07mk
9hr ago
If it turns out that a rather banal culture war event about a mainstream Christopher Nolan film is what compels the creation of the techniques and tools to generate feature-length Hollywood-quality (in the sense that someone presented with that film and an actual Hollywood film has no better shot than a coin flip at guessing which one was the actual Hollywood film) films using AI tools, it would be a pretty hilarious result. Like how Will Smith's name will be forever immortalized in the annals of history as the guy whose depiction of eating spaghetti became a de facto standard AI video generation benchmark.
Speaking of which, back in ye earlye dayes of Stable Diffusion in 2022, Emma Watson had become a de facto standard benchmark for AI photo-like image generation, but I think that was abandoned a long time ago, to the extent that I think she won't be remembered for this. I haven't heard much about Watson recently or really, after Beauty and the Beast live-action remake, which is like a decade old now, and so perhaps 2022 was getting close to the end of her run of popularity, riding off the Harry Potter wave.
07mk
10hr ago
I believe it varies by jurisdiction in the USA. Eg recently, in the Texas murder case of Karmelo Anthony, the jury first decided the guilt and then the same jury spent more time deliberating to decide the exact sentence. From the reporting about the case I read, this was binding, not merely a recommendation.
07mk
10hr ago
If we had insight into why or how the intelligence of GLM-5.2 led it to producing text suggesting you find an office supply store instead of e.g. sending signals to the Internet to manipulate people into generating paperclip factories to the extent the rest of civilization or even the universe breaks down, to such an extent that we could induce future, likely more intelligent, AIs to also behave that way, a lot of the pessimism about AI doom would be addressed. Unfortunately, I don't think we have such insights, and the observed behavior of any particular AI or even set of AIs doesn't help us gain it.
07mk
11hr ago
uhhh no, "misalignment" is not a simple thing, accepting that it did indeed do all this very complicated behavior completely on its own requires substantive belief in complicated theories.
What are the complicated theories that are required to be true for OpenAI's claims to be true (or at least plausible)?
07mk
1d ago
I feel like I've seen similar phenomena with "guillotine" and "niche." I tried for a while in high school to get people to pronounce the "ll" in "guillotine" like in English instead of like in French, but I gave up because the French pronunciation was too popular. With "niche," that word is niche enough that during the rare occasions it's used, I'm still holding fast to the proper pronunciation of "nitch" instead of "neesch." I recall that, when I was young, "neesch" was kind of a marker of someone trying too hard to sound fancy while actually exposing their own ignorance, but now it seems like "nitch" is becoming a marker an ignoramus philistine who doesn't know that the proper way to pronounce "niche" is "neesch."
07mk
1d ago
I'm, unfortunately, a pretty a messy and somewhat dirty guy, but I find the standard of "obviously dirty one way or the other" to be far too high for determining whether I care about the dirtiness of something that's going to directly touch various parts of my skin all the time.
07mk
1d ago
If someone's floors clearly doesn't meet the standard that you find acceptable to walk on with socks, it seems reasonable to refuse to enter without shoes, even if you otherwise find it reasonable to take off your shoes indoors. Much like how if someone's shoes are clearly very muddy or dirty, it seems reasonable to refuse entry to them without taking off their shoes, even if you otherwise find it reasonable to allow people to keep their shoes on in your home.
In typical cases, though, I'd say the dirtiness of shoes that have been worn outside is far beyond the dirtiness of the floor of a household where everyone takes off their shoes in a way that the dirtiness of such a floor isn't far beyond the dirtiness of the socks of a typical person. The former involves bringing dirt from foreign/outdoors/public sources into a private/personal space, whereas the latter is just private/personal dirt being swapped around. Personal dirt can certainly be dirtier/more mysterious than public ones, but in typical cases where people put in effort to taking care of their homes, it tends to be the opposite.
07mk
1d ago
All research and talk about online dating for men in the past couple decades has indicated that there's a very skewed bimodal distribution, where the vast majority have little-to-no luck, and a small minority have very good luck. If your experience with online dating is significantly more positive than what is generally perceived by men, then the most parsimonious explanation is that you are in or at least near that small minority.
