Average. It’s just an average. I haven’t verified whether the number is accurate (and often it’s probably debatable what qualifies as an empire and at what point it fell) but some empires lasting way longer does nothing to disprove 250 years being the average lifespan.
The second part of what you said is still entirely correct of course, that number has no real predictive capabilities for the collapse of the USA.
I never even thought it was that deep (idk if in other countries ppl go over it in school or something, I first heard of it online) so I never really understood how people are relating it to any economic system. All it’s saying to me is that one bad actor can be enough to ruin something for everyone - as far as I’m concerned it’s just prisoners’ dilemma in a larger group. So we need some way of enforcing that, if a shared ressource is vulnerable to singular bad actors (which isn’t all of them, e.g. some people abusing welfare doesn’t suddenly skyrocket costs), it won’t be abused.
Edit: just realized I forgot whether tragedy of the commons was about some few fucking up the pasture for everyone, or everyone slightly overusing it. The latter is ofc a bit different, but “ah I can cheat the system a little, I need it after all” isn’t an uncommon sentiment. That one usually just means you need a bit of a buffer, though, because most people won’t grossly abuse something. (And of course, it’s still quite independent of economic systems - regional software pricing for example is ultimately a capitalist thing to sell more, and yet would fall under this as it’s usually possible to get these prices from other regions.)
Das mit der sichtbarkeit gehört eigentlich bundesweit als Bedingung für die Zulassung, Ausnahmen nur mit triftigem Grund (trecker z.B. können das bestimmt nicht, kp wies bei LKWs aussieht, und manchmal braucht man vielleicht wirklich offroad kapazität und ein trecker wäre etwas overkill).
I think there’s a blurry line here where you can easily train an LLM to just regurgitate the source material by overfitting, and at what point is it “transformative enough”? I think there’s little doubt that current flagship models usually are transformative enough, but that doesn’t apply to everything using the same technology - even though this case will be used as precedence for all of that.
There’s also another issue in that while safeguards are generally in place, without them llms would be very capable of quoting entire pages at least of popular books. And jailbreaking llms isn’t exactly unheard of. They also at least used to really like just verbatim repeating news articles on obscure topics.
What I’m mainly getting at is that LLMs can be transformative, but they also can plagiarize. Much like any human could. The question is then, if training LLMs on copyrighted data is allowed, will the company be held accountable when their LLM does plagiarize, the same way a person would be? Or would the better decision be to prohibit training on copyrighted data because actually transforming it meaningfully can not be guaranteed, and copyright holders actually finding these violations is very hard?
Though idk the case details, if the argument was purely focused on using the material to produce the model, rather than including the ultimate step of outputting text to anyone who asks, it was probably doomed to fail from the start and the decision makes perfect sense. And that doesn’t seem too unlikely to have happened because realizing this would require the lawyer making the case to actually understand what training an LLM does.
Yes, but that doesn’t make the comparison to all countries with over 500 000 people meaningful. It’s specifically that part that seems dishonest to me.
Though I suppose it is also possible that the full data has a few states where incarceration rates are more around the global average, which then would actually have a point in including other countries. Those weren’t part of the image posted here though (which was also dropped without context as to why it was posted)
Edit: yknow it occured to me i could click the link and yea, some states are indeed more normal, though still kinda high. That’s really the interesting part far more than the top of the list.
Yes, but that is not how the graph is framed. It’s framed as “look, if we put US states on a graph with other countries, they have such a high incarceration rate that there are almost no countries even on the graph!”
If it was honest and just trying to compare the incarceration rate of US states amongst each other (and the national average) it wouldn’t be titled “[…] in U.S. states and all countries […]”. It’s a clearly manipulative title.
The reason that a graph with this title could maybe make a point if it was absolute numbers is that most U.S. states’ population is less than most countries, so if individual states were still high on such a graph, that would be shocking.
They are, and I agree it’s misleading. It’s implying that it’s somehow shocking that the individual states of the county with the highest incarceration rate in the world also have a high incarceration rate. If it was absolute numbers, it would maybe make a point. As it is, it’s stating the extremely obvious and framing it as “look, it’s even worse than you thought”.
