faul_sname
Fuck around once, find out once. Do it again, now it's science.
No bio...
User ID: 884
faul_sname
Fuck around once, find out once. Do it again, now it's science.
6hr ago
I expect AI to never figure out how to do maintenance on its own datacenters
Never is a long time. It still seems plausible that even if continual learning is the most important bottleneck, and even if algorithmic improvements fail to crack it cleverly, brute force will still work eventually. Maybe not until 2050 instead of the 2030 the more breathless people are predicting if we have to wait on Moore's Law, but it seems fairly inevitable unless some catastrophe kicks industrial civilization significantly back before then.
faul_sname
Fuck around once, find out once. Do it again, now it's science.
3d ago
... if the shooter was pushed into this by feds who were trying to sell a particular image of a danger the country faces that only they are able to address, does that count as a "false flag"?
faul_sname
Fuck around once, find out once. Do it again, now it's science.
4d ago
Are there any places you would want to live where violence along partisan lines is commonplace? If not, why do you think that is?
faul_sname
Fuck around once, find out once. Do it again, now it's science.
4d ago
Practically every commenter where? If you're looking at your Twitter FYP that's specifically tailored to you personally to maximize your engagement, which in practice tends to cash out as "that which makes you personally angry or afraid".
Do we even know yet that the shooter was a leftist rather than a random crazy person with no coherent ideology?
faul_sname
Fuck around once, find out once. Do it again, now it's science.
4d ago
the DNA lounge in San Francisco has a "neck shot" special tonight
That is in exceptionally poor taste.
At what point is it defensible to take up arms in defense of one's community?
You're glowing.
faul_sname
Fuck around once, find out once. Do it again, now it's science.
6d ago
Yeah, "there are 4 experts on this equipment in the world and none of them are US citizens and we need one of them for 4 weeks" is a problem which is not well addressed by my proposal (or by our current system).
faul_sname
Fuck around once, find out once. Do it again, now it's science.
6d ago
I have heard complaints that this would probably drop them all in high-cost-of-living areas,
Which is a problem why exactly? Most of the people who object to immigration are living outside of those extremely HCOL areas anyway, seems like it would do a good job of ensuring that the areas that wanted lots of immigrants had lots of immigrants and the places that don't like immigrants don't get them.
faul_sname
Fuck around once, find out once. Do it again, now it's science.
6d ago
... man, it would be nice if the US could just have some areas that are primarily there as economic zones, and some areas that were reserved for protectionism and preservation of American culture. SF / NY / Boston / Seattle could host the global megacorporations that want to bring in the world's best no matter the cost, while much of the rest of the country could be reserved for cultural preservation.
faul_sname
Fuck around once, find out once. Do it again, now it's science.
6d ago
It’s absolutely grating hearing people claim the US economy relies on these “high-skill” workers as if the majority of them are doing groundbreaking technical research.
I mean the issue is that we're not selecting very hard for actual high skill / rare skill people.
If we just made the H1B system conditional on paying the employee $250k we’d get all of the benefits of the actual high-skill immigrants and not the army of Indians undercutting everyone who just wants a normal boring office job.
Yeah this seems like a good policy decision and I expect it doesn't happen because of specific lobbying to prevent it, not because it's unpopular with voters.
faul_sname
Fuck around once, find out once. Do it again, now it's science.
6d ago
The general pool of foreign workers does not have to be identical to the pool of foreign workers we bring in. We can and should be selective.
The top 10% of people who would come to the US if they had a chance are genuinely more skilled than the median American, and there are enough people who want to come to the US that we could pull from that top 10% for a long, long time. We're a country of immigrants, and we should be strategic about that.
faul_sname
Fuck around once, find out once. Do it again, now it's science.
6d ago
Or the idea that perhaps we should have functional immigration pipelines specifically for high-skill workers, rather than almost ignoring human capital in our immigration process.
B-1 visas do allow some types of work - I bet LGES argues that the workers were there "to install, service, or repair equipment/machinery purchased from a foreign company", or "to train U.S. workers to perform these services", both of which are permissible activities under a B-1 visa per CBP's own documentation provided that the workers do not receive compensation from a US source.
