, but I'll just leave it here anyway.

[silverspawn]

something really important that I'm too much of a wuss to say is that                                                           ,                                                    is total bullshit. Any                       where                                                                   shows that                                                                                          .

[silverspawn]

I really like how there is no luck in chess

[Awaclus]

I really like how there is no luck in chess

There's also no luck in Prismata.

[silverspawn]

Review of Don't Look Up from one of my favorite people

[silverspawn]

Using pie charts as the default is so dumb

Bar charts are superior in almost every way. Basically the only advantage of pie charts is that they have an in-built visualization for sums of adjacent responses. If this is really important, and if there aren't too many responses, then they're fine. I recommend them for election results. But in most cases, just use a flippin bar chart

[silverspawn]

[silverspawn]

The One True Principle That Is Responsible For Every Philosophical Confusion Ever

is thinking you're special. Or to be more precise, not treating yourself as a perfectly normal part of the environment

[silverspawn]

1. Decision Theory

The academically dominant decision theory is causal decision theory (CDT). According to CDT, you should choose an action such that the expected outcome when doing that action is maximized. The keyword here is doing; you do not treat the fact that you did that action as a thing you learned. This is why, e.g., if there is a gum such that chewing it makes you mildly less likely to have cancer, but chewing the gum is highly correlated with having cancer because both are caused by the same gene, CDT says to go ahead and chew.

Then comes  Newcomb's problem. Omega (hypothetical superintelligence for thought experiments) presents you with two boxes. One is transparent and contains 1000$. One is opaque and contains either 1000000$ or 0$. Your choices are (a) take both boxes or (b) take just the opaque box. However, before you have joined the scene, Omega has predicted your decision and put 1000000$ into the box if and only if it predicted you one-box.

CDT argues "regardless of what's in the opaque box, the choice has already been made; no matter what it was, taking both boxes yields 1000$ more than taking just one; I'll take both" and thus goes home with 1000$.

The mistake: treating yourself as special. Everything else is modeled accurately, but your own actions are treated as magic. It pretends as if you are on the outside looking in, and get to choose freely without interacting with anything. The previous sentence is not exaggerated; this is exactly what the math does.

[silverspawn]

2. Anthropic Probability Theory

Anthropic probability theory is probability theory where the number of observers is weird in some way.

Sleeping Beauty Problem: Sleeping Beauty is recruited for an experiment and put to sleep -- go figure. A coin is flipped. If the coin lands heads, Sleeping Beauty is woken up and interviewed once on Friday. If the coin lands tails, Sleeping Beauty is interviewed twice, once on Thursday and once on Friday.

Sleeping Beauty wakes up and is being interviewed. What is the probability that the coin came up heads?

"it's 50%", argues the very serious professional philosopher. "being interviewed was inevitable, we cannot conclude anything about the coin".

The mistake: treating yourself as special. If you looked at the setup from a bird's eye view and the experiment was run 2000 times, you would see 1000 interviews with the coin on heads and 2000 with the coin on tails. The answer is 1/3.

[silverspawn]

3. Quantum Physics

"According to your collected data," says the experiment, "the wave function propagates like this".

"But we only see one world," says the very mature serious and humble senior physician who is very serious and humble and not arrogant, "therefore everything except a single point of the wavefunction spontaneously disappears as the only non-continuous and non-deterministic and non-local... [refer to the list of absurdities I've quoted earlier; I don't know physics well enough to recite it] phenomenon in all of physics".

The mistake: treating yourself as special. Of course the wavefunction doesn't bloody disappear just because you don't see it.

[silverspawn]

4. Phenomenological Identity

"I expect to wake up as myself tomorrow," says every person ever, "and every other day for the rest of my life. If someone with my exact genes is born, they are different from me. But if I completely change over 50 years, it's still me. This is all very obvious and normal. If I'm cloned, then it's not me even though the other entity is physically identical. Unless I die first, in which case I have to admit it is me since dying and t hen being cloned is a well-known concept used in science fiction for teleportation, also it's basically what happens in every normal moment anyway. So it somehow depends on the order in which (dying, being cloned) happens because that totally sounds like a thing that could be true. Also it several clones are created or the experiment is tweaked in a bunch of other ways, my position necessarily becomes incoherent or nonsensical. The above is obviously the only correct theory and is obviously true, obviously."

"Maybe you're identical to everyone?" says the Open Individualist.

