Fully agree, although it’s interesting to consider the perspective that the entire LLM hype cycle is largely built around the question “what if we punted on actual thinking and instead just tried to memorize everything and then provide a human language query frontend? Is that still useful?” Arguably it is (sorta), and that’s what is driving this latest zeitgeist. Compute had quietly scaled in the background while we were banging our heads against real thinking, until one day we looked up and we still didn’t have a thinking machine, but it was now approximately possible to just do the stupid thing and store “all the text on the internet” in a lookup table, where the keys are prompts. That’s… the opposite of thinking, really, but still sometimes useful!
Although to be clear I think actual reasoning systems are what we should be trying to create, and this LLM stuff seems like a cul-de-sac on that journey.
Blame marketing, branding and pressure for value-add. There's no reason for anyone to have a VPN client you can see. It's all IPSec, OpenVPN, or wireguard under the covers. It should be handed off to the OS, but every corp needs to be a special little snowflake and show you their logo. (Yeah, I'm bitter, I fought with way too many crappy GUI wrappers for openvpn CLI) Once the app exists, there seems to be little pressure to make it actually good - one of them updated and literally removed the necessary client but just from the MacOS releases.
(ZeroTier gets a pass though because it's an uncommon protocol and I can't remember the last time I've actually seen the app open - as it should be.)
Strongly agree that this one is both very fun to think about and rings true. I sometimes imagine imagine it as parallel to the advancement of the world of computers, which has sort of been like watching the development an entire civilization in miniature. Early computer pioneers were, we know, incredible minds whose talent was the very thing that put us on the hard road to progress. Sure, nowadays, random people are able to casually accomplish much more in absolute terms, but it's because they're standing on the shoulders of giants.
All that said, I do tend to be sort of a Graham Hancock apologist. My take is that most people go too extreme with him. They either think he's a crackpot loony who must be taken at face value and debunked as a purveyor of pseudoscience OR they think he's a rebel truthteller: the only one who will look at the real facts, bravely pushing through the corrupt academic swamp.
It seems obvious to me that he's neither. He's just an author who stumbled on a compelling, mind-expanding idea. Roughly stated: what if we know less about the past than we think, and thus underestimate our ancestors? I think the interesting thing about Graham Hancock's spiel has nothing to do with any of his specific pieces of archaeological evidence that he digs up, which are very clearly marshaled to make a point he has already decided on making. (This is bad science, 100%.) Rather, the thing he brings to the table is more like a philosophical approach that is genuinely fresh and interesting. And I do think he will one day be vindicated in some way, because we act like we have way more precise knowledge about the past than we actually do. This is sort of an epistemology thing, so appealing directly to the evidence and the current anthropological understanding isn't really engaging with him in good faith. He's pointing out that the Troys of history prove that we consistently overestimate how completely we've understood history and what is and is not reasonable. Over time we tend to acclimate to that picture, and then the problem multiplies, because we tend to only accept things that seem to fit with the now-banal-seeming history we already know, leading to even more banal hypotheses gaining traction. Some of his best writings relate to the systematic bias against catastrophism that existed, and showing how these types of errors in epistemology lead to actual errors of science down the line.
I was going to write something similar: if I had been in creative control back in 1993, I definitely would have fallen for the trap of thinking that I was making things better by forcing the artwork to have a homogeneous, consistent feel to it. In retrospect, the idiosyncratic uniqueness of each card made it like a little piece of art that you could own. As a young kid, that made it so cool! I think a lot of young kids only bought a relatively small batch of cards and never even played the game much. I remember owning a scant few cards but that made me treasure my favorites that much more.
You should know: Eckert and Mauchly had already conceived the idea of the stored program long before "Johnny" arrived on the scene, but they made the pragmatic decision to finish the ENIAC under their original design. The stored program would have to wait until the next machine: the EDVAC. They were under constraints that prevented them from publishing the ideas, whereas Von Neumann was free to swoop in, grok the ideas (it was Von Neumann after all) and publish them without attribution. Though in fairness, he just wrote the draft - it was really Goldstine that put the icing on the cake by distributing the draft and ultimately cementing the term "Von Neumann architecture". Because he was an intellectual giant, it was easy for outsiders to just bin the whole thing as another one of Johnny's great ideas. Von Neumann was an incredible mind, but he and Goldstine are "the bad guys" in this story. Eckert and Mauchly have been written out of history despite being originators of perhaps the most important concepts underlying all of modern computers:
a. Make it fully electronic
b. You can store the program the same way you store the data
(You can quibble about whether Zuse or Turing described stored-program approaches earlier, but these were significantly more hypothetical / non-practical than the designs being discussed for the EDVAC. This is kind of a theme: there are many sideshows in early computing which were flirting with the right ideas, but the ENIAC and subsequent Moore school lectures stand out as the ancestors of everything that actually went anywhere. The cambrian explosion started there, which is why people get strung out about who gets the credit.)