If the workers did receive compensation from a US source, that means someone fucked up somewhere, but my guess is what actually happened is that CBP disagrees with LGES about whether the activities these workers engaged in qualify as installing, servicing, or repairing equipment/machinery purchased from a foreign company, and decided that the appropriate course of action was to chain these workers up and make a self-congratulatory press release about it, and that we will hear any follow-ups about the outcomes of this raid in terms of findings of actual wrongdoing.
(Disclaimer: IANAL, TINLA)
faul_sname
Fuck around once, find out once. Do it again, now it's science.
6d ago
I don’t think you need step 2. A theory that passes the gates of "makes novel, interesting, and falsifiable predictions" and "and those predictions end up panning out" is already rare enough that the volume will remain manageable even with the deluge of AI slop. Most AI slop frameworks won't even pass the first of those two gates, and that's the easy one.
I would not be surprised if half of the workers were arguably violating the terms of their visas, but I expect the modal case here looks like "a worker for one of Hyundai's subcontractors was here on a B1 to suprvise the installation of equipment, and demonstrated to a worker on site how the machine was supposed to be hooked up when they're technically only allowed to describe" not like "Hyundai shipped in 500 Koreans on tourist visas to do unskilled construction work building the factory". In other words, I expect that the majority of detainees were authorized to work in the US, but I would be unsurprised if some were doing types of work they were not authorized to do, though I expect the majority were at least ambiguously authorized to do the sort of work they were doing.
Under my model I would be unsurprised if e.g. DOJ and Korean company disagree about whether work should fall under "contracted after-sales service" or "supervising installation of equipment". But under my model "chain them all up" is not a reasonable response to "people who are not flight risks were doing normal business things but we think they might have technically violated the terms of their visa, we'll find out in court".
I am unsure if there are any good and timely metrics but I would be quite surprised to see e.g. table 42d here showing 475 more (or even half that more) noncitizen enforcement returns to South Korea in 2025 than in 2024 - for reference the current latest data is 713 returns in 2022. (The latest available year here is 2022, so it might be a while before 2025 daya shows up). And my read is that DHS would enforce if they have even a vaguely plausible case of visa violation, so I think absence of this particular evidence would be evidence of absence of such a case.
There might be less janky ways to operationalize this, I'm open if you have ideas.
faul_sname
Fuck around once, find out once. Do it again, now it's science.
7d ago
Would you put up money on a bet that at least half of the people arrested were unambiguously violating the rules of their visas?
faul_sname
Fuck around once, find out once. Do it again, now it's science.
7d ago
If you think that changes in your qualia would have noticeable effects, what is your reasoning for saying
What if it's assigned randomly, like the first thing you ever see is assigned what I see as "red?" There's just no way of knowing, no conceivable test to find out.
(Emphasis mine)
If you expect that your experience of red would be different if it were "assigned" first vs second, such that you would notice if the first "assigned" quale switched places with the second, including all associations, then it seems like there are at least conceivable tests.
faul_sname
Fuck around once, find out once. Do it again, now it's science.
9d ago
Assigned by what? By "qualia" are you referring to anything you've ever experienced? If so, how do you know you've experienced qualia?
If the first color quale you ever experienced was "assigned" to red, and the second to blue, and then one day they magically switched, would you notice a difference?
If no, why do we care about "qualia"?
faul_sname
Fuck around once, find out once. Do it again, now it's science.
11d ago
This "color" is how you perceive that wavelenth to be
My perception of the color is not a simple function of wavelength - see example of the blue+black/white/gold dress. My perception of a color is the effect it has on my mind. There is no perception of a color outside of the effect it has on my mind.
In terms of the effects colors have on our minds, we currently have limited direct visibility into this, but
- "limited" is not "zero"
- that ability is improving over time
- we have good reasons to believe that, for the most part and with many caveats, minds that sense data downstream of a set of causal processes develop highly analogous internal maps of those causal processes, even if their sensory data is not the same modality. When the sensory data is of the same modality, the internal structures will likely become more analogous, not less.
