Just this guy, you know?
Dagon
Karma: 14,192
Sure. My point was more along the lines of “fixed price throughout the day has been known to be silly for decades—what’s different now that would make power districts move to a more complicated and rational system”?
Most large-market subsidies are inefficient and imperfect—they’re always a political compromise.
Do you have any examples of someone arguing that net metering is even close to correct for any purpose besides political acceptability? It seems obvious that a move to demand billing (with rates that change every minute), and a legible spread for consume/provide rates is much closer to tracking real costs and values. But that doesn’t get anyone elected.
It’s not possible for all and only progress since ~1100CE to be wiped out. Some knowledge will be durable—possibly things like the germ theory of disease and similar. Some artifacts will be durable—evidence of cities and such. Raw materials may be less accessible (easy coal/mineral access, old-growth forests, etc.) or may be MORE accessible if they’re recoverable from the artifacts.
The weights of a frontier model are a LONG way from useful—possibly long enough that however they’re recorded doesn’t last long enough to use them.
httpi is a premature label for a set of semi-related huge projects.
no part of it is solved.
- zkPassport doesn’t solve any identity problems. It only separates identity (government-provided, very limited) from attributes like age.
- examples of contract-as-code or automatic adjudication are limited in scope and ALREADY are showing some fragility and error.
- enforcement (why would a consumer want this) seems to require government mandate or something.Some of these technologies and behaviors would be good for some interactions. It wouldn’t help very much for most types of agent misbehavior.
The sentence is the clearest indication of crank bullshit. The rest of the post doesn’t even acknowledge how far it is from a workable proposal, which also supports this.
we could say that GPT controls OpenAI at least as much as Sam Altman ever did.
it’s not plausible in the short-term, and long-term there are plenty of other attack vectors for highly-coordinated superintelligences. More directly, this is a much lower level of autonomous arbitrary control than the post seemed to be talking about. “leading and managing” has different implications than “weilding as tools of power”.
If you misbehave you lose money (a fine) or time (jail) or both.
Not even close. a tiny fraction of human misbehavior is legibly punished. We keep the world safe from humans by the fact that they’re generally not powerful enough to destroy the world. This has been changing in recent history, and we probably will NOT keep the world safe from humans.
I will now dive right into a concrete and crazy idea, however, I am trusting you to be smart and extract the general principles instead of focusing on the exact tunable parameters:
Yeah, this pattern-matches so strongly to crank bullshit that I am unable to take it seriously. This comment is a courtesy to let you know why you’re being downvoted.
httpi adds a third job: the client has to prove who it is too.
How? Even assuming the client WANTS to prove anything, how would it do so? Remove the passive voice from all of this—what human, agency, or AI is performing any of these validation/adjudication/enforcement actions?
AFAICT, you’ve written a bad sci-fi scenario, handwaving all of the difficult parts
free and reliable identity
free and perfect adjudication
side-effect-free punishment and enforcement
If you can actually solve, or even slightly improve, ANY of these, great. Otherwise, you’re not really furthering the discussion.
Do you have a gears-level description of what it actually means for “a model gets effective control of its host corp.”?
My experience and understanding of super-dunbar organizations is that they’re not controllable. Members and shareholders can influence them, but they’re too big and chaotic to quickly shift their culture or behaviors. This is the same problem in a different aspect as https://www.lesswrong.com/w/moral-mazes.
That said, “take over” in the minor sense of owning a lot of stock and being treated as powerful employees is certainly possible.
[ edit: fixed typo durbar->dunbar]
Do you have a better link for what “move 37” means than a piece of fiction? In what domain of risk assessment is it udervalued?
Thanks for the great example of utterly failing to address the problem. “if you HAVE an ought (called a preference), you can generate less-interesting oughts to support it”.
I like the style, and somewhat disagree with the substance. The hammock example is good, but a lot of people could be confused between static (force needed to hold it) and dynamic (force needed to raise it). you and I know that any tiny amount over the holding requirement will cause it to raise, but that’s not an obvious thing to everyone.
