abecedarius

Maybe a better goal is some representation for "COCONUT [with these 3 letters occluded]". Then the consumer might combine this with other evidence about the occluded parts, or review it if questions come up about how accurate the OCR was in this case.

I already said ID is not security. How many 70s hijackers got away unidentified? How many would not have even tried if they also had to show ID? I can think of one case, "D. B. Cooper".

This conversation just doesn't seem promising. Your initial wildly-off claim got me worried I might misunderstand the world by that much (I remember the 70s, fuzzily). I wasted time checking it, wondering if the actual peak might be in a year not mentioned, trying to find the paper Wikipedia cited: it was a non-open-access paper with at least four links to check; one of them did show the first page, which at least gave the figure in question, if not the ultimate source.

I agree that Y is bad, actually.

https://en.wikipedia.org/wiki/Aircraft_hijacking cites two figures for the U.S. rate of hijacking attempts near its peak. The higher one is "Between 1968 and 1977, there were approximately 41 hijackings per year." meaning once per 9 days, an order of magnitude below yours. The other is once per 13.3 days over 1968-72. If both Wikipedia numbers are reliable then the peak must be somewhat higher than 1/9 in 73-77.

Which country was that?

ID is not security.

Yes, it was. One thing there seems to be more of is bundling of politics, like "Y is bad" as rejoinder to "X is good" with no necessary link between them.

I think I asked what deposit they would need. It had never been an issue before.

They divorced before I met the one. Whatever the reason might've been, it'd be something like this, so why ask? Eh, anyway. It's cool that more of these cases have come up here.

That reads basically like "if you seem cooperative enough with our extra effort to check your identity, we'll probably let you through". That's just a very different country from "you need greenbacks to buy a ticket over the counter."

Similarly though admittedly more permissive, with Amtrak iirc it's like: you need a credit card and they probably won't check your ID. There were several more years after 9/11 before the "papers please" principle came for the trains and buses too.

E.g. search turned up a claim Greyhound "employees are not allowed to sell a ticket without an official ID" https://travel.stackexchange.com/questions/140703/i-have-no-...

Reminds me of the O’Whielacronxes! Though I'd be surprised if that was ever their legal name.

There was a 70s science fiction future where customarily both parents adopted a new last name when they married. Seemed the obvious sensible policy for women's lib without a 2^N name-length scaling with the generations.

I only know one person who's done that. In some ways there's been less future shock than we expected.

I guess a very shallow angle of impact was part of it:

> brushed by the mountain summits of Malapert and Mouton

Similarly, 70s Steve Wozniak was supposed to be kinda chubby, but people this millennium seeing those old photos are like "what? he's totally average?"

Yeah, similar experiences on both counts in early 2000s. The hotel's refusal to honor a reservation even with cash in advance was part of the new "not the country I grew up in" feeling, those years.

The U.S. used to take pride in freedom of movement. In fiction, demands for papers were a classic sign of a government gone over the line. Now you need ID even for a Greyhound or Amtrak. I guess you can still walk or ride a bicycle.

We know how they're trained. We know the architecture in broad strokes (amounting to a few bits out of billions, albeit important bits). Some researchers try to understand the workings and have very very far to go.

> personally not like to call whatever that is science

A counterargument written pre-LLMs: https://arxiv.org/abs/1104.5466

(I think the difference between the OP's calculation and my estimate all comes from their 80% efficiency vs. my assumption of 100%. My 'roughly' twice as far as high is actually exact under ideal assumptions. Where the slope of the inverted parabola is -1, the chord is 4 times the height, but in this parabola only half of the launch energy goes into upward motion, accounting for the other factor of 2.)

Yep. But calculating what horizontal distance that implies, in your head, doesn't seem trivial the way my note above was.

Checking: it's 6 tons dropped 2 meters to launch a 100kg payload. If straight up, it could go 6000*2/100 or 120 meters. Visualize a launch for range vs. for height, and guess it can go roughly twice as far horizontally as up.

(The OP ends up at 210 meters.)

Yes, my judgement too from messing with Claude and (previously) ChatGPT. 'Ridiculous' and 'cultish' are overton-window enforcement more than they are justified.

You're right, it was kind of rude. Apologies. I really would rather be wrong, for a reason I gave in another comment.

The thing is, you can interact with this new kind of actor as much as you need to to judge this -- make up new problems, ask your own questions. "LLMs can't think" has needed ever-escalating standards for "real" thinking over the last few years.

Gary Marcus made a real-money bet about this.

I have a similar opinion of Claude Sonnet. Superhuman knowledge; ability to apply it to solve new math/coding problems at roughly the level of a motivated high-schooler (but not corresponding exactly in profile to anything human); less ability to stay on track the longer the effort takes.

But ChatGPT a couple years ago was at more like grade-school level at problem-solving. What should I call this thing that the best LLMs can do better than the older ones, if it's not actual reasoning? Sparkling syllogistics?

Sorry, that's sarcastic, but... it's from a real exasperation at what seems like a rearguard fight against an inconvenient conclusion. I don't like it either! I think the rate of progress at building machines we don't understand is dangerous. (Understanding the training is not understanding the machinery that comes out.)

Compare the first previews of Copilot with current frontier "reasoning" models, and ask how this will develop in the next five years. Maybe it'll fizzle. If you're very confident it will: I'd like to be convinced too.

Chiang makes some insightful points, e.g. about what we mean by magic.

Then I come to

> [LLMs] can get better at reproducing patterns found online, but they don’t become capable of actual reasoning; it seems that the problem is fundamental to their architecture.

and wonder how an intelligent person can still think this, can be so absolute about it. What is "actual" reasoning here? If an AI proves a theorem is it only a simulated proof?

Thanks!

Neat term, debaiting -- though I misread it for a moment.

And I was just trying to help. :) Easy to miss that link. Unfortunately Wikipedia generally isn't great for motivation on math topics.

Tensor ops in e.g. pytorch are pretty opaque to me, too much implicit on shapes which can change as you go down a pipeline. Maybe I'll come to appreciate it better.

HN guidelines: "Otherwise please use the original title, unless it is misleading or linkbait; don't editorialize."

Perhaps you would also describe the Manson family as a Christian death cult.

I should've said I didn't do any control experiment. I looked more closely at what Claude did in another case and it was console-logging some rough features out of the canvas data that time. If it actually was "looking at" the image the first time, it had to have been through a text encoding -- I think I remember a data URL briefly starting to appear in the transcript.

The submitter of this story inserted 'rationalist' into the title; it's not in the headline and IMO it's calumny. Would you take it back out?

See the link at bottom right to "Penrose graphical notation".

(I have nothing to do with the site.)

It says "In 2016, Congress passed Public Law 114-190, which among other things banned the use of biographical assessments as a first-line hiring tool for air traffic controllers."