“I want to be the one posting details about constructing WMDs”—you do? ;-)
Roger Scott
Karma: 47
Unless the terrorists in question are actually from, or, at least, acting on behalf of, the neighboring state said neighboring state would have a strong incentive (in the form of wanting to avoid a retaliatory strike from their neighbor) to not be a “host” for such a terrorist attack. That’s not to say it still couldn’t happen, but it would mean your neighbor would have similar incentives as you to prevent such an attach on you.
Flying cars weren’t a rational extrapolation of any existing or reasonably foreseeable technology in the 20th C. At most they were a very abstract extrapolation of the pace of technological progress in general, perhaps with a near-total disregard for more mundane engineering limits.
Are there LLMs that don’t do next token prediction? If not, isn’t the associated misalignment just a baseline for all LLMs and the other flavors are layered on top of it, not alternatives to it?
I think what you’re taking to be an attempt to excuse a problem is instead an explanation of why your apparent surprise is unwarranted. Nothing in the examples you give suggests denial of the problem, but the way you frame the complaints not only states the problem but attributes fault and presumes some “obvious” way things should have happened that would have avoided it. The pushback seems to mainly be on these latter elements.
The Viking I lander situation seems somewhat different in that the problem with the patching system wasn’t something that arose mid-mission. It must have been either a bug in the software written back on earth prior to the flight or an operator error, again on earth. In either case it could have been prevented by greater effort or diligence pre-launch. Physical/mechanical problems occurring mid-mission are not amenable to such prevention. A problem that shouldn’t even require one shot to discover seems qualitatively different from one that requires that one shot to even reveal.
One could argue that if the primary purpose of a government is to protect the lives and wellbeing of its people then the French course of action in WWII was far superior to that in WWI, given that, as a percentage of the pre-war population, French casualties in WWII were roughly a third of the WWI figures. There were certainly terrible consequences of French capitulation, but there’s no reason to believe similar or worse outcomes wouldn’t have resulted from putting up greater resistance.
You (implicitly?) raise an interesting point: is character, as we usually think of it, sometimes defined as much by what someone doesn’t know as what they do know? If so, someone who knows everything isn’t going to be like any character we’ve encountered in the human world. I suppose I should clarify that it’s not ignorance, per se, that shapes character so much as the pairings of knowledge and ignorance. To a significant extent a person’s knowledge (if interpreted more broadly than encyclopedic knowledge that can be gained without actual experience) is a proxy for their experience, so what someone knows (or doesn’t know) clues you in to what they’ve done (or not done). That all breaks with an LLM.
ChatGPT offers multiple “personality” choices these days. You don’t say which one you’re using, so I assume it is the default. Have you tried any of the others? If so, was your impression any different?
You’re only communicating if people are listening to you. If you talk in a way that causes fewer people to listen to you you’re being anything but more efficient. The goal isn’t to issue forth your great truths into the empty void—it is to be heard and understood by the greatest number of people. Tact is the avoidance of needless offense to other people. It is a cop-out to say that the only way something brutal or ugly can be communicated is to offend people in the process. What that statement actually means is that you can’t be bothered to find a tactful way to communicate.
I think if you actually dig into the facts here you’ll see that while there may be a gradual increase in restrictive laws over long periods of time the majority of restrictive measures introduced—just like the majority of nearly all measures introduced—fail. This is particularly true at the Federal level, while at the state and local level there’s a lot of variation from place to place.
Deflecting demands for proof, particularly as that term is understood in the “hard” sciences, is not unreasonable. Deflecting demands for stronger arguments is just intellectual laziness.
Are you equating access to information and morality?
“A fast government is more able to process new information, identify new threats and respond appropriately...”—I was with you right up to that last part. In what way is a fast government more able to respond appropriately?
If we accept your premise that slow governments immediately lead to human extinction that in no way implies that fast governments are any better. Yes, if a fast government made the right decisions it could, in theory, implement those decisions more quickly, but I see no logical reason to believe that a fast government is any more likely than a slow government to make correct decisions and quite possibly the opposite. I’m afraid I don’t find the contents of cyberpunk literature, classic or otherwise, to be a basis for large-scale political decisions or structures.
Just FYI, virtually no one who genuinely lacks tact believes that they lack tact. That’s the nature of the problem: there are societal norms which, if you’re aware of them, implicitly guide your behavior and, if you’re not, you will violate without even knowing it (pretty much by definition).
Just curious: do you limit all of your choices and activities to those you are certain are 100% safe?
Since we established right up front that power should be thought of a multi-dimensional (and I agree with that structure), it isn’t really meaningful to ask whether one kind (or linear combination of kinds) of power is more or less important than another.
”In countries where women are the most oppressed, feminism doesn’t succeed as a movement, and in countries where women hold a lot of power, it ends up being very popular.”—I find it misleading to compare “most oppressed” with “hold a lot of power”. At the very least we need to modify the latter to “hold a lot more power”, but far better would simply be to place these both along a power spectrum—more power vs. less power. The implication of “hold a lot of power” is that in these societies women hold more power than men, and that certainly wasn’t true in societies such as America in the 1960s when feminism gained significant popularity.
In general, I suspect that ideas such as feminism only become widely popular when they stop seeming outlandishly far-fetched, regardless of how appealing their theoretical objectives may be.
It sounds like you’re arguing that every restrictive law that’s ever attempted is eventually enacted. Is that really your claim?
Isn’t what information shouldn’t (and should) influence people constantly changing?
I’m no expert on international terrorism, but isn’t there a force opposing large scale, indiscriminate killing, namely that such behavior makes the terrorists seem less rational and thus less likely to alter their behavior in response to any “concessions” by their target? If it appears that all you want to do is randomly hurt people the only response you’re likely to encounter is your target trying to eliminate you.