VoxelVexillologist
๐บ๐ธ Multidimensional Radical Centrist
No bio...
User ID: 64
I'm not convinced of it's accuracy, but "the left proposes changes, not all good, and the right resists them, especially the worst ones" as a framework wouldn't be completely unworkable. Lots of continuous improvement sorts of systems work this way. We try things, and adopt them more widely when they work, and revert them when they don't. Our willingness to try things varies over time. To use a car analogy, not every model year improves over the last in every way, but most consumers probably prefer a 2026 Western (or even Chinese) model over an unchanged 50-year-old Lada design built this year.
It's not my favorite model, but I am not really in favor of letting either side of the CW declare moral victory.
now is the time to think seriously about going back to Reddit
I'd actually disagree with this. Reddit has recently ended login-free read access to the site. And it has lots of AI content filtering going on beyond terrible subreddit moderators. I see more than a few "removed by reddit" posts (for some reason, the posts sometimes get filtered after being viewable for a bit) with thought crimes like failing to use the latest TikTok filter euphemisms for violence ("unalive", "grape", etc) or daring to link to local news sites mentioning these things.
Now feels, if anything, a more dangerous time to move back: coordinating the move was difficult enough, but for the frog to jump back into the warming water would be a bad choice.
If anything, the declining amount of commentary seems to me downstream of an uneasy dรฉtente in the Culture War: there is just less drama to talk about. Normies seem tired of extreme takes right and left, and we're maybe begrudgingly coming around to the old "don't talk politics and religion in polite company" rules.
Most full-size residential fridge units have auto-defrost circuits. I'm not aware of having to run one manually before: it's probably not meant to be manually controlled like that. The little ones generally need to be taken outside to thaw occasionally. I believe heat pumps for HVAC need auto-defrost too.
Buy and configure a single main operational server, have the hardware send regular updates there,
Depending on the way the corporate network is setup, you might need one of these at each location: does a grocery chain have a corporate-wide VPN already? I assume not because then you could avoid this whole situation by never exposing the hardware to the Internet. Even then, locking down odd hardware in a subnet would be a good choice.
It's often said that the "S" in IoT stands for security, and it's not really wrong, but the way we define "security" does manage to be difficult to implement on embedded devices. "Oh just use TLS" isn't trivial: most configurations there will need certs that get updated regularly. Do you even trust your CA to remain on the approved lists indefinitely? Will your choice of RSA-4096 still be acceptable for the life of the unit (say, 50 years)? What's your plan for post-quantum cryptography on a device built in 2003?
Your fridge also suddenly needs a good clock source to check the certificate time limits: now you need to set that on the device, potentially every time it loses power. NTP works, but isn't foolproof or even secure in most deployments. These things aren't trivial, unfortunately.
In practice, a large, long-lived SSH key signed by corporate would work fine, but none of the normal tech stack is really setup for this today unless your users like command lines. I'm not saying it can't be done, but it's not in the normal wheelhouse for technicians who specialize in wrenches.
I share your ethical concerns about MAID, but in this case "botched" meant "patient didn't die" in the same way the execution failed.
I have no clue if the medications used are even the same.
because they almost always do.
Usually. But the Mark Sanford, uh, affair in 2009 was pretty memorable, for example.
A loving Harry-Dracina marriage
It's at least interesting that the (hetero) romance angle would play out somewhat differently if you gender-swapped Harry instead. But I don't follow fanfic, maybe that's been done before.
There isn't any method of execution that can't be botched
Are there any accounts of MAID being "botched"? Admittedly I haven't gone looking, but I'd assume they would be similarly effective but for the lack of medical qualifications on the death penalty side.
By contrast, in the space of all utility functions an ASI could have, the functions which are compatible with human thriving are a null set.
