site banner
Advanced search parameters (with examples): "author:quadnarca", "domain:reddit.com", "over18:true"

Showing 25 of 1800 results for

domain:cafeamericainmag.com

Fact is that adapability and agentic behavior are key things to consider when discussing whether a robot can replace a human worker

are you sure?

you do not either to replace human workers - combine harvesters replaced vast amount of human workers, without really having either of that

LLM that would hallucinate far less and be better at following orders (with no agentic behavior whatsoever, though I guess adaptability would be be needed) would allow to do the same to programmers

Claude or Grok has suddenly gone "FOOM" and turned into Skynet

if we assume bizarre increase in intelligence of LLM then agentic behavior is absolutely not needed - you would only need human to prompt them once (and surely there are jokers doing "turn into Skynet and murder all humans") several times a day

An LLM on its own is little more than a tool that turns words into math, but you can combine it with a second algorithm to do things like take in a block of text and do some distribution analysis to compute the most probable next word...

that seems quite unusual definition of LLM - who else is using it?

Muppet side-eye.png

Come on, be charitable. It's not a perfect analogy. The point I'm trying to make is that it's a dangerous thing to be carrying around in public. It does require volition, but volition may be influenced by rage, or alcohol, or psychosis, or mental illness, or one bad day.

Humans are fallible. They can just be mistaken about whether they should use a gun in self-defense and end up killing someone anyway. The difference between justified and unjustified can be seconds.

and humans are stupid. They do incredibly dumb shit (warning, death) like shoot each other over literal garbage.

sig owners shift uncomfortably in their seats

This is the polar opposite of an "effort-post;" this might actually be the Platonic ideal of "Reasoning From First Principles" being nothing more than Rationalists making wildly invalid assumptions, insisting they're the only logical position to have, and adamantly refusing to verify if the conclusions accurately reflect reality.

The notion that intelligence agencies would ever say "why bother, these guys are already in the bag?" is profoundly stupid; even if one's only experience with intelligence is exclusively through fiction, one would not say something so incredibly clueless, yet hewre we're holding it up as proof "Epstein was Intelligence" is tinfoil-hat territory?

For decades, some of the most successful (and aggressive) intelligence operations against the US are run by our allies; Japan, South Korea, France, Israel, and Taiwan in particular have consistently been labeled as tops threats by the U.S. National Counter Intelligence Center. Most of these efforts are focused on obtaining business, industrial, and technological secrets, but no small amount also goes into gauging just how sincere America's commitment is to our alliances. Anyone raising their hand and saying "why worry about blackmailing this rich and powerful man, he's already on our side?" would be quitely assigned to work nothing of any importance, with "advise" to their immediate supervisors to find any pretext to fire them, that wouldn't result in any problems.

you'd target rich Chinese, Indians, gentile Russians, and above all rich Sunni Muslims

Not if you wanted any chance of your intelligence operation working.

Epstein bragged about working for intelligence agencies; that is the one thing you don't want your agent of blackmail to be doing.

Again, "Reasoning From First Principles" being utter nonsense. Yeah, you'd like them to not do that, but if they're successful - and it is a known-fact that Epstein was named as a middleman for various African and Middle Eastern deals - you're gonna ignore that problem, until such time as you no longer can. "I can't believe X would do Y, because that would be stupid" should not ever be something that occurs to you.

Man, don't shoot the messenger here. I'm trying to explain my understanding of an ethos here, not grandstand about it being my position. Either way, the thing is that the rule against killing is, again to a first approximation, fairly absolute; and to someone who actually believes in an absolute rule, asserting that you actually want to break it in a fairly broad special case is not persuasive. Going with my previous metaphor, you may be saying something to the effect of "but they only want to have sex with minors who are really asking for it" - the difference just does not matter to those who perceive sex with minors to be intrinsically wrong, no matter the details, and a lot of people in Europe also likewise perceive killing to be intrinsically wrong, no matter the details. There is really a complete disconnect of moral intuitions here, with both sides finding the other barbaric - if the story is "Texas home owner shot robber who was running away with his TV", classical Americans will be cheering, while Europeans (+Europeanised American urbanites) will be cheering to lock the home owner up. Can you muster the theory of mind to understand that some people actually believe that there are no "bad guys" who it is a good thing to kill?

