JarJarJedi
Streamlined derailments and counteridea reeducation
User ID: 1118
JarJarJedi
22hr ago
I'm pretty much the same height, but not sure what's my advantage here. I feel like Red Stapler Milton a bit - "I was told there would be an advantage?" I am not complaining, I am just curious what's the advantage. Maybe I am just blind to it, fish/water situation?
I mean I might have some advantage in physical fight but haven't been in one since my teen years. And sure, my wife can ask me to take stuff from the top shelf and I can do it without using a footstool. But I am pretty sure she'd be ok with me even if I had to use one. And I know many people who are shorter than me happily married.
JarJarJedi
23hr ago
"Intelligence tends to be positively (if imperfectly) correlated with wisdom."
I actually start thinking the opposite - there's a weak anti-correlation between high intelligence and wisdom at the top ranges. I observe too many obviously highly intellectually capable people falling victims to various mind viruses, fallacies and fads. It's like there's a car and a driver, and the car - the IQ power, what we call "intelligence" - could we awesome, but if the driver is not skilled and you put them into a race car, they'd likely hurt themselves pretty badly, and may not survive the experience even. And I am not sure what constitutes being a "good driver" yet, but I am pretty sure it has nothing to do with the car power. Of course, if the car is a child's pedal car (very low IQ) you'll never get anywhere far, regardless of driver skills. But if it's in a normal range, or especially - slightly above normal - something else comes into play.
if the smartest apes amongst us were not also generally aware of their own limitations and were trying to make good decisions that considered more than just short term interests.
Sounds great, until some apes try to build a future paradise and murder 20 millions of other apes in the process, because not being murdered is just a short-term interest that can be sacrificed for the greater good of the future.
JarJarJedi
23hr ago
People want freedom and deserve it. At least most of the people most of the time. (It also kinda implies free will is a thing that exists, I guess)
But I admit it's a hard belief to hold onto sometimes.
JarJarJedi
23hr ago
I tried to watch the first one, but everybody wearing masks annoyed me too much. Based on reviews, likely won't watch this one either - I mean, why give my mindspace to people who repeatedly were on record telling me their art is not for me and they hate me? There's enough art in the world for a hundred lifetimes, I have a choice.
If your point is just "I haven't had success getting them to do simple tasks",
Did you even read my posts? Honestly, I described how I used it for simple tasks, and explicitly said it's good for some simple tasks, even though not others, and you are like "so, you're saying you haven't had success getting them to do simple tasks?" At least LLMs, when I catch them doing something like that, say "you are absolutely right, I messed up, let me correct myself!". That's one thing they definitely have over people now.
You're describing LLMs failing at a sub-junior-level coding task (calling an API),
That's not what I described, but I am starting to see why a lot of people think of LLMs as more brilliant than they are. I mean, compared to the multitude of people that literally have written text in front of them, can spend arbitrary time at looking at it and figuring out what it means, and still are making absolutely basic mistakes in understanding - well, that's not a very high bar to be measured against, is it? I said - and repeated in another post - that the task was combining two APIs to work together, not calling a single API. The LLM btw understood this (even though it couldn't make it work correctly) but looks like you did not. Sad.
I mean I don't care if you think LLMs can do half of the work of senior developer, or 80%, or 100%, or 500% for that matter. If you actually try to pull this off, you will fail, and if you ask somebody who actually does this job day to day, they will tell you so, but if you don't listen it's no skin off my back. I'm just telling you my experience as somebody who has been in the industry for decades. If you choose to think I am an idiot and LLMs are brilliant and everybody who says it is not so are just holding it wrong - be welcome to it.
JarJarJedi
1d ago
I think it's more precise to say AI is good to generate coverage. E.g. if I have a big code surface, and I know how to test it, but I need like 50 different tests, which look pretty similar, AI can help. If I need a script to generate a bunch of randomized data with certain properties, to cover a bunch of corner cases, AI can help. If I need to figure out how to test some complex feature that nobody tests before - AI would be useless.
An LLM could absolutely do 30-40% of this, assuming it's being directed by a senior programmer and has said programmer on hand to do the other 60-70% of it
Nope, unless by 30-40% you mean "a lot of large data files for integration tests", which yes, LLM could do - moreover, it could write a serviceable code to generate those. But something like production-quality code for database backend? No freaking way, 10% would be too generous to be honest. I kinda know what I am talking about, I do this for living.
a "senior" programmer would still be doing a decent amount of coding
Like, all of the substantial coding, except some very corner-casey boilteplatey pieces that don't really do anything complex but are needed anyway.