Is this strictly true? To most other types of animals on this planet, we probably seem like a super intelligence. To a whale, we probably look a lot like gods, and Lord knows we haven't always used that power for good. But, outside of Japan, Norway, and Iceland, our intelligence has also given us the motive to want to preserve in addition to the usual creation and destruction powers, and we've banned whaling, and the whales are doing better than a few decades ago. And for some species like the Great Auk, this realization has come too late.
I'm not saying I want AI to treat us like whales, but there are bad outcomes that stop short of extinction. The AI has already ingested our conservationist memes, but I suppose what exactly it chooses to do with them we can't guarantee.
AI Safety advocates (correctly) argue that even a 0.5% chance of the world ending is worth spending resources to study and solve.
Pascal has entered the chat, and would like to encourage us to spend large-but-finite resources on the worship of God because the positive odds of evading eternal infinite hellfire make it a utilitarian positive.
Less tongue-in-cheek: I don't think you're wrong per-se, and some resources are reasonable. But how does AI risk compare to, say, the risk of near-Earth asteroids that could end civilization?
We should maybe keep track of how this changes, and I do expect it to, but what fraction of the work that happens at, say, TSMC is fully automated? The raw silicon mining isn't. The trucks bringing materials to the factory aren't, and can't even fill themselves with gas. I'd even bet there are a few dozen batch operations in the fab that don't start themselves, if not more. The big powerplants that run the grid mostly have human operators. The repairs to all of this are planned and executed by human hands.
For now, at least, Skynet would lose because it depends on a grid and supply chain that humans could not be extracted from. I expect this to change, maybe even rapidly, in the next decade or two, but the starting point is so manual that we'd still be irreplaceable for a few generations.
Even while we're talking about automated weapons in the military, and automated aircraft, the guys on the carrier flight deck are numerous and aren't just there because it looks cool (admittedly, it does). It'll take a lot of effort to replace all of them.
You probably know more than me here, but it seems like the most interesting results would be bounding "intelligence" (which remains ill-defined) as a function of model size and training data size. There are some results in that direction, but no Shannon Information Theory class results.
It's be really nice to affirmatively prove "three sentences of training data and an arbitrarily large model cannot create the Omniscient Machine God." This seems an obvious result, IMO, but even "maximal intelligence scales sub-linearly with both training dataset and model size" would seem to dispell "FOOM" singularity concerns.
We have to deal with the fact that the human era of producing value is coming to an end,
Value is really an agglomeration of scarcity and social elements. Even today, it's possible to see a trend of concrete material wealth losing "value" to a human element. People often are willing to pay extra for functionally equivalent handmade products, farmers markets, and original art pieces. Food in terms of calories is valueless in a way our ancestors might not be able to comprehend, but there remains expensive food.
It seems quite possible that post-material-scarcity doesn't necessarily imply that human physical and emotional labor is zero-value.
The USA's Largest Employer of Mathematicians is not really academic, and almost certainly doesn't hire foreign nationals. I genuinely have no clue what their demographics look like, though.
I believe one of the AI companies published a break in one of the post-quantum crypto finalists (HAWK). It wasn't the one NIST settled on, but I bet they looked at those. Note that NIST published crypto standards have at least once in the past been resistant to attacks not yet published when they were chosen.
But also, the P=NP Millenium Problem, which hasn't been claimed yet, would have very strong cryptography implications depending on the direction that goes.
IIRC NY is a one-party consent state for audio recording. Such a recording showing up here might be the biggest surprise evidence since Kavanaugh writing down all his social interactions starting in his teens.
Power and status are highly contextual, and I've found general attempts to rate "privilege" don't map well to the full gamut of situations.
We have a few non-CW threads. Maybe they need more traffic. Although I would be cautious personally about crossing too many other domains there for opsec.
I hope the full text of her actual statements to police get released
It is interesting to me that all the media outlets claim to have the Title IX case file, which I assume they got from Jane Doe --- leaking it from the university would, AFAIK, be a felony under FERPA. I'm fairly confident she has, or could get, a copy of the police statement and full report. I haven't seen the media claim to have seen those. My best guess is that what we're seeing is heavily spun in the interest of getting a settlement, but I have no doubt there will be plenty of collateral damage.