I don't really see why that's bad - presumably even Europeans want their military, soldiers, spies etc. to do their jobs.

It's complicated. I think a decade or so ago the answer from a narrow majority actually would have been "no", if that job involves actual killing. Now, some of them still will say "no", many more will prefer to not think about it, and many will think something to the effect of "yes for external enemies, but this is not a principle on which you can run a civilised society internally". I'd imagine that even people who are deep in pro-UA brainrot territory in countries like Germany would more often than not balk at the idea that counterintelligence should kill Russian spies inside Germany.

It arguably wouldn't actually be a good swap for the US to get European gun violence levels if it also meant getting European attitudes and regulations towards air conditioning! And that's without getting into values-based stuff like free-speech rights.

I'm not so convinced that they are strongly correlated at all - East Asia has ubiquitous AC but no guns and an atrocious free-speech situation compared to Europe as well, Russia flip-flops but at least intermittently had quite liberal gun laws with no relation to its AC or speech situation. Either way, the heat death figures you refer to always seemed fairly cooked to me - Eurocrats have an incentive to inflate them to support the climate change narrative, while the US figure seems pretty inappropriately small for its burgeoning homeless population.

This was quite different back then than it would be now. If you live your entire life in one neighbourhood and there are a dozen other ethnicities living there, soon enough everyone will adopt a common tongue.

Except that's not true. New York had significant yiddish, italian, bulgarian, lithuanian, greek, etc. communities, where those languages were spoken alongside, or even to the exclusion of, english in the early 20th century. Chicago had polish, ukranian, etc. Los Angeles today has several areas where spanish is predominant, as well as several suburbs that are at least duolingual with many/most advertisements in mandarin, vietnamese, etc. Up until WWI huge swathes of the midwest spoke german, usually as a second language, but in some areas to the exclusion of english.

Immigrant ghettoization is extremely common, and tends to preserve language use.

What percentage of pigs are cannibals today?

Its a record, typically on paper of purchases or debts. They are less common now, but a typical example would be when you go to the dry cleaner and drop off a suit, the carbon paper they hand you is commonly called a chit.

It's interesting how open carry has changed in the US in the past 30 years. I grew up in a place with many guns and where open carry was legal, but only the most trashy of rednecks would open carry, and they were derided by other gun owners. "Whatsamatter, you think the Russians are going to invade today?"

Comparing modern to founding-era and 19th century gun discourse is also fascinating; back then there were laws against concealed carry because that was viewed as covert, sneaky, and dishonorable. What do you have to hide and who are you trying to surprise? Whereas open carry was considered completely normal. Nowadays it's the sight of a gun that freaks people out, so concealed carry is much more popular; allow the gun person their hobby without scaring everyone.

the telos of a gun ... is to kill people

There are around 400 million civilian-owned guns in the US now. Over the last century US guns have been responsible for something like a million homicides and a million suicides. Were the other 99.5+% of the guns designed or made wrong?

we can begrudgingly haggle over exceptions

How can you avoid exceptions? Should the enforcers of a gun ban have guns? If so, then you're making an exception before gun supporters even have to lobby you for it. If not, then it's not going to be much of a ban.

I hope that exception gets more sophisticated than "of course people Following Orders still get guns". Europeans under those rules were responsible for way more than just one million homicides over the last century, even if you only count the civilian victims.

like self-defence against someone who tries to kill you first

This seems like an important one, right?

We're happy to drip literal poison into our veins in some cases, because the poisons used in chemotherapy kill cancerous cells faster than non-cancerous cells. The direct purpose of those chemicals, what you're calling the "telos", is to kill human cells: if only 0.5% of doses of a prospective chemo drug killed any human cells, then we really would conclude that it was probably designed and/or made wrong. Their direct effects are pretty lousy. Their specificity kinda sucks and they kill good cells too. But they're still worth it, so long as you realize you can't ignore the distinctions between targets or the indirect effects of the killing in a final analysis.

The common European value system says that basically to a first approximation there should not be a legal way to kill people

About a quarter of Europeans live in a country where assisted suicide is now legal.

a leashed tiger, or a running chainsaw.

