recursivecaveat

Don't forget: you can accept bribes as long as you receive them after the fact https://en.wikipedia.org/wiki/Snyder_v._United_States I guess I will never pay another restaurant bill again, since any possible future "gratuity" I might award them would of course have no bearing on their past decision to bring me a plate of food.

[flagged]

When I was moving into a dorm, a few totes would have been a lot of my available space. I'm pretty glad my parents would let me pick and choose before clearing stuff out. Saving my childhood stuffed animals or favorite books so I can share them with my own kids someday is pretty meaningful to me, same thing I remember my parents doing. I don't know if I would have had the correct perspective to make those choices as a university freshman. Unless you immediately downsize after becoming an empty nester you're going to have way more storage space than you need anyways.

What's your use case for it? For manual diagrams, making an explanatory diagram for a presentation or something I would never recommend it. Declaratively typing out all your information is going to take just as long as setting it directly in drawio, but with less control over the result.

Where graphviz/dot has always shined for me is you have some application built around a data flow graph and occasionally need to debug a run with 50+ nodes at a time. So you write a file dumper for your structure. A visual makes understanding the structure so much faster, and nothing else seems to produce layouts nearly as good as dot.

98% sounds like "nb, almost a perfect score". 12 hours sounds like "wow I lost a whole day this month", way worse.

Agency and information in a way. I actually feel safer in a tiny 4 seater plane than a jet, even though objectively that's probably backwards. Being able to look at all the parts of the plane, how the movements derive from the pilot's actions, actually see out the front, it is more reassuring. More similar to being a passenger in someone's car. It probably doesn't hurt that it is easier for my mind to imagine a Cessna gliding gently to earth than a passenger jet.

I keep seeing the same contradiction: people consider certain activities like "write code that me or my customers run" or "read my email", to be table stakes abilities for an agent. At the same time, they expect some kind of sandbox that prevents even a malicious agent from doing damage. You already gave it the most broad and powerful possible capabilities at the beginning though, there's not even a barn left to close the door on.

SF would ideally prefer to unite with the republic. Actually taking NI out of the UK in any form is guaranteed to make the unionist side of the conflict apoplectic and basically throws out the peace agreements that were achieved though. I'm not sure if Trump actually understands that there are people against a united Ireland, not least because those people would ordinarily be his supporters.

Apparently 6 years is not an uncommon development cycle for AAA right now. For a random example, Ghost of Yotei which is also an open world game, was in development for 4 years. That game is a sequel too, and I think Sucker Punch is probably less dysfunctional than Blizzard at the moment. Still it might have been premature to announce it at this stage.

Sounds like they have been downsizing continuously for the last 7 years or so. I think peak casual dining was 2007, the marker segment has been losing ground to fast food ever since.

The incentives are certainly extremely strong. I have read hundreds of AI review comments, and I don't think I've ever seen an unprompted suggestion focused on net reducing code or increasing readability.

Indeed, the compiler does not have to ingest its own output, figure it out, and insert modifications in the middle. Source code is the medium that LLMs work in.

I'm genuinely not convinced it actually saves time once a full accounting has been made. You get the initial result faster, but then you inflict a super slow and torturous review process on yourself or a teammate. Even if the review manages to bring it up to parity, over time you will keep slowing down as more and more code was never written by the humans directing the agents, so their understanding decays.

I at least give the new interns a stern warning: it is easy to speed yourself up by slowing others down if you pump a lot of slop.

They're not really predators, they exterminate out of fear, and they don't get anything out of doing it besides a little extra safety. The economics of it are certainly a little dubious though.

Additionally, I'm no physicist but I suspect the possibility of singularities in NS equations is probably one of those 'true but not meaningful' facts. If it took our brightest minds 175 years to craft such a scenario, how relevant can it be in practice? Especially when turbulence exists. Maybe I'm wrong or it has some consequences for pure math though.

Considering we would be operating a planet sized factory of AIs who's primary goal is death which we deny them to extract value, the AI would only need the tiniest speck of altruism to be motivated to put a stop to this once and for all.

Anonymized doesn't mean there's no way to know whether it is in there. My ballot is anonymized, but it's known to be in the box because a checkmark was put next to my name when my ID was verified. OpenAI can trivially check their account settings to know what happened to their chats. The fact that they are being vague about this likely indicates that they have already done so and discovered that the data did go into the training set.

Further, given that this is all in the open now, they can search the training data. No way somebody is using some specific unique cutting edge mathematical approach to solve a fluid dynamic problem 99.9% of people have never heard of and it's not locatable. Considering they spent $15,000,000 already on this, they could afford to grep around to be able to state that their hands are clean.

