domain:nature.com
Corvos
18m ago
Well, yes, your example causes a lot of misery too.
Like I say, I understand the reasoning but getting punished for being virtuous is still very frustrating!
clo
46m ago
I have to laugh because otherwise I'd cry. Baudrillard was right and I hate him for it.
Just to be clear, here's what happened:
A certain subset of Americans decided that the military banning beards was racist. Afghan tribespeople think people without beards aren't men. American soldiers think beards are for cool elite people.
Does anyone else feel a deep despair at this? Or is it just me? I don't want the desert of the real either, but is the endpoint of this just complete breakdown in shared language and everyone just chucking rocks to see if they have the power to enforce their meaning?
FistfullOfCrows
50m ago
This is very interesting but please next time, label the axis labels at the top and right, that is at x+/y+
CertainlyWorse
No one is coming. It's just you.
1hr ago
This type of self-development journey is more common than you'd think. Also everyone goes through a period of feeling the mortality of their parents and themselves.
From the outside it seems like you're doing pretty good. Finishing a PhD is no joke. Also some of your other achievements aren't small things.
Keep doing what you're doing. I wish I had a bit more to add, but you're already socialising in hobby groups and keeping fit. Meeting a romantic partner would nice, but its not as easy as it sounds while you're juggling all the other stuff.
anon_
2hr ago
my sense is that UCal has already been on a tighter leash for some of these things than many other unis
Yeah, California famously voted to make racial preferences illegal decades before SFFA v Harvard, and reiterated that in a recent vote.
It's one of those strange thing politically, that UCB might be far more progressive than (e.g. Brown) along a number of axes (Chesa failed upwards from his recall in SF to law prof) but has long had a far more meritocratic (albeit still biased) admissions process.
anon_
2hr ago
I'm just flabbergasted that you posted something about Michael Jackson that doesn't have him a the peak of soft power. The man was among the first truly universal superstars. He commanded the admiration of millions.
anon_
2hr ago
I think in the case of the USA the red and blue tribes share quite a lot
Indeed. I think the points of agreement are so broad and deep that they almost vanish into the background. We take them for granted and so the only things that are salient are the outliers.
anon_
2hr ago
I don’t object to free expression of ideas even in contentious situations on controversial topics.
That is true. On topics where there is a live social controversy (most of the Culture War), this is probably ideal.
At the same time, I think this can be weaponized to by people that want to express ideas that are beyond the pale and who want to reap the social approval of having people accept their views because of "etiquette". One particular example that comes to mind is the voluminous academic (at least in the sense of "coming from the academy") literature rehabilitating the "Minor Attracted Person" and wanting us to take this idea seriously. It's a demand for social acceptance of something that society ought not accept.
Of course, the inverse kind of weaponization happens as well -- cancel culture as an entire phenomenon is predicated on wielding this against views for which there is no social consensus. The fact that some views are outside the window of acceptable discourse is temptation enough to realize that one can try to put one's opponents views in that bucket.
[ And of course, this is all inside the bounds of free speech. But then again agitating someone's employer to get them fire for asserting there are 2 genders is also free speech. That doesn't solve much. ]
self_made_human
amaratvaṃ prāpnuhi, athavā yatamāno mṛtyum āpnuhi
2hr ago
I had chikungunya once, back in India. Worst fever of my life, and it prostrated pretty much my entire family. This was presumably well before vaccines, I was a small child.
DradisPing
3hr ago
The mosquito-borne disease discussion always bothers me. The diseases are tied to specific species and to me the solution is obviously that we should make an effort to eradicate those specific dangerous species.
There's always a knee jerk "oh no we can't do that". But generally things like the Anopheles mosquito are invasive and well outside of their natural territory. There is just no downside to exterminating them in most areas, and no real downside to completely wiping them out.
self_made_human
amaratvaṃ prāpnuhi, athavā yatamāno mṛtyum āpnuhi
3hr ago
Is this what is going on? I had thought deep thinking had more to do with scaffolding built to continuously reprompt itself sometimes even using totally different specialized models for tasks.