This is a wild comparison; the gun is inert and has no volition of its own. Nor is it always in a state of active danger like the running chainsaw. Firing a gun is not something that is done by failing to pay sufficient attention to the gun - it requires volition and active intent across several particular bodily motions to draw, aim, turn off a safety, and fire a gun, just like it would to grab someone by the head and try to break their neck, or try to stab someone with a knife or pen, etc.

I am trying my best to be charitable here, but I literally explained why that paragraph was wrong, over and over, and you... just repeated that same paragraph?

I will say it for the last time. That paragraph is pure fiction from your part. There is no interface layer, there is no second algorithm like you described, and you have completely misinterpreted how LLMs work. Ironically, that paragraph sounds like an LLM hallucination.

Am I out of bounds by saying that this is constitutes trolling at this point? This is genuinely upsetting.

Dude, look, here's code for the core functionality of a GPT2 model taken from the most simplified but still functional source I could find: https://jaykmody.com/blog/gpt-from-scratch/

This is the ENTIRE code you need to run a basic LLM (save for loading it).

import numpy as np

def gelu(x):
    return 0.5 * x * (1 + np.tanh(np.sqrt(2 / np.pi) * (x + 0.044715 * x**3)))

def softmax(x):
    exp_x = np.exp(x - np.max(x, axis=-1, keepdims=True))
    return exp_x / np.sum(exp_x, axis=-1, keepdims=True)

def layer_norm(x, g, b, eps: float = 1e-5):
    mean = np.mean(x, axis=-1, keepdims=True)
    variance = np.var(x, axis=-1, keepdims=True)
    return g * (x - mean) / np.sqrt(variance + eps) + b

def linear(x, w, b):
    return x @ w + b

def ffn(x, c_fc, c_proj):
    return linear(gelu(linear(x, **c_fc)), **c_proj)

def attention(q, k, v, mask):
    return softmax(q @ k.T / np.sqrt(q.shape[-1]) + mask) @ v

def mha(x, c_attn, c_proj, n_head):
    x = linear(x, **c_attn)
    qkv_heads = list(map(lambda x: np.split(x, n_head, axis=-1), np.split(x, 3, axis=-1)))
    casual_mask = (1 - np.tri(x.shape[0])) * -1e10
    out_heads = [attention(q, k, v, casual_mask) for q, k, v in zip(*qkv_heads)]
    x = linear(np.hstack(out_heads), **c_proj)
    return x

def transformer_block(x, mlp, attn, ln_1, ln_2, n_head):
    x = x + mha(layer_norm(x, **ln_1), **attn, n_head=n_head)
    x = x + ffn(layer_norm(x, **ln_2), **mlp)
    return x

def gpt2(inputs, wte, wpe, blocks, ln_f, n_head):
    x = wte[inputs] + wpe[range(len(inputs))]
    for block in blocks:
        x = transformer_block(x, **block, n_head=n_head)
    return layer_norm(x, **ln_f) @ wte.T

def generate(inputs, params, n_head, n_tokens_to_generate):
    from tqdm import tqdm
    for _ in tqdm(range(n_tokens_to_generate), "generating"):
        logits = gpt2(inputs, **params, n_head=n_head)
        next_id = np.argmax(logits[-1])
        inputs = np.append(inputs, [next_id])
    return list(inputs[len(inputs) - n_tokens_to_generate :])

def main(prompt: str, n_tokens_to_generate: int = 40, model_size: str = "124M", models_dir: str = "models"):
    from utils import load_encoder_hparams_and_params
    encoder, hparams, params = load_encoder_hparams_and_params(model_size, models_dir)
    input_ids = encoder.encode(prompt)
    assert len(input_ids) + n_tokens_to_generate < hparams["n_ctx"]
    output_ids = generate(input_ids, params, hparams["n_head"], n_tokens_to_generate)
    output_text = encoder.decode(output_ids)
    return output_text

if __name__ == "__main__":
    import fire
    fire.Fire(main)

Let me walk you through the important parts:

First, the prompt is encoded with a byte-pair encoding tokenizer. This groups letters together and turns them into integers to use as ids. This is just a look-up table.

