@benmmurphy's banner p




0 followers   follows 0 users  
joined 2022 September 06 20:04:30 UTC


User ID: 881



0 followers   follows 0 users   joined 2022 September 06 20:04:30 UTC


No bio...


User ID: 881

Yud seemed to say LLMs could play chess and therefore could reason. However, the games I've see it play it has tried to make illegal moves which seems to indicate its just pattern matching and the pattern matching breaks down in some spots. of course maybe reasoning is just pattern matching and the LLMs aren't good at it yet or the LLM hasn't been trained on enough chess games. i guess chess players would also say chess is heavily about pattern matching but it also involves some kind of explicit reasoning to double check lines.

did the suicide nets lower suicides or did people just use another method? i would expect they did lower them somewhat because presumably it increases the cost of suicide at the margin. but i guess if there are good substitute options then it might not have much of an effect.

we are already having people doing coding interviews with us and using chat-gpt to generate the solutions

this is actually true. i know there is a third party framework that is installed in a lot of popular apps in iOS that could be used to deliver zero day exploits targeted at individual users. one of the co-founders of the company had a senior position in the DoD. the only reason i don't think this is such a big deal is because anyone who has access to such exploits could probably just find/buy exploits for safari in order to deliver it so the infrastructure is not that useful. also, i'm not sure how it is deployed at customer sites. the framework interacts with software running under the customer's control but i'm not sure if the software is capable of calling back home or not. it could be that the software is run completely firewalled off in the customer's data center in which case it would be difficult to use as an attack vector because the 'attacks' would need to be pushed as software updates.

they knew it would break people's setups but they did it anyway. https://github.com/kubernetes/enhancements/blob/master/keps/sig-cluster-lifecycle/kubeadm/2067-rename-master-label-taint/README.md one good thing about this linus philosophy of not breaking userspace is this happens less often. its very difficult not to break your users because they end up relying on behaviour you are not even aware of but breaking your users in order to fix 'offensive' naming seems to be a bad trade off.

teaching stuff in the school that reflects negatively on certain groups is always going to be tricky. it seems like it is an instance of the [Cardiologists and Chinese Robbers problem] (https://slatestarcodex.com/2015/09/16/cardiologists-and-chinese-robbers/). you can imagine kids being subjected to an avalanche of facts that reflect negatively on a particular group and then when challenged the people who set the curriculum claim these are just facts and you are anti-history. Then there is also the duality that if someone doesn't like some facts being taught they can claim its a 'Cardiologists and Chinese Robbers' problem.

i don't quite understand what is happening in that video. he takes the corner quickly and almost hits the pedestrian but then there seems to be some people chasing the car. did the people start chasing the car when they saw him take the corner dangerously or were they already chasing the car?

I don't understand why the Biden administration didn't try to pass this as law while the Democrats controlled congress.

This twitter dude sounds like a compulsive liar / fantasist. I don't think they can help themselves and they say the most absurd unbelievable shit and I think most of the time the people around them either enjoy the stories or tolerate it so they don't usually get meaningful pushback.

yeah. obv there is something not normal between the rate of rapes between the two races but i think a lot of articles from the 'pro-white' perspective exaggerate the discrepancy because if you have a model of perpetrators selecting random victims then blacks are going to naturally commit more rapes against whites per population. however, in the article emil does make the case that victim selection is not random and perpetrators tend to target victims that they have access to and there is a lot of racial segregation so its a lot more complicated. this 'random' model might not be appropriate.

there seems to be two things

  1. blacks are committing more rapes than whites per population and this seems to be undeniable

  2. blacks are targeting whites more than some kind of 'normal' model of rape would predict. this seems much harder to prove mostly because wtf is this normal model of rape. it seems like some random model of rape is inappropriate due to neighbourhood distribution. so how do you come up with some 'normal' model of rape.

i guess it depends on what you mean by rates. if you compare the ratio ([W rapes W] / [W rapes B]) against ([B rapes B] / [B rapes W]) then you would expect the ratios to be out of whack. but if you compare (W rapes B) against (B rapes W) then these should be the same numerically because it takes 2 to tango so the ratios in the final equations are the same but just ordered differently. but i guess if you doing something like dividing (W rapes B) and (B rapes W) by population numbers of the offender (or victim) then you are going to get ratios that are out of whack because the numerators should be the same but will have different denominators. but i'm not sure what the justification for doing this division would be...