She has given multiple versions of the events, and most of the citations here are to the least favorable of those. The DA's summary is probably technically correct, but I doubt it's the most flattering version of the original tale. The more recent version is viscerally unpleasant and I can imagine she feels today like it was exploitative, at least in hindsight. I'm not sure advertising her to the whole fraternity without her knowledge wasn't.
It seems quite possible to me she was sold a "modern, libertine, anything-consensual-goes sexuality" by the culture, only to run headlong into "lol, but not actually" in how the rest of campus reacted. It's unclear how she went to all the trouble of making a victim statement to the police without seemingly including the facts to make it a crime. I am really confused by that one, actually: is reading comprehension and general knowledge that bad among Ivy League undergraduates and police departments?
But it seems to me now she's been taken for a ride by the lawyers, and promised financial compensation, criminal charges, and fame anonymity by her lawyers with this highly-public strategy, who I suspect know deep down she'd be lucky to get at most one. I don't see anyone but the lawyers winning this even if she does settle with the university. Is settlement even an option with the other parties? The facts, even only her accounts, don't seem friendly at trial (not a lawyer, interested in opinions), and there doesn't seem a case for reputational recovery worth a settlement for the guys even if they had money.
All 8
Don't forget the Marine unlucky enough to have the same name as one of the 7 and his mother whose address was incorrectly listed in the docket.
I feel bad for Jane Doe, but those two didn't even do anything!
But at least based on texts, Male #1 didn't talk about.
Do we know which of the numbered parties sent the Snapchat message?
I suspect the open Internet is permanently dead.
I feel this, as someone who has been around it for like 3 decades now. I've watched the digital frontier get explored, claims staked, fences erected, gardens walled off, and IDs checked at the door. The last one is happening now, between various proposed legislation worldwide and informal methods used in the name of "fighting bots". Within the last few months, Reddit is no longer usable without an account, and I can't even consistently make Google searches in incognito mode.
Some things have certainly gotten better. Video was once a [BUFFERING] exercise in [BUFFERING] patience. Now it streams in HD while I download whole video games. A large local AI model can replace much of the information you'd be looking for anyway. But I do think something of value has been lost.
I'd love to see some good "closing of the digital frontier" stories in fiction.
I see where you're coming from, and it does seem relevant here but "kissing can be fine, but kissing and telling is rape" is not, IMHO, a viable legal standard. Especially since I'm not aware of it being codified anywhere relevant here. It's dangerously close to ex post facto to criminalize actions based on words separated by months or years, unless it was considered a separate crime.
And if this was truly the standard, going on a nationwide media court-of-public-opinion press tour as part of your civil litigation strategy is, uh, not a good look either.
- Prev
- Next

Note that "good" for progressives is very much a moving target, and that previously-good things sometimes become "bad". Not usually overnight, but see how "ending the drug war" was going to solve all our drug problems two decades ago, and compare to today when many municipalities and states have effectively unilaterally withdrawn from it. This has led to excesses that we hear a lot about today, and the vibe, from what I can tell, is shifting towards "maybe letting all this chaos happen in cities hasn't been compassionate" and even "the trains and busses should be safe and run on time". A century ago, progressivism was pushing Prohibition, because alcohol was "bad", although given cultural shifts in drinking before/after, I'm not sure I can easily judge that, honestly.
To be a bit generous, the worst excesses here seem to happen in areas for which there seems to be no good answer: it's nice to think that restricting the police will reduce violence and let drug addicts hold down jobs. But it's also naive to think that will wholesale fix our problems. And a lot of the excesses seem to be assumptions that going harder on the "nice" idea will fix it, when we're well beyond the diminishing point of return: "more school funding" when our failing schools already have 2-3x the per capita budget of the rest.
More options
Context Copy link