The generate loop gets logits directly from running the LLM. What are logits? It's a probability that's assigned to each possible token id.

With that, you just need to take the highest value and that gives you the token.

See how the LLM directly outputted the highest probable word (or token rather)? Where is the "interface layer"? Where is the second algorithm? No such thing.

And yes, this is pretty much how ALL modern LLMs work. It's extremely simple. They just predict the next token, by themselves. All the sophisticated outputs you see arise purely out of that. THAT is the miracle that no-one could believe for a while.

The specific fantasy you seem to be upset at is "killing a bad person who is trying to do a bad thing." Most gun owners who are interested in self-defense are interested in self-defense. Movies and gun manufacturing ads and the NRA website and all of those things you're discussing aren't promoting the idea of unlawful violence or mass shootings. (It's actually imho the liberal-leaning press and gun control groups that do the worst to spread mass shooting memes, because they amplify the contagious meme of mass shootings to advance, in the case of the latter, their policy agenda). They are promoting the idea of stopping a bad guy. You can go read the NRA magazine, they (at least used to. maybe they stopped) pull accounts of robbers, rapists, mass shooters etc. getting stopped by "the good guy with a gun" which happens pretty often, honestly. If there's a fantasy here, it's specifically the same fantasy that people who join the military or police often have. I think it's fine to criticize certain aspects of this but fundamentally wanting to stop bad people from doing bad stuff is an honorable impulse.

Moreover, this fantasy is in large part a response to another fantasy failing to become a reality, that of police being able to keep you safe if someone intents to hurt you or your loved ones.

Shady dealers, private market sales, straw purchasing, theft, etc. These are all much easier to prevent in a society with few guns and strict controls then a society with guns everywhere.

Is this actually true? I fully accept that prohibition lowers the general public's ability to access the prohibited substance (e.g., alcohol prohibition actually lowered drinking significantly and permanently changed U.S. drinking culture), but is there any evidence that it significantly impacts illicit dealing in the prohibited/regulated substance?

No, I mean it in a practical sense of "whether I can legally buy and carry a gun". Once you start agreeing that background checks are OK, there's no fence on that slope; the OP already got to invasive background checks and even psych exams, for crying out loud. A psych exam to exercise a right? When a lot of psychiatrists are straight-up anti-gun? That's obviously a vitiation of the right.

What makes it superior for sporting over either something like a hunting rifle, or something like a fairground gun that shoots tiny bullets of a few millimetres calibre?

All of those guns are for different purposes. The .223 is a cheaper, lower-performing round compared to say a .308. It's also more fun to shoot (less recoil, semi-automatic), and it is a tiny bullet (same diameter as a .22, so not dissimilar to a fairground gun most likely). But people do sporting events using all sorts of different calibers.

Personally I think the .223 is a very good varmint round, and that's how I've used it.

a legal way to kill people

To be clear, when I say this, I mean "it should be legal to own deadly things" and that's about how I took your phrasing. Most pro-gun-people (including me) don't support it being legal to execute people randomly, but perhaps my phrasing was...unclear. But, to your point, at least in the US, they think that the Second Amendment is an important backstop to liberty. It's hard to tease out the correctness of this, but the US of A is doing much better than Europe in this regard. (For instance, just as a wacky example, in A/C unfriendly Europe, heat deaths kill more people than firearms in the US of A. It arguably wouldn't actually be a good swap for the US to get European gun violence levels if it also meant getting European attitudes and regulations towards air conditioning! And that's without getting into values-based stuff like free-speech rights.)

there still were large and massively funded organisations deliberately binge-drinking to the point of getting it

I'm not sure what the organization has to do with it. Alcoholism is much more dangerous problem in the US than firearms, but alcohol is much easier to procure (and is also glamorized in the media, much as guns are!) If all of the gun-rights orgs shifted their focus to sporting, I doubt that gun control groups would be assuaged, because at the end of the day their goals are things like "stopping school shootings" not "stop optics we don't like."

the most popular fictional depictions of alcoholic drinks all involved flashy celebrations of how they induce cirrhosis

I mean - most popular fictional depictions of guns are of people, often those who are legally permitted and encouraged to have them (cops, spies, soldiers, etc.) using them to stop bad people. I don't really see why that's bad - presumably even Europeans want their military, soldiers, spies etc. to do their jobs.