I suppose one thing to check - do you agree that two identical-to-the-atom clones observing identical-to-the-photon sensory inputs would have identical qualia? Or do you think even that is not something we can have high confidence in?
faul_sname
Fuck around once, find out once. Do it again, now it's science.
11d ago
Do those sensory inputs exist as an experience that I have outside of my memories, emotions, etc. though?
No? What would they even be sensory inputs to?
If they don't, then how am I able to identify colors of entirely new things sans context?
I don't think that's a thing you're able to do sans context. Infants, lacking context, aren't able to identify the colors of anything.
I suspect we're using the word "context" differently - what exactly do you mean by "sans context"? Are your memories a part of the context? Are the innate saccade patterns that all humans use to look at things (e.g. gaze snaps to contrast, edges) part of the context? How about the learned saccade patterns (e.g. scanning in reading order)?
so I conclude that I do experience sensory inputs, i.e. those sensory inputs are a form of qualia
You don't experience unmediated sensory inputs. The map is not the territory, and you can only experience the map, never the territory directly. See exhibit 1932741: the blue/black or white/gold dress. There's an excellent diagram on that page which shows how the exact same colors on the screen can lead to the perception of a white/gold dress or a blue/black dress, in a way that makes it very easy to verify that your raw sensory data really is the same for the blue on one dress and the white on the other.
And so if you have a quale of seeing white on the ruffles of the dress, that quale is not just your raw sensory inputs.
Which then raises the question of if the qualia of me experiencing the sensory input from observing a stop sign is similar to that of someone else doing the same thing.
I would say the question should be "how similar is it" rather than "is it similar", but yes.
that wouldn't actually get us to the similarity of the sensory inputs themselves.
True, but since we don't directly experience the raw sensory inputs, I don't know how much it matters how similar the raw sensory inputs are. We could quantify the similarity of those raw sensory inputs (e.g. by doing the same dimensionality reduction trick on optic nerve spike frequencies), but I don't think doing so would buy us anything beyond pretty pictures to look at and maybe some cures for diseases.
It's also possible that, since qualia is intrinsically and, as-of-yet, inescapably subjective, the very concept of comparing qualia between two people is incoherent, and the best we can do is to figure out if the qualia of the meaning that we ascribe to sensory inputs are similar, as a proxy that we can never get better than.
I reject the idea that qualia are inescapably subjective. People talk about qualia all the time. Therefore, those qualia are causally upstream of what they're saying. If you can figure out the full chain of causality from sensation to perception to meaning making to conversion to language to speech, I don't think there's anything left to explain. It's a lot of stuff to understand, and we don't yet understand all the links in that chain, but that's a statement about the inadequacy of our knowledge, not the unknowableness of the phenomenon.
faul_sname
Fuck around once, find out once. Do it again, now it's science.
11d ago
You're associating your sensory inputs with your memories, emotions, concepts, behavioral associations, etc. If you sever your optic nerve, and then you point your eyes at a stop sign, you will not experience redness.
I claim that qualia are what it feels like from the inside to ascribe meaning to your raw sensory experience.
If you see something crimson, and then something cardinal red, are those "the same red" to you? My guess is that you can distinguish those colors, if they are put side-by-side next to each other, but that the associations that each color in isolation brings up in your mind are quite similar.
I don't think there is "something it is like" to see the color crimson, aside from the associations with your memories, emotions, concepts, behavioral associations, etc. And if you ask whether other people have the same associations, we dissolve the philosophical question of whether the qualia are "the same", and replace it with empirical one of "how similar are they". We know how to tackle that one.
Let's say you were to take a set of 500 colors, and a set of 50 random memories you have, you could rate how strongly you associate each color with each memory on a scale from 1 to 10. This would give you a 500 x 50 matrix of association strengths, which you could think of as a 50 dimensional space where each orthogonal direction in the space is how strongly one of the 50 random memories is associated, and each of the 500 colors is associated with one particular point in this 50 dimensional memory space. But those points will not be randomly distributed within the space, and in fact you can probably map those points to a 3 dimensional space without losing much information. The position of colors within this 3 dimensional space would be a fairly faithful representation of the association of colors with those 50 memories.