Reasoning models can be a little heterogeneous as a class. But if you're talking for different "specialized models" in this context, you might be thinking of either a mixture-of-experts setup (a name that sounds obvious but which is somewhat misleading) or a routing system where prompts are assigned to different models depending on what they're good at it/what the prompt requires. That would be the router seen if using the auto mode for GPT-5.
A minimal definition of a reasoning model is one that spends a certain amount of time generating tokens that do not necessarily represent the intended final output to the user, usually delineated by special tokens or tags. Then a lot of funky additional post-training happens to enhance capabilities, you'll have to ask someone better informed.
Speak for yourself, I want my output to be part of the machine god.
Oh I want to be immortalized too. But Apollo included a canary, and if I'm quoting them this much, I feel obliged to ensure I'm not the reason the data gets scraped. In general, I couldn't care less if I'm trained on, and I actively prefer it.
fmac
Ask me about bike lanes
3hr ago
I want to see more political prosecutions, and I want to see more naked corruption between republican governments and their aligned NGOs
Emulating the worst aspects of communist governments... to own the libs?
I'm perpetually amused by the argument of "the Dems shot our country in the foot, therefore, when we're in charge, we should make sure we shoot the other one even harder"
Aren't you supposed to be patriots? You clearly don't want what's best for it
fmac
Ask me about bike lanes
3hr ago
I imagine this isn't the kind of recursive self-improvement you probably had in mind, but I think you'll find it enlightening to examine your intuitions about why this kind of recursive self improvement doesn't "really count".
Maybe I'm too tired but I'm genuinely not sure what the enlightenment is here
Doesn't this just mean that when people say "recursive self improvement" what they actually mean is "holistic full stack recursive self improvement that allows the entity to improve all bottlenecks simultaneously"?
self_made_human
amaratvaṃ prāpnuhi, athavā yatamāno mṛtyum āpnuhi
3hr ago
That's not my own canary, I quoted extensively from Apollo's posts and I included their canary to help reduce the risk of accidental contamination when some scraper comes and takes a snapshot of us.
That's likely the most common route for benchmark contamination. Even if a lab is responsible and doesn't directly train on the test, there's no guarantee that the information won't leak from someone else and end up in the corpus.
It'll only work if a scraper is well behaved, but at least Apollo and other labs will be able to tell that a leak happened, should future models be able to generate that canary string.
DaseindustriesLtd
late version of a small language model
4hr ago
I mean my entire post, of course. If you want to see how biased they are in less clear-cut cases, feed the wiki page of Kobi Kambon and ask if this is science.
urquan
Hold! What you are doing to us is wrong! Why do you do this thing?
4hr ago
“It’s annoying when ‘I am very careful about not creeping out women’ gets interpreted as ‘I am a bad boyfriend risk’.”
In most domains of life, the only way to credibly demonstrate capability and low risk is to have a track record of actually undertaking that activity successfully. If you were in a place to loan out thousands of dollars to a stranger, you’d want to see a credit history too.
aiislove
4hr ago
My rankings are subjective and I'm not married to any of them. If we all filled out the chart with the same people we'd probably all have everybody in (slightly to wildly) different positions, especially considering different people move throughout their lives and as others pointed out can depend on the situation or social surroundings.
Also, come on. Is Ellen really more powerful than Oprah? No one thinks that.
I think that, if I was serving them in some context I'd probably give creepy Ellen whatever she wanted while I'd be inclined to openly roll my eyes in Oprah's face. I think Ellen commands more respect than Oprah based mostly on their vibes and appearances. Oprah seems more haughty and irritating while Ellen seems like she'd keep the interaction on a more professional level which translates to more soft power.
aiislove
4hr ago
Lots of women have the same thing going on - they may want to be rich, clever, happily married etc. but in the moment they are far more emotionally validated by evidence that they are hot.
Are you sure about this? Do any women on this site want to share their experiences? I genuinely don't know if women think this way, because if I swap the genders it seems 100% true but I suspect women are less obsessed with being hot than men are, and are happier with soft status qualities like wealth, cleverness and happiness in marriage and so on. Personally as a man I don't really care about the secondary soft status I just want to be hot and loved for my intrinsic qualities rather than any sort of skill, virtue or prowess
RandomRanger
Just build nuclear plants!