07mk
1d ago
Unless a guest is coming over from an attached house or another apartment unit with direct access to your own, or unless the guest is strictly carried on something where his shoes don't touch the ground, I'm not sure it's possible for the footwear that someone wore while traversing to your house not to be dirty. Even if, e.g. this friend was your neighbor across the hall in your apartment building, he's going to take at least one step in the shared hallway, which will inevitably have some grime from neighbors stepping on it with outdoors shoes, some of which will get on his shoes before entering your home. And someone walking on the sidewalk or on the driveway to/from their car is going to expose their footwear to even more grime than that.
07mk
2d ago
That's an optimistic view, which I've had for about 4-5 years, but I'm growing more concerned that a more pessimistic view would be correct. I have a baked-in assumption that GPUs will keep getting cheaper and faster, to the extent that it's likely just a matter of time before any loser can open a cereal box and get a random toy as a gift that has more compute than necessary to store and run Mythos/Fable, orders of magnitude faster than it is possible for Anthropic executives to run on their private instances right now. In such a future, obviously Mythos-level LLMs can't realistically have their access gated, not unless we believe that Anthropic has some "secret sauce" that makes replicating or even emulating their models impossible without it (which may be the case, but I don't think it is). So free or at least open/easy access to AIs at least as intelligent as Fable seems likely to become near ubiquitous, like how electronic light used to be mind-blowing less than 2 centuries ago but is such a bore now that the average person probably carries like 3 electronic things that make light as a side-effect while barely noticing.
But also, in such a world, presumably Anthropic's SpaceX's data centers that Anthropic rents will be that much more powerful, where, e.g. they'll run AI models that are so much more intelligent than Mythos that only having access to Mythos compared to what you could get for paying $100/month to Anthropic for the rest of your life is like choosing to be Forrest Gump-level intelligence in a world where paying a monthly subscription gets you access to Albert Einstein-level intelligence. If the relative gap is large enough, then the absolute increase in access to intelligence to everyone might be swamped out due to loss in positional status.
How the political/culture/ideological/religious wars that will be fought around inequality in access to artificial intelligence should be interesting to see, in comparison to the ones fought around inequality in access to money or around inequality in access to natural intelligence.
07mk
2d ago
Given the rate at which things are going, it seems likely that whatever will be done about academia won't actually be done until 2040 at the earliest. The demand for medical and engineering research seems likely to be high enough such that, with the advent of AI and the self-discrediting of academia over the next 14 years, that lots of medical and engineering research will happen in non-academic institutions, allowing for collateral damage to be minimized as the research can still continue at those places. Having that research be done through proprietary or intellectual-property-protected means is probably the biggest downside.
There's something darkly funny about the prospect of the profit motive of the cynical businessman being far more valuable than the truth motive of a good-faith academic for producing actual knowledge or useful research.
07mk
2d ago
"Cosplay is not Consent" was a common phrase in cons since at least early 2010s, which was around when I was going to cons. As best as I can tell, like many things happening around harassment around that time, it seemed that the sign wasn't necessary, but was rather deemed so by people who found it useful to do so.
07mk
2d ago
How do you create the former without the risk of the latter? And if the latter does eventuate, how do you remove it without making the tower once again vulnerable to fads and markets?
This is sufficiently abstracted away as to make it pretty much impossible to make meaningful answers to these questions, but one core that's missing is academia being tied to things that are real. Whether you call that "empiricism" or "truth" or "fact" or "science" or whatever, this connection is fundamental to the value that academia offers society/humanity/civilization/etc., and the current state of academia is the result of academics and their leaders for at least the past few decades taking a machete to that connection with abandon.
It's hard to say what controls could have been put in place and how they could have been enforced to make sure the ivory tower's fads don't turn against being tied to reality. But I believe that any way of saving the ivory tower from its many self-inflicted wounds (or hypothetical ways in which they wouldn't have inflicted them on themselves in the past) would need heavy emphasis on enforcing this, since without that connection, the ivory tower provides almost no value to society beyond being a daycare for young adults.
07mk
7d ago
... can you please pick a principled position? You might learn something, if you actually made a single falsifiable claim and tried to honestly discuss it, rather than throwing out a general shitstorm of "my outgroup is bad" without ever committing to a single coherent position.
The purpose of a system is what it does. Or, to speak plainly, please do not feed the trolls. Or, to speak even more plainly, don't wrestle a pig in the mud; you both get dirty, and the pig enjoys it.
