ulyssessword
No bio...
User ID: 308
ulyssessword
2d ago
You did not say "no"
Why would anyone answer a thought experiment with a direct factual analysis? I wouldn't use the trick calculator because I would use a normal one, or possibly specialized software that has error-checking that goes beyond faithfully calculating my button presses. Wow, I'm so insightful.
I notice that you haven't answered the question either: Have you seen humans? I personally see dozens of humans on an average day, but I wouldn't want to assume anything about your answer.
I know its long but seriously watch the video essay on Badness = 0 I posted up thread. It is highly relevant to this conversation.
Where's the relevance? Was it "Using an LLM to answer your questions will cut your workload by 99% but not 99.99% because you have to follow one link to confirm its response"?
0-6:00 Detail orientation!
6:00 - 9:00 Instead of watching >100 videos each about 10-30 minutes long and assessing them himself (or using any other research strategy), the author used a (now) old model with 5% the parameters of GPT4, and it confused a video about error correction algorithms with a video about admitting to and correcting your errors. He got his answer within minutes.
9:00-12:00 Intro to LLMs and his toy example.
12:00-19:00 BoVeX, which is a typesetting software he made that rewrites text to eliminate "bad" breaks in text (e.g. hyphens, overspacing).
19:00-22:00 Conclusion/credits.
ulyssessword
3d ago
Let's try a concrete example. Excerpted from here:
The o1 model identified the exact or very close diagnosis (Bond scores of 4-5) in 65.8% of cases during the initial ER Triage, 69.6% during the ER physician encounter, and 79.7% at the ICU
65.8% accuracy isn't that great, but buddy, have you seen humans?
—surpassing the two physicians (54.4%, 60.8%, 75.9% for Physician 1; 48.1%, 50.6%, 68.4% for Physician 2) at each stage.
The state of the art for generating accurate medical diagnoses doesn't involve gathering the brightest highschoolers, giving them another decade(-ish) of formal education, then more clinical experience before asking for their opinions. It involves training an LLM.
If I ask an LLM about anything, I need to do the research that I would have done even if I had not asked the LLM.
I'm almost with you there. I need to do some of the research I would've had to do without the LLM, but it adds enough to displace a Google search or two while being faster and easier.
ulyssessword
3d ago
When you find something via Google, do you immediately and unconditionally trust it? I don't, because Google's results are full of nonsense. In response, I've developed google-fu to both refine my queries and judge the results. The same goes for every other source there is, from physical libraries to subject-specific Discord servers.
Do I compare LLM output to Google results? Sure, but that's nothing special. Comparing what you find in different sources is a pretty basic tactic.
LLMs are part of a complete breakfast research strategy, and a pretty good one at that.
ulyssessword
4d ago
...as anything other than nonsense generators.
As opposed to the other sources you can go to, which are...?
I am grading on a curve, an LLMs look pretty good when you compare them to traditional sources. It's even better if you restrict yourself to free+fast sources like Google search, (pseudo-)social media like Reddit/StackOverflow, or specific websites.
ulyssessword
6d ago
has to be subsidized by the state government because it’s just not profitable anymore.
AKA: charging what it costs would be unacceptable. I'm sure there's some price where it makes sense to offer flood insurance in a floodplain, but the government decided that people should pay less than that.
At least it isn't a price control forcing the insurance companies take an (expected) loss on every policy.
ulyssessword
8d ago
What is the definition of an "enemy of the United States" though?
Hamas is designated as a terrorist organization. If that isn't sufficient to meet the definition, then I'm not sure what is.
ulyssessword
9d ago
I always treated the 1-10 pain scale as logarithmic, like earthquakes or sound. 7% of a magnitude 9.0 earthquake ("Near total destruction – severe damage or collapse to all buildings.") is 7.8 ("Causes damage to most buildings, some to partially or completely collapse or receive severe damage."), not 0.63 (the scale only goes to 1.0, which are not felt).
Or going the other way, stubbing a toe might be a 3. Stubbing three toes is definitely not a 9. It might not even reach a 4.
ulyssessword
10d ago
Where are you finding that?
ulyssessword
10d ago
I've also seen arguments that a particular distribution of values for IQ, Openness, Conscientiousness, Extroversion, Agreeableness, Neuroticism, etc (including many factors that psychologists don't measure) is the best for a nation, and the only way to get spread is to select based on race.