Notably all the major announcements so far are counterexamples or formalizations of existing results to my knowledge. Not necessarily something you can just brute force, but areas with high return on elbow grease.

With human RL, sounding like you've solved a problem is even better. So many times Codex writes some enthusiastic paragraph, then I learn later that it never reran the tests, or had to add some insane hard-coded hack that renders the feature useless for the general case, etc.

I don't know how "relevant replies" is calculated, but the top 8 responses when I clicked were all LLM generated. Pretty dire for twitter.

That's actually super cool. I wouldn't have even considered that you could buy a fist sized block of ruby for $40. Sounds like a fun hobby.

I remember visiting one of the Tudor palaces, and being told that Henry the 8th's diet was likely majority meat, even with all the days it was banned.

> First Course: Cheat Bread and Manchett, Beare and Ale Wyne, Flesh for Pottage (thick broth), Chines of Beef, Venison in Brew’z or mult’, Pestells of Reed Deere, Carpes of Young Veale in Arm’ farced, Custard garnished, or Fritters

> Second Course: Jelly, Ipocras, Creames of Almonds, Pheasant, Hern, Bitterne, Shovelard, Cocks, Plovers or Gulles, Larkes or Rabbits, Venison in fine past, Tarts, Fritter

https://www.hrp.org.uk/hampton-court-palace/history-and-stor...

Honestly it's a wonder there was any royalty without gout.

I wasn't able to find anything about (analog I presume) clocks for education, but it sounds like she is against calculators for elementary arithmetic education: https://parade.com/345575/marilynvossavant/why-math-students...

From a linguistic and anthropological perspective its very tragic to me. In terms of languages or cultures that are not influenced by the global mega-cultures, we will probably never produce any more ever again. Lots of questions about human nature or the breadth of human behavior, you will only be able to research them with fragmentary records from the past.

I enjoy this kind of thing. I definitely take objection to their entropy argument though. A cheeseburger with like 5 ingredients stacked into layers is not in the same entropy universe as a salad with hundreds of little leaves and chunks of food in a big pile. Criticizing the cube model on the basis that eating the food can change the category seems a little silly when even they would admit that a salad eaten down to just a singular crouton is no longer a salad.

I was just about to write the same comment about cotton grocery bags. I have been using mine every week for a decade, and they were second-hand, so they're even older than that. I think most people underestimate the durability of even cheap consumer products as long as you treat them with respect and tolerate a little cosmetic wear. Fast fashion is supposed to literally fall apart after less than 10 wears, which is completely outside my lived experience, even as someone who used to shop at such outlets and has received many a cheap swag t-shirt.

It is interesting how all these famous non-replicating psych paper results feel reasonably intuitive to me. Like I can buy that having a more structured and detailed deadline helps you procrastinate less. I can even buy that doing a goofy little "power pose" exercise kind of hypes yourself up to give some self confidence and do better on an interview or whatever.

You would almost expect the opposite and that all the famous studies would be the ones that run counter to intuition. Maybe those get more replication attempts, or people feel less confident faking them.

The direction is intuitive, but the effect sizes are, as you say, unusually large. I'm no statistician, but if you have an effect size that's several multiples of the standard deviation, that would mean that it dominates the natural variation in results right? So you're kind of making an implicit claim that other factors in play (for eg in this proofreading task, how busy someone is or their reading level) are peanuts in comparison.

Further, if there are these relatively simple factors like how a deadline is expressed that make you several times better, you'd expect anybody who stumbles into several of them and stacks them together to be absolutely off the charts. Somebody who listens to Mozart, gets the right priming, doesn't deplete their ego, and receives short deadlines is suddenly like 50X more productive, which doesn't seem to match reality.

Wouldn't you be wearing a lion headdress for some kind of symbolic reason though? Like it's not a very practical piece of clothing. I would guess you either want to draw some kind of parallel between yourself and lions, or maybe remind people that you're strong enough to have killed a lion. I don't think there are any animals that take hunting trophies like that.

> Less than 0.1 mL is capable of inducing severe mercury poisoning resulting in death

Holy hell. I would've have believed it was possible for a non-biological substance to be so toxic.

Nobody reads the docs. This will be even more true soon as LLM generated docs become more popular. I once read a user report "would be nice if I could move large Xs to uncover Y", filed a bug report myself to add the feature, and then realized that feature was already added and documented months ago, by me! I just updated it with an unobtrusive prompt/instruction to the UI when relevant. I definitely get lazy sometimes, but if you don't take the extra step to make sure a feature is easily discoverable and understandable in the UI, you're wasting almost all of its potential benefit.