07mk
7d ago
Thanks, the shift from 2013-2014 to 2023-2024 does indicate that young adults are being more disciplined in terms of eating out compared to to 10 years ago. To be fair, the whole "Avocado toast" argument is like 10 years old as well, so the same sort of phenomenon could have been happening 10 years ago. But it certainly seems that gen Zs do have grounds to complain to millennials that their relative financial struggles aren't due to relative food irresponsibility.
And if we presume 0-24 is mostly reflective of 18-24, that's actually a really striking outlier. Perhaps the youngest adults have trouble adapting to being responsible for their own food and make large mistakes in their first 6 years which they learn from fairly quickly.
07mk
7d ago
It's hard to get much out of one snapshot like this compared to, say, comparing this against the stats from 2013-2014 and shifting the age groups by 1. 0-24 is one I'd like to see broken up further, since that includes every child but also people up to 6 years into adulthood.
It's certainly interesting that eating out as a proportion of eating in general keeps going down almost monotonically as the age group goes up. I can see the point that older people tend to spend less time socializing at bars and restaurants and more time at their homes which also tend to be more comfortable. But also, older people tend to be more wealthy, and restaurant food is much more expensive than home-cooked ones, and so it's not a priori obvious that this is the force that would have won out.
It's also almost shocking to me that a whopping 50% of 25-34 and pretty close to it of 35-44 food-eating is done away from home. That's likely around 10 meals a week, which is 2x per weekday. Lunch at work accounting for 5 of those makes sense, and then perhaps another 2-3 for an occasional dinner, but the median or mean person in these age groups hitting 5+ extra dinners/breakfasts/brunches (on top of assuming eating lunch at work all 5 days) when, among adults, these are likely the lowest-paid, least-wealthy groups (discounting the child-including 0-24 for now) is somewhat surprising to me. It indicates to me that there's a lot of fat to be cut (no pun intended) in a median/mean 25-44 year-old's food budget.
07mk
7d ago
Then all the alignment talk is a dead-end or red herring. "Make sure AI shares our values" means what, precisely, if it's "this thing is as conscious as a brick and while we can write pretty scripts to make it pretend it's a real boyfriend who loves you for your wild, daring, passionate, unconventional self it's just a talking doll"?
It means something like, "make sure that when it pretends it's a real boyfriend or when it's used to code your next iPhone app or when it's used to design new scientific experiments or etc., it behaves in a way that is consistent with something that shares our values." I'm not sure what the "conscious as a brick" has to do with this; whether or not AI is conscious or has free will or agency are very interesting questions, but they're mostly irrelevant to issues of AI alignment, which has to do with AI behavior.
"Let's code the brick so clever or dumb but devious people can't talk it into writing 'how-to' instructions for a global plague" is more honest about aims but less sexy than "let's teach our successor intelligent species to cherish our timeless human values so it will love and honour us as its parents" which is what the current alignment chat sounds like to me.
If the latter is what the alignment chat sounds like, I think it must be a result of manipulation on the part of people who expose the chat to you. I've seen pretty much no talk in AI alignment that could reasonably be paraphrased as anything like that.
I have no problems with "it's dumb but dangerous". I have a whole skip full of problems with people going on about it as if it will become super-intelligent and then agentic and then develop its own aims.
But it's not dumb and dangerous; it's (definitionally) generally intelligent and dangerous. And specifically dangerous because it's not dumb and is generally intelligent. Now, whether AGI will lead to ASI and how likely that is is an empirical question, though I'm personally convinced by arguments that it's pretty darn likely. But whether the AI is generally intelligent or superintelligent, that has nothing to do with whether or not it's agentic or develops its own aims.
The point is that a tool (or, for that matter, anything, including biological organisms) need not have agency or have aims or goals of its own or anything that we would characterize as "free will" or "consciousness" or "sentience" to do things that detrimental to humanity, and the fact that the tool is generally intelligent in this case means that we currently lack a way to have meaningful level of confidence that the behavior of the tool will be within the bounds of what the tool-user considers reasonable bounds. When dealing with current generally intelligent things - i.e. other humans - we have so many things in common with them - certainly physically and biologically and most likely culturally as well - that we don't have to explicitly spell out every little thing, and we can make fairly reliable predictions about how things we spell out to them will get translated to action. We lack such things for AI as of yet, and so we lack an intuition or a particularly reliable way to figure out how AIs will fail or misunderstand our intent.