"LOL don't be ridiculous," says every person ever. "Sounds like spiritual Buddhist woo. Have you taken LSD? Normal very serious people don't believe in such hippy things".

The mistake: treating yourself as special. No, there aren't any irreducible continuous entities in the universe that remember who something happens to. This contradicts reductionism, there is precisely zero room for something like this in physics, it doesn't survive thought experiments, and there is zero evidencefor it. Qualia having identity is no more coherent than matter having identity.

[silverspawn]

It's much more fun to write all this without having to worry about being persuasive or not offensive

[silverspawn]

5. Reinforcement Learning

The standard mathematical model for Reinforcement Learning is AIXI. The agent keeps a probability distribution over possible states of the environment and has list of possible actions. At every time step, it receives an observation. It updates its probability distribution, and chooses the action that, weighted by this distribution, has the highest expected return.

Then the agent learns to self-modify and the model implodes into a logical black hole of infinite recursion.

The mistake: treating yourself as special. No, the agent is not outside of the environment looking it. No, the agent cannot have a probability distribution over all possible models of the environment that includes the correct one. Trivially impossible because the correct model of the environment includes the agent itself, which includes the correct model of the environment, which includes the agent itself, which includes the correct model of the environment, which includes the agent itself, which is an infinite recursion.

The agent is part of the environment like everything else, made out of smaller parts like everything else, and can be modified like everything else.

I admittedly don't know much about this (even possible I got AIXI wrong somehow), but apparently this is the the fundamental obstacle when trying to develop a theory of intelligence that permits self-modification.

[silverspawn]

USB Sticks

"Other teachers forget their USB sticks in school," says me. "I do not because I am special."

The mistake: NOT REMOVING THE USB STICK BEFORE GOING HOME. I am special anyway, of course.

[silverspawn]

As evidenced by the fact that I made a backup because I did, in fact, anticipate losing it

[silverspawn]

Look at all those great red words like beautiful crowns and funny kangaroos and magical genies next to those terrible blue words like glass you will bump into or loud ambulances that mean someone is sick or apples that have grown too big and mealy.

[silverspawn]

5. Free Will

Why did the Rock fall down?

Bryan Caplan: Because it was pushed.

Why did the hurricane hit our town?

Bryan Caplan: Because of weather conditions and climate change.

Why did the wolf eat me?

Bryan Caplan: Because it was hungry.

Why did Hitler kill babies?

Bryan Caplan: he was FREE to choose so through the only non-deterministic and non-random process in the universe. When particles come together in just the right way, they form a single indivisible ontologically fundamental entity with FREEDOM that is FREE to do whatever it pleases. This cannot be formalized and hasn't been observed anywhere else in the universe. Also it contradicts determinism but I simply deny that determinism exists. Also Hitler didn't just kill babies, he did more bad things than that, which is why I would choose to torture him if we were stranded on an Island together and nothing good could come of this ever.*

The mistake: treating yourself as special; this one is rather obvious.

*Point he actually brought up out of his own FREE WILL and without being asked about it. I believe he made it as a point about why FREE WILL matters.

[silverspawn]

Flip this was number 6

[Awaclus]

Do I need to consider myself special to think that some extent of free will exists? I believe the wolf could probably choose not to eat me even though it's hungry, and while I also believe that humans have a higher level of free will than any other known species, it's definitely still limited.

[silverspawn]

No, not yourself specifically. This actually goes for most of the others as well. Read "you" more like "humans".

To be clear though, the position I was referencing in #6 is explicitly libertarian free will that exists outside of determinism, not some kind of Daniel-Dennet-ian compatibilism that locates free will as "degrees of control" while acknowledging determinism

[faust]

No, not yourself specifically. This actually goes for most of the others as well. Read "you" more like "humans".

To be clear though, the position I was referencing in #6 is explicitly libertarian free will that exists outside of determinism, not some kind of Daniel-Dennet-ian compatibilism that locates free will as "degrees of control" while acknowledging determinism

It is however conceivable to talk about free without regarding humans as special. All you have to do is extend your notion. Not just humans have free will; the wolf might have chosen not to eat you. Not just animals have free will: The rock could have chosen not to fall down.

In this view, laws of physics are just predictions of the highest-likelihood outcome, similar to laws of economics or sociology. Gravity is a good law because we can observe that most matter has a preference towards falling to the ground.