A worked example:

Using this population ratio:

A: 3/4

B: 1/4

Where the population is made of of 50% rapists who rape from the population randomly:

A_r: 3/8

B_r: 1/8

A_r_A: 3/8 * 3/4 = 9/32

A_r_B: 3/8 * 1/4 = 3/32

B_r_A: 1/8 * 3/4 = 3/32

B_r_B: 1/8 * 1/4 = 1/32

BrA is 3x more likely than BrB


ArB is 3x less likely than ArA

but A_r_B == B_r_A

i think the joke is 'but none is laughing, so i'm going to stop'. i think it's meant to be making fun of what conservatives think, or the jokes a conservative comedian might make. but it is AI so it's hard to tell.

https://www.twitch.tv/watchmeforever a seinfeld spinoff hosted on twitch that is produced by AI (originally the davinci model) was banned after it made jokes referencing transgender people. weirdly the joke seemed to be making fun of comedians making fun of transgender people but i guess using the word transgender in standup comedy is a third rail that cannot be touched.


"anyone have any suggestions

i'm thinking about doing a bit about how being transgender is actually a mental illness

or how all liberals are secretly gay and want to impose their will on everyone

or something about how transgender people are ruining the fabric of society

but no one is laughing, so i'm going to stop

thanks for coming out tonight --"

This could just be a result of incompetence. My experience from reporting security issues is that people don't do root cause analysis. So if you report security X they are just going to fix issue X they are not going to grep the codebase to see if issue X is repeated. So its quite possible that someone reported an issue where chat GPT made some argument saying black people were bad. The developer 'fixed' this issue but didn't enumerate all the races to ensure that chat GPT didn't say X race was bad. It's very obvious if chat GPT responds to some prompt about X race in a bad way that you should also check if chat GPT responds to Y race in a bad way for same prompt. But your average jira code slave is just resolving tickets in the most efficient way possible so you end up with this weirdness.

Also, there is a new motion in the FBI Seth Rich FOIA case from the plaintiff that seems to make the claim that the FBI covered up Seth Rich's involvement in the email leak.


Previously from the FBI:



i did slightly better after i think after 4 guesses i just decided left->woman right->man because i assumed the leftist parties would have more women than men and the MP selection would be selected randomly and the the ideological composition would be random. given i have NFI about finnish politics these seemed like reasonable assumptions. tho the ideological composition or the quiz random selection are very dubious assumptions.

that's kind of a weak man. they are going to say that blacks can't succeed with AA because of the systematic racism and that is why they need affirmative action. not because of something intrinsic to blacks but rather something that society is doing to blacks.

sarah is just GPT. KEKW. the funny thing is probably everyone in your timeline i just some kind of AI but just more realistic so just go with the flow.

being in the top 10% of players who have played > 1 game is not necessarily that good. it could still mean you are performing poorly compared to top humans.

don't mess with chicken or pork when it comes to cooking

his latest defence is FTX didn't have a bank account so customers were transferring money to Alemeda and then OOPSY DAISY the money was never transferred from Alemeda to FTX. (https://www.vox.com/future-perfect/23462333/sam-bankman-fried-ftx-cryptocurrency-effective-altruism-crypto-bahamas-philanthropy) I'm not sure how that covers all the deposits because presumably you could transfer crypto as well.

there is a bunch of interesting DMS between the reporter and SBF including this:

REPORTER: you were really good about talking about ethics, for someone who kind of saw it all as a game with winners and losers

SBF: hehe

SBF: i had to be

SBF: it's what reputations are made of, to some extent

SBF: I feel bad by those who get fucked by it

SBF: by this dumb game we woke westerners play where we say all the right shibboleths and so everyone likes us

Maybe HR in some of these tech companies are acting more like a concierge for the other workers. That could explain their high numbers.

SoS claimed there was 10,200 petition sheets. i'm sure after they have finished properly reviewing all the signatures they will find the missing ones that weren't part of the initial count.

it could be related to this: https://twitter.com/Skip_Borders/status/1574542329465036801

My last panel I was ejected for asking simple topic based questions by a radicalized individual that was triggered by the very notion that there were different thoughts on the topic.

Hoping future panels are open to market place of ideas & interested in addressing key issues!

it sounds like they had some wrongthinkers at the event

kiwifarms has also now been removed from the internet archive according to keffals (https://twitter.com/keffals/status/1567259373671522307)