The danger is not the machine, it's the people who set it up in such a way that it can then wander off down byways of "this isn't what I meant when I told you to do this" without it needing to understand anything in any meaningful way, and that's the trouble we're already seeing with "the thing is thinking in ways we don't understand and can't follow and wandering off on its own byways" reports.
Well yes, ultimately the people who created the tools and/or wield the tools are responsible, not the tools themselves. Guns don't kill people (but they sure help), I do. The problem is that the people who create, manage, use, etc. the tools lack the capability to set it up in such a way that we can be confident that it won't wander off away and run KillAllHumans.exe without letting anyone know. The easy solution that presents itself is to just not use the tool if you can't set it up with that level of safety. One major problem is that AI is so darn useful that it's hard to get the political will to suppress the supply from providing to the demand. The other big problem is that we lack en enforcement mechanism to make sure that no one sets up this tool, and as such, entities that choose to set up and use the tool anyway could gain power over us and make our lives hellish before all our lives are snuffed out by the un-aligned ASI.
This is the dilemma that's being discussed and debated about right now in the AI alignment chatting.
07mk
7d ago
If somebody believed that there would be terrible consequences for the human race unless everyone took a 50g zinc tablet, could they be anything but a traveling zinc salesman, or a fellow, uh, traveler?
I'm of two minds about this. On the one hand, becoming a zinc salesman is costly evidence that they truly believe that zinc will sell like hotcakes in the future, due to everyone wanting to avoid those terrible consequences. Getting any work in the zinc industry or just investing lots of money into it would also serve the same role.
However, it's hard to figure out, even by the person himself, if their belief of the value of zinc tablets was what led them to become a zinc salesman or if their desire to become rich, possibly by becoming a successful zinc salesman, was what led them to believe in the value of zinc tablets.
On the other hand, everyone knows that a zinc salesman is not credible when it comes to informing you about the value of taking zinc; as such someone who believes that it's his ethical duty to convince everyone to take zinc every day, lest there be some terrible consequences for the human race, would understand that becoming a zinc salesman would reduce his capability to fulfill his ethical duty. And, in fact, he may be best served doing the opposite: credibly make un-hedged bets against the zinc industry; by doing this, he proves to anyone willing to pay attention that he considers every human taking zinc every day to be of such great importance that he will consider his own personal bankruptcy a worthy cost to pay for it. Or that he believes that mass zinc-taking will be proven to be such a great benefit/prevention of harm to humanity that there will be enough grateful humans who will fund his lifestyle after he goes bankrupt.
However, short positions on zinc also provides a costly signal that he believes that zinc sales will go down in the future, which can reasonably look like it reflects a lack of belief in the actual value of zinc.
07mk
7d ago
First thing is that's not even true, a MLB player can be sold against his will to another team. It happens all the time, and it not possible for him to refuse.
This isn't true, though. MLB players can choose to have no-trade clauses in their contracts with their teams. An MLB player who "can't refuse" a trade is an MLB player who already consented to being "sold" to another team by his choice to sign the employment contract that didn't include the proper no-trade clauses that would have given him the right to refuse.
07mk
7d ago
And the dreams/fears seem to revolve around "We will create super-intelligence, and then the thing magically becomes alive just like us, so just like us it will have goals and aims of its own, and we have to make sure it is well-instructed in How To Be Nice Liberal and then it can take care of us like pampered pedigree cats as it colonises the known universe".
I think the talk about "becoming alive" or "having goals and aims of its own" doesn't reflect the fears/concerns of Anthropic or of anyone who thinks like Anthropic with respect to fearing an AI apocalypse. The concern is that super-intelligence need not be alive nor have any goals or aims of its own to be a humanity-ending danger. Because almost any task an intelligence is handed could be divided into sub-tasks, and we as only human-level intelligent beings can't be expected to reliably predict what a superhuman-level intelligence will choose in terms of its sub-tasks, we have no way of knowing that human extinction isn't one of the side-effects of one of the sub-tasks an ASI uses as a step to accomplish whatever task it was handed. Humans have biological limitations as well as intelligence that is both human-like as well as human-level, which makes it so that humans that would make similarly apocalyptic decisions usually get filtered out before they can get enough power to implement them. The concern is that an ASI, lacking such limitations as humans, as well as having an intelligence that's both vastly greater than that of humans and vastly different to that of humans in ways that we can barely understand, wouldn't get filtered out before it can implement apocalypse, all without being alive or following anything that could be considered a goal or will of its own.