If you were to repeat the above procedure with 50 random concepts you know instead of 50 random memories you have, you would also get a 3 dimensional space with colors in particular points within that space. Generally, I would expect that the positions of colors in this space generated by concepts would be pretty similar to the positions of colors in the space generated by memories.
Well now let's say we repeat this experiment with the same 500 colors, and the same 50 concepts, but a different person, Bob. I would expect that that person maps probably maps colors to concepts in a similar way, as long as they speak the same language and neither you nor Bob are colorblind. If crimson maps to a similar location in your color map as it does in Bob's color map, I think it's fair to say that you see a similar red to Bob.
This also tracks with how we teach colors to our children. We don't say "red is an ineffable experience which I experience and you might too", we say "red is the color you see when you look at a fire truck, or a stop sign, or a strawberry". This provides anchors so that our children know how to bind qualia to language. We can see evidence that they really do bind qualia to language in similar ways to each other too.
Take two kids, Alice and Bob. Teach them red by example. This fire truck is red. This strawberry is red. This stop sign is red. Teach them orange by example. This carrot is orange. This traffic cone is orange. This orange (fruit) is orange (color).
Take Alice into a room with many objects of many colors. Ask Alice to bring you things which are halfway between red and orange. Note the things she brought you, then put everything back exactly where it was at the start. Bring Bob into the same room, and ask him to bring you things which are halfway between red and orange.
Alice and Bob probably both chose similar things. They both took two of their qualia, interpolated an intermediate quale, and mapped that quale back to the physical world. When they did, they got similar results to each other, implying that their qualia were similar (unless Bob is colorblind, in which case they got very different results, implying that their qualia were very different).
faul_sname
Fuck around once, find out once. Do it again, now it's science.
12d ago
I imagined/conceived the above. So it has to exist. As all impossibilities exist [...] Everything Is Happening All At Once - Every Possibility Exists - To Stabilise the 0th Dimension
Under your world model, most observers would be Boltzmann brains". I observe that my experience is mostly ordered, where things that happened in the past are more likely to happen in the future, and past states of the world seem to influence present states of the world.
More generally, this does not seem coherent. Instead, it feels like you are going off the rails in the particular way that happens when your mind starts flagging random experiences and thoughts as "this was deeply meaningful".
If "sometimes I have an experience or insight which feels deeply meaningful for reasons I have trouble articulating" applies to you, there are a couple of things which might help:
- Get enough sleep
- Reduce stimulant usage
- Run ideas past others (importantly not LLMs, which have a strong tendency of agreeing with whatever you say. If you must use an LLM, try figuring out what the exact opposite of your idea is, and try running that through the LLM)
- Ask yourself "if this were not meaningful, what observations would I make that are different from the observations I would make if it were meaningful". Then actually look for evidence either way.
faul_sname
Fuck around once, find out once. Do it again, now it's science.
12d ago
I press x to doubt on the SJ debate remaining salient in the event of a full nuclear exchange
faul_sname
Fuck around once, find out once. Do it again, now it's science.
12d ago
From the page you linked:
Constitutions are chains with which men bind themselves in their sane moments that they may not die by a suicidal hand in the day of their frenzy.
Senator John Stockton, 1871
The constitution should be treated as much closer to a suicide pact than feels reasonable if you don't think about the long term. It is not a literal suicide pact, but if you want to do something unconstitutional and you also don't have sufficient support for the thing for an amendment, that's a sign that you should perhaps not do the thing, and instead you should do something constitutional, or you should change the amendment until it is popular enough to pass.
Moore's law, strictly interpreted as "number of transistors you can fit on a 2d die doubles every two years", will be running into physical limitations soon. Moore's law, loosely interpreted as "amount of compute you get for a given monetary or energy budget doubles every two years" is still relatively far from physical limits - Landauer is still about 6 OOMs away, so there's some headroom even without reversible computing. 6 OOMs of continuing halving of energy cost every two years would take us to 2065.
I'd expect further architectural changes to be required before then, because I don't expect the compute to be shaped optimally for the ML techniques of 2025, but "the SOTA ML algos are the ones that are able to take advantage of the hardware available at the time" is nothing new.
More options
Context Copy link