> he (von Neumann) and Goldstine are "the bad guys" in this story
Von Neumann and Goldstine are clearly at fault in not crediting Eckert and Mauchly at all in the 'first draft'. The precise role of the players in developing the concept is disputed though. The 'first draft' is the first complete description of a stored program computer - there is no comparable document from Eckert and Mauchly - and it's inconceivable that Von Neumann didn't contribute significantly to the design.
And if we are focused on practical designs, then Eckert and Mauchly weren't even the first to implement a working stored program computer using the approach set out in the first draft.
We can disagree whether distributing the 'first draft' was morally right. My personal view is that doing so and hence invalidating the patents was a huge net good.
I hear "this put the computer in the public domain" a lot. But it was patented, and that didn't stop the digital revolution one bit.
Eckert and Mauchly filed a patent (on ENIAC) and it was believed to be valid for the next 25 years. That doesn't mean anybody tried to create a monopoly. IBM quickly traded some punch card patent rights with the owner, Remington Rand, for the (still pending!) ENIAC patent. Once it was finally granted Remington Rand pressed for royalties from the other computer companies.
Evidently they asked for too much, and that fomented Honeywell's suit to invalidate the patent, which succeeded. That, technically, put the ENIAC in the public domain, but by then there were thousands of other computer related patents on the books. They annoy, especially the bad ones, but business goes on.
What you're missing is that up until relatively recently, folks were still holding out hope that Reddit could remain a place that was enjoyable to interact with regularly. It's only the last 5 years that it has completely degraded into an obnoxious hellscape that you occasionally have to pinch your nose and deal with to extract some information that is unfortunately buried there. Facebook, on the other hand, degraded into a hellscape a long, long time ago. The facebook wounds have healed, whereas the reddit wounds are fresh. Plus, people liked using Reddit a lot more than they ever liked Facebook at their respective peaks. You're seeing people coping with a deeper and more recent loss.
I don't get where the stopping point is. When does using your resources to gain more resources start to become immoral? You concede that the nameless workers ought to be capturing some economic power, which acknowledges that it's a basic human right to accrue economic power. What you won't seem to entertain is the possibility that an investor is fairly accruing economic power by using his existing economic power to fuel another profitable enterprise. If you keep repeating that process, your capital compounds, but there's no principled reason to say that any step in that process involved extractive / exploitative gains.
Why don’t we start with just raising the floor on expectations for care such that it’s impossible to be involuntarily homeless or die from a preventable disease
We can probably figure it out faster once the majority of the population is not grasping for air
This is imminently possible today but the social will isn’t there
If you critically examine the source that provided you this number and follow the rabbit hole, you will probably learn that it is wildly pessimistic AND not telling you what you think it’s telling you. E.g. these estimates often include estimated casualties from sectarian violence that was merely uncorked by the Iraq war. You are clearly not taking pains to prevent the casual reader from assuming that 600k (already a specious number) were all directly killed at the hands of US soldiers.
> they're literally worse than Stalin in this regard.
> and accurate body counts.
Ironically, not really, the numbers people typically refer to were calculated by extrapolating birthrate dynamics, not an actual accurate body count.
Applying the same method to calculate the victims of the Great Depression will yield about 7 mil dead. I remember some Americans being real pissed when some Russian historian did this kind trolling, mirroring Stalin-related western propaganda.
There are only 160 million US passports in circulation (out of a population of 330 million. So roughly half. Of course HN people tend to be middle class or higher so they'd more likely be in that 50%. But still, the US is huge (roughly the size of Europe (Europe is approximately 10,180,000 sq km, while United States is approximately 9,833,517 sq km). There's plenty of places to go in the US without needing to go elsewhere. And look at the tiny passport numbers in the 1990s! That's because before 9/11 you didn't need one to go to Canada or Mexico if you had other identification (like a driver's license) showing you were from the US. I've heard that you can still go to Canada that way if you drive rather than fly, but I haven't tried it.
Until about 20 years ago you didn't need a passport to travel to Mexico or Canada. And passports only last 10 years so it is possible some have traveled outside the US in the past, but no longer have a passport.
Although to be clear I think actual reasoning systems are what we should be trying to create, and this LLM stuff seems like a cul-de-sac on that journey.
reply