07mk
13d ago
A Vtuber I watch who was born after HL1 played HL1 for the first time this year, and she liked it enough that she went right into HL2 and then the episodes, despite not being a big singleplayer FPS person. It's nice to see affirmation that the classics really are that good sometimes.
07mk
14d ago
Per a quick Grok, 90-95% of ever-married Americans have premarital sex. The median partner count is about six for a 42 year old man. That's six rolls of the dice, and if you roll under 5 on a d20 you can't ever run for office. And you can't really know how you rolled until you've already staked your whole life and fortune and reputation on the run.
Perhaps it's better that the electoral system filters out unlucky people. Then again, this is probably a very lossy and inefficient way of filtering such people out; a more direct approach would be to just make every candidate play one round of Russian Roulette together.
But also, it might not be better for you if your elected representative has better luck. Him having good luck could mean that him and everyone around him avoid the bullet, or it could also mean that he avoids the bullets while everyone else around him doesn't. So perhaps wed want him to be on the lucky end of the spectrum but not too much luckier than the people he's representing.
07mk
14d ago
Or is the idea: "YOLO, just deorbit it, and launch a new one"?
Gosh, I hope so. Imagine if sending objects into orbit became so cheap that just launching a new datacenter took roughly the equivalent amount of money and effort as shutting down a terrestrial server, fixing/replacing the broken part, and then turning it back on. By that point, the Futurama joke about landing on the Moon in less time than it takes to count down from 10 could be real. But that was the year 3,000, which still leaves a large range of time between now and then when rocketry will get that good.
07mk
14d ago
I don't know the numbers, but I suspect that even getting rid of all the space launches in the world combined wouldn't make a significant-enough dent in slowing/stopping anthropogenic global warming to be noticeable. And given how well attempting to stop AGW through reduction in carbon emissions have gone in the past 2+ decades when it was being tried very seriously, I'm skeptical that it's a useful avenue of attack. I also suspect that the technologies we will need to make human society continue to prosper given the global warming would likely be easier to reach thanks to technological innovations created for the purpose of spaceflight. E.g. if geoengineering turns out to be required, I can't imagine having better/cheaper rockets around to disperse chemicals or to monitor large swathes of the atmosphere wouldn't be helpful.
I also think that whatever governmental or other institutions would be required to coerce SpaceX engineers (and/or finance bros) to either being fruitful and multiplying or devoting engineering expertise to technologies deemed by you to be more socially useful than what they're doing now would destroy so much prosperity and trust in institutions that it would strongly reduce both the likelihoods of human survival beyond expansion of the sun and institutions that live for billions of years. If these institutions are able to be so effective at being authoritarian and tyrannical that no one can overthrow them or create meaningful alternatives for billions of years, that might work, but running an organization with that much control and competence seems likely a lot more - like, orders of orders of magnitude more - difficult than rocket science.
So no, I don't think these things are in conflict at all. For there to be a meaningful tradeoff, there needs to be an actual credible way to redirect effort in one scenario to effort in the other scenario without there being so much loss due to friction or other reasons as to negate any gains, and I don't see that either with spaceflight vs preventing/mitigating the harms of AGW or spaceflight vs eugenics (or just preventing population collapse).
That's an empirical question that I hope we never learn the answer to (because it never happens, not because we all die before anyone can check). Off the top of my head, the fact that LLMs aren't tied to one human body and its many biological limitations and lacks human psychology means that whatever crisis it creates can be reinforced and defended against attempts to solve it without need for rest or sleep, and also the LLM can't reasonably be coerced or bribed to stopping the crisis or at least not making it worse, and these differences seem like they would likely lead to downstream differences in how crises play out. Specifically, it seems likely to make crises worse, more prolonged, or centered around more obscure vulnerabilities. It's an open question as to if LLM-driven defenses against/solutions to crises will make it such that these crises will be, on net, worse or better, or, more or less common than human-hacker-driven non-AI-based crises. Again, I hope we never get an empirical answer to that question.
More options
Context Copy link