Interesting, here in germany I’d say that’s like a medium level swear (the sexual ones are what I’d consider high level, the religious ones usually no one cares about, at least in the north).
I wonder if the 3 usual categories - religious, sexual, fecal - can be approximately ranked (at least with one being the clear “worst” or “least bad”) in every culture or if sometimes it’s all over the place.
It’ll never stop being funny to me how in the US it’a apparently considered pretty bad and even gets censored while here in germany it’s one of those child friendly swears that you expect kindergarten kids to say.
Tbh, just because I’m capable of writing well, doesn’t mean I always do, especially online. If OOP is an english teacher it’s a bit concerning since not making it a run-on sentence would just be better storytelling, but otherwise eh.
Get your own domain and use it for mail routing to whichever email service of your choice. Afaik gmail offers this, and so does probably any other decent email provider. That way if a provider turns to shit, you just need to set up with a different one, but don’t have to change any accounts.
Downside: you will have to pay for that domain for the rest of your life (or change all accounts again)
I’ve been meaning to do this for a while, maybe I’ll finally do it now.
Oh damn. Very good article btw.
According to numbers floating around online, thiat would mean one llama query is around as expensive as 10 google searches. And it’s likely that those costs will increase further.
It still seems like the biggest factor here is the scale of adaptation. Unfortunately the total energy costs of AI might even scale exponentially since the more complex the queries get, the better the responses will likely be. And that will further drive adaptation.
This pace is so clearly unsustainable it’s horrifying, and while it was obvious to some degree, it seems it’s worse than I thought.
Using it, not all that energy intensive (one llm use is roughly the same as 3 pre-ai-bullshit google searches iirc). Training it, very energy intensive.
Yes it would but we haven’t even replaced all our previous needs with renewables so it aint helping.
Da gehts ja auch um Geld, und nicht sowas unwichtiges wie Menschenleben.
/s wenn das da noch nötig ist.
It still hurts russia. Azerbaijan (or any other country proxying russian gas) will take a cut, and russia will almost certainly end up with less money for their gas than before. Strategically it’s still not a terrible move. Long term we just need to phase out gas entirely, but that necessarily takes time.
Should’ve started a decade or two earlier, but you can’t change the past.
I didn’t actually know this (all my math knowledge comes from what my cs degree forced me to ingest) but google says yes (since natural numbers, being a countably infinite set, are apparently an example of the smallest possible cardinality of infinite sets, so any infinite subset of natural numbers is always the same cardinality.)
The answets to this stackexchange question go into detail but that’s beyond what I understand without a lot of effort tbh.
Funnily enough that’s an example of two infinities that mathematically have the same cardinality (which is very often conflated with size since for general mathematical purposes that’s what it is) since you can map a bijection (i.e. every number in the first set has one and only one mapping in the second and vice versa) between the two (and it’s as simple as f(n)=2n).
And intuitively that makes about as much (or rather, little) sense as the infinite hotel.
An example of infinities with different cardinalities would be rational numbers vs natural numbers.
Many countries have restrictions on what you can and cannot post (hate speech being a common one). Turkey in particular has been moving towards autocracy over the last decade or so, so I wouldn’t be surprised (to be clear this is speculation feel free to correct me) if it had restrictions on lgbt issues or political dissent or something.
This is quite literally the opposite of how you actually get people to support your cause. Ask any psychologist.
Shutting them out completely might work fine when they’re a tiny minority, but when in some cases a quarter of the population agrees with them enough to vote for them, doing that is simply impossible. They will have reach.
I also still think we need people throwing their morals about manipulating people to the wind and starting to peddle left wing conspiracy theories just like the right wingers are doing. For example with how perfectly it fits there really should be an actual movement behind the whole trump=the biblical antichrist thing in the US, but I’ve only seen it as satire in spaces that are already left wing.
I mean yea that doesn’t surprise me in the slightest honestly, even outside of the number itself being pretty meaningless in the first place it’s very fuzzy what the actual dates are.