I think it’s not also clear in the hammock example how to multiply the effect. Once you add pulleys / block-and-tackle, it scales as much as you want—each additional loop adds to the effect; N times as much distance with 1/N as much force. This also maps to gear ratios and how they multiply if multiple gears are in play.
Much depends on the economics, and on the range of behaviors these robots can perform autonomously. I suspect there’s a long road to get them cheap enough to really replace human workers in most of the supply and maintenance chain, let alone “trillions of robots” (100 per human?). I could be wrong here—once they CAN replace all humans in the mining, refining, manufacturing, etc. of themselves, it could very rapidly take over. Much like AI self-improvement could very well foom.
Philosophy posts: where have you hidden your solution to the is-ought problem in all these words?
Thanks—this helps me clarify why the post didn’t resonate with me.
Summary: Individuals who act in ways that are harmful can be categorized into three groups based upon their motivation: Pure evil, or causing harm for personal enjoyment, instrumental evil, or causing harm to some other selfish end, and righteous evil, or causing harm in hopes of benefiting society.
Why does this categorization help in any way, and how do you know which category applies in which instance of behavior you dislike?
I think if you’re set on categorizing, I’d add a 4th: social evil—acting in ways that are socially advantageous, to increase one’s status, power and/or wealth. I can’t think of many cases where this isn’t a large factor.Many of history’s most evil people are motivated by righteous evil, contrary to popular opinion, which asserts that pure evil is a primary motivation due to reluctance to accept that anyone could be delusional, immoral, or harmful.
How do you decide what motivates them? how do you even separate them?
I’m a big fan of studying history and looking into what factors and forces made someone simultaneously powerful and horrific. I don’t think “evil” is a useful part of this study.
I think you should ask the people who say it’s low, how they know and what specifically they mean. Find out if they’re talking “awareness” or “understanding”—what’s the threshold for “knowledge of...”?
My impression is that many are aware of it—anyone who follows tech even a little bit heard about it, anyone close to coding or AI saw some details. My mother asked me about it. And there’s a pretty steep curve for who understands how much of it. 3% of gen pop understanding any LW topic would be asking a lot.
Am I reading this right (only the first bit—I don’t actually care that much)?
Trevor claimed you threatened to sue. Your post is that you didn’t sue, but did threaten to do so. So he was exactly correct in that part of his horrified reaction to your behavior.
Then you go on to say that you think your threats were proper because you consider his post to be libelous.
You probably just need to pick a different word, or use “legal contract” rather than just “contract”. it’s ambiguous in many people’s prior expectations.
I disagree about intention being out of scope for law—intention and mental state is inferred/adjudicated as part of MOST criminal prosecutions, and gets messily decided along with other facts about each case.
I am highly skeptical of any attempt to come up with a simple, consistent, and complete set of principles that apply to human behavior. Actually, strike that, s/and/or/ .
I lean libertarian, but I’m not dogmatic about it. In truth, there are almost no goals or behaviors that don’t affect others. Many in very hard-to-predict ways, many others in ways that benefit some more than others. The best plan seems to be let people do what they want, and then make them pay to repair (or to discourage) the egregiously-bad outcomes.
I’m confused. I think this is a mixed definition of “contract”, sometimes meaning “area of agreement/alignment”, and sometimes meaning “specific document with legal precedent and bureaucracy about interpretation and enforcement”.
Moral reasoning and political/social arrangements DO handle many things differently, at least in part because morality is personal and political/social is about enforcement on people who may or may not share your beliefs.
neither upvoted nor downvoted, but this would benefit from a TL;DR or thesis statement, and would benefit a LOT from some description of whether you’re describing a thing you think is real/objective, or describing a framework/belief system that some sub-group uses for some explanations/decisions.
I suspect the elitist-populist dimension (as you put it; in common use, they have slightly non-opposite connotations) is not central enough to most political coalitions to be important. It’s not elitists vs populists, it’s various conflicting groups of elites fighting each other and various conflicting groups who claim popular alignment.