"fantasizing about killing"

The specific fantasy you seem to be upset at is "killing a bad person who is trying to do a bad thing." Most gun owners who are interested in self-defense are interested in self-defense. Movies and gun manufacturing ads and the NRA website and all of those things you're discussing aren't promoting the idea of unlawful violence or mass shootings. (It's actually imho the liberal-leaning press and gun control groups that do the worst to spread mass shooting memes, because they amplify the contagious meme of mass shootings to advance, in the case of the latter, their policy agenda). They are promoting the idea of stopping a bad guy. You can go read the NRA magazine, they (at least used to. maybe they stopped) pull accounts of robbers, rapists, mass shooters etc. getting stopped by "the good guy with a gun" which happens pretty often, honestly. If there's a fantasy here, it's specifically the same fantasy that people who join the military or police often have. I think it's fine to criticize certain aspects of this but fundamentally wanting to stop bad people from doing bad stuff is an honorable impulse.

It seems to me that you are making the vibes-based argument that "Hollywood thinks gun violence is good therefore guns are bad" but my argument here, on the whole, is that if you look at actual use cases and not vibes vast majority of use even of guns that are e.g. derived from military designs is for peaceful purposes. The same way that most drinking isn't to die of liver failure even though that's a not infrequent outcome.

I mean, for me, it was the realization that principles mean nothing. A sufficiently motivated adversary will find some way of maliciously using your "principles" against you, and then chiding you for defending yourself. You see it with free speech, and malicious actors doxing the families of "free speech absolutist". Congratulations edgelord, you found the edge. Nobody doubted you could.

I think if you grew up in a high trust society, you take for granted that principles are just another part of the social contract. As it descends into a low trust hellhole, where there is no social contract what so ever, principles just amount to handing your daughters over to literal roving gangs of barbarian rapists and sitting idly by because you wouldn't want to step out of your lane and violate the state's monopoly on violence. If you want to stick to your principles, you must allow savages to repeatedly rape her.

As civilization crumbles around us, we repeatedly see what our "principles" are earning us, and it's suicide.

Being wary around actually dangerous things isn't neurotic!

These things are reasonably safe as long as you pay extra attention, but carelessness can kill or maim very easily.

I’ve said before, my solution to this slippery slope is that the gun regulatory body is elected by current gun owners with high mandatory minimums for breaking the rules- that were made by gun owners themselves- and strict penalties for localities trying to get around them.

That doesn't solve the problem, as current gun owners may well be happy to be part of an exclusive club. And of course the regulatory body is likely to be even more so.

I evaluate such proposals first by "Would I be able to lawfully get a gun under these rules?" -- this one fails that simple test.

I’ve said before, my solution to this slippery slope is that the gun regulatory body is elected by current gun owners with high mandatory minimums for breaking the rules- that were made by gun owners themselves- and strict penalties for localities trying to get around them.

I mean I have gained 10 pounds since that post, which puts me at closer to 22.

It's the same forced attention I get around high cliffs, or heavy machinery, or a busy highway. I might know that the leashed tiger is tame. I might be aware that the running chainsaw has a safety guard, but I can't put it out of mind

Really? This seems… neurotic. Just don’t go on the highway, or too close to the cliffs, or stand next to the guy using a chainsaw.

I have a hard time believing that if there was demand bringing guns into european countries would be that difficult, considering how big the external frontiers of the EU are and how open the internal ones are, and considering how drugs make it there, illegal refugees make it there, etc...

I think that for the most part, there is little demand because local hard criminals are still civilized enough to understand deadly shootouts are not worth the hassle because they bring down a lot more heat on them. North American street gangs are not civilized; they believe shootouts are cool. As for terrorists, if you distinguish mental health cases (random nutjob just starts stabbin') from organized terrorist cells, I think the latter are usually packing (and packing military equipment).

Cracking down on straw purchasing is like this too, IIRC- democrats refuse to do it even if it’s doable, likely to work, and a major issue with gun crime.