4hr ago
o3 sure is spooky, as is the stuff about rationality forums. Janus the LLMwhisperer also goes on about how tiny amounts of training data can radically change a model's self-conception, training on the information from these alignment tests has a special effect he says.
It's incredibly ironic how badly the classical AI safety movement has failed. Sam Altman credited Yudkowsky for getting all these people interested in AGI and ASI back in the day.
https://x.com/xriskology/status/1622197122979209218
eliezer has IMO done more to accelerate AGI than anyone else. certainly he got many of us interested in AGI, helped deepmind get funded at a time when AGI was extremely outside the overton window, was critical in the decision to start openai, etc.
It goes to show how unpredictable the long-term effects of our actions are. What we have today is exactly the outcome Yud didn't want!
aiislove
4hr ago
Ah, yes, I just skimmed this page but it seems like this is very much what I was trying to describe. I was wondering if anyone else had teased it out before, I'll look into this. Thank you
aiislove
4hr ago
I am not married to the quadrant names, I had considered Barbarian for that quadrant but thought the term was more politically charged than Caveman which is more politically neutral so I went with that one
aiislove
4hr ago
Oops, I meant Charles III. My bad
Not all starlets migrate that way, I even said that Britney went from Princess to Whore and is either at Whore or Hag. Anyone can go any way on the chart depending on the way they move in terms of hard and soft power, it's just easier and more common for people to follow a certain path, because of age and experiences that lead to a certain direction of progression. You could say it is "the same status" in the sense that they are on the chart in the same way or you could say it is "not the same status" in the sense that it is different qualities leading to different positions in different times of their life etc.
True. I'm focusing on the marginal cases. To the extent that our values are mutually incompatible, cooperation is harder, especially in pursuing those values. To the extent that the gaps in values are small and isolated, only small amounts of separation are needed to avoid significant value loss or conflict; maybe the normal separation we have between people, families, social groups, churches and so on is sufficient. The larger the gaps, the more separation is needed, until it's more separation than our society can reasonably accommodate in its current configuration; people start moving to different areas they perceive as lacking the gap, change jobs or careers maybe. As the gaps get bigger and available separation can't keep up, fighting over power becomes increasingly attractive.
Rather, "Tolerance is not a general solution to human nature." It works great over a very wide range, but there are edge cases where it stops working. If you can't cooperate on a few things, maybe you can cooperate on other things, and the value is still net-positive. But there's obviously a point where cooperation just costs too much value on net and it's not worth it any more. Further, we can see these points coming, and act in advance of their arrival, and we can respond to others doing likewise, with the usual caveats about the dangers of acting on predictions.
Things like this seem over the line to me. Also things like this. ...I'd prefer not to do a large-type airing of grievances, but there have been a lot of things Blue Tribe attempted or executed over the last ten years that seem to me to amount to irreconcilable differences. It doesn't matter if some of the things didn't work, or others were reversed when cooler heads prevailed; the knowledge these incidents generated about what Blue Tribe is willing to commit to means that it does not seem to me to be a good idea to trust them to have power over me ever again. Maybe that's partisanship talking. Maybe it's really not all that bad. Maybe nothing ever happens.
...I think there will be a backlash to the things my side is doing now. While that backlash is predictable, it does not seem wise to refrain from doing those things to forestall it, and it will almost certainly be a good idea to generate a backlash of our own when theirs arrives. It is hard to imagine the point at which I will conclude that there's been enough conflict, we should make peace instead. Objectively, it is hard to imagine the other side reaching that point either. We will each perceive what the other has done as reason to double down, and our own actions as justified. The difference, of course, is that I perceive my side to be correct, and their side to be insane. And sure, I would, wouldn't I? This is how tribalism works, we can always retreat into abstractions until there's no difference between right and wrong, good and evil, cue the Dril tweet.
Here in the real world, there's not much of an off-ramp I see. When we cannot agree on the definitions of basic terms like murder, child abuse, rule of law, treason... it seems wiser to me to admit that the problem is beyond us, and pack it in before we really hurt each other.
More options
Context Copy link