Even though the "good" numbers might be higher on other people, naive number-maxxing would lead to a failure mode of some kind. It's often unspecified, but the ones I can remember involve out-of-touch highly [good trait] people making norms that are legible and achievable to them, but disastrous to everyone else. Liberalization of sex and drugs are the main culprits.
By extension wether a man is black or white must matter more than whether they are an aged Supreme Court Judge or a Twenty-something meth head.
Maybe. I saw some stats on the Trump 2016 election, and the survey showed that being Black was a better predictor of presidential vote than being
ulyssessword
10d ago
Nice strawman. But even the most hardcore HBD believers would accept that the worst whites are likely worse in some aspects than the best non-whites.
Ethnonationalists are (often) also HBD believers, and they say that the important aspect of a person is their race, full stop. You could point to higher intelligence, longer life, better health, or lower criminality among other ethnic groups, but that still wouldn't convince them that someone from another race is better in the ways that matter because that's not what they're judging people on.
ulyssessword
12d ago
I think it's pretty clear they reduce overdoses...
They may reduce overdoses, but I think that's far from clear.
As a (sort of) counterexample, "...B.C. has implemented every harm-reduction program that has been proposed, from safe-injection sites to safe supply and effectively making all drugs legal. As each new measure has been introduced, drug overdose deaths have increased, except for a brief drop in 2019."
The toy model is that they would otherwise stop (often from from death), but instead they continue for longer before stopping (slightly less often from death), and the time they spend doing drugs is higher and therefore the social cost is higher as well.
ulyssessword
13d ago
That being said, overperformance of multiracial students would be consistent with heterozygote advantage.
I suspect that it's some amount of selection bias as well. Specifically, being multi-racial is higher class than being single-race, so people with X% of their genes from one race and Y% from another would answer differently on the survey based on their class.
ulyssessword
13d ago
Showing your ID in person is not a privacy risk. Sending it over the internet is.
ulyssessword
14d ago
Okay? So ban porn advertising on any site that targets children. I'm pretty sure that law isn't even necessary because websites have a lot more control over their ads than stores have over their neighborhood.
If a city council sorted their areas by crime rate and excluded adult bookstores from the bottom X%, then I'm pretty sure the (prospective) store owners would have a good case the restrictions are illegal. If the city pulled all its cops and banned private security from them, it would be a slam-dunk case.
ulyssessword
14d ago
No. There isn't a crime rate minimum for opening an adult bookstore.
Uploading your driver's license is a government-mandated risk to your privacy, while going to the bad side of town is just an unfortunate coincidence.
How does someone that blatant stay employed long enough to reach your house?
Doorbell cameras are in 1/4 houses, and let's say half report it. That means she could get away with it 8ish times, or about one full day of normal deliveries. How unlucky do you have to be to be in the first eight(ish) thefts of her career? Alternatively, how bad can the companies be that they let her keep her job after getting caught?
ulyssessword
15d ago
The usual progression for small organizations is:
- Some minor punishment
- Blacklist them, revoke their membership, ban them from the premises, etc.
- Commit to calling the cops if they return.
Any organization can (theoretically) do #2 on a whim and #3 if needed.
ulyssessword
16d ago
Have twin studies been ignoring that factor?
Yes, routinely separating people from their family and other support structures is slightly insane.
ulyssessword
16d ago
Can you be more specific?
How would social status vary more among fraternal twins than identical twins? Are there family dynamics that push fraternal twins apart (on whatever stats you use) while pulling identical twins together? Do fraternal and identical twins live in different climates?
ulyssessword
16d ago
Isn't the whole point of martial arts (at least some technical ones, like BJJ) to make this possible?
Same with sports, and yet regional U16 boys teams routinely beat world-class women's teams in hockey and soccer (at least).
Being better than every woman in the world at a physical activity isn't too outlandish for a man, and I'd bet that being better than every woman in a given city isn't uncommon among dedicated amateurs.
ulyssessword
18d ago
Stranger things have happened, but dying four days after a leg wound is certainly up there IMO. I'd think you would either die in a few minutes or get enough help to recover, but maybe there's a middle ground.
More options
Context Copy link