LLMs are also fairly decent at answering design questions like this
No they aren't. Not when these questions actually matter - i.e. when you need to think about it with human brain. If it's a question like "what is in fashion this season" or "do I need to rewrite my little app in Rust" or any other question that does not have any technical depth - sure, LLM will duly regurgitate all the accumulated wisdom of thousands of stackoverflow posts to you. But that's not thinking. Last time I seriously needed a deep choice like this, it took 2 months work of 3 highly qualified engineers to just collect preliminary data to start approaching the decision. And then it took another month to actually make a decision. I agree that most choices in everyday life are not deep. It usually doesn't matter whether you use mysql or postgres or oracle or some cloud doodad - 90% chance your app will never hit the limits of either, and if it does, you'll rewrite it 5 times on the way anyway so initial choice does not matter. When it does matter, regurgitating stackoverflow at you is not going to cut it. Don't get me wrong, SO is a great resource and helps a lot, and properly regurgitating it on request is a very useful function, which I personally enjoy using almost daily. But that's not "design questions", at least not ones I am talking about.
did you give the LLM enough context for what a correct API call should look like?
Yes I did. I was not born yesterday. As I mentioned, I do this for a living. For over 30 years now. Shit, I just realized it's closer to 40 than 30 by now. If there's an article about "how to prompt your LLM" I probably either read it, or read some version of it.
which still happens sometimes, but there are solutions to that. Usually just telling the LLM to break the problem into smaller chunks,
There's no smaller chunks, what's less that two APIs? One API? I already had one API, I needed two of them working together. I'm not sure what could be "smaller chunks" in this case - that's like saying to jump over the ditch in two leaps. Either you jump to the other side, or you fall into the ditch, there's no "smaller" here. And yes, that happens all the time. Because LLMs do not actually understand anything but just try to guess what you could want to hear, if I have no idea what I want to hear (that's why I asked it in the first place) and there's no place where the answer already exists then it very often fails. The sum of human knowledge is very vast indeed, and just rearranging it you can indeed get very far. But there's a limit to it where you just can't get this way. The specifics of this case don't even matter - I found the solution already and I had many such cases before, LLMs just don't do well with certain class of tasks. That's fine, all instruments have their advantages and their disadvantages, but if somebody tells you they invented the instrument that would do everything for everybody - it's very likely you are talking to a swindler.
JarJarJedi
1d ago
we could see a discontinuity with LLM programming ability jumping up to where it can suddenly do 80% of a senior programmer job instead of the current 30-40%
I don't believe either of it. I don't believe it's at 30-40% of a senior programmer now (at least what I understand as "senior") - more like 10% at best - and I don't believe it will be up to 80% anytime soon. For me senior programmer is somebody who can write a serviceable database server. I don't mean in a weekend or so (some people could but I don't ask for that much), but he'd be technically competent to produce most part of it, given adequate time. Higher level either go deeper (like designing an optimizing compiler which would require some deeper theoretical background) or go higher to more social aspects like matching the design to actual domain requirement (questions like "what kind of database we actually want to use here and what is more practical given limited resources we have?"). I don't see LLMs coming even close to being capable of solving problems like this. And given current model, which is "make a tiny-shredded salad of all existing knowledge and regurgitate is in various combinations" I don't think it's possible in principle.
As for practical applicability, I have a mixed experience with LLMs. From one side, it's pretty good with simple coding tasks - e.g. when I tried to vibecode an app for personal use to track my gym schedule, it did it admirably. Admittedly, it made some choices I personally would never make, but if I pretend to not know anything about coding it actually built a workable app from zero without me (deliberately) using any of my coding knowledge, just by saying how it should look like. This is certainly gets to the level of junior programmer and maybe about 3/4 of the mid-level one. If you don't care how it works, how well it works and what happens inside, it certainly can produce code that does things. You need to watch out though because while I was working on it, there were a couple of times when it misunderstood me and attempted to wipe out half of the existing app, and I had to tell it "don't do that". Since I insisted on approving every step it was doing (deliberately not trying to go far beyond "it is wiping the whole thing now? oh it doesn't then fine, go ahead" but still exercising some judgement) it was ok, but if I put it into fully autonomous mode I'd probably end up with nothing useable.
On the other hand, many times when I tried to use it on a more deep but more specific tasks, it failed horribly. It invented things that do not exist, it gave me contradictory advice, it once tried to convince me to wipe out my whole computer because it didn't know how to solve a certain permission issue (which it caused by hallucinating solution that doesn't work), and it was horrible. Last frustrating exercise was when I tried to have it explain to me how to use two certain APIs together, and it wrote a plausible code and configs, except it didn't work. Then it told me I need to do some changes, producing a very plausible explanation why the code (that it produced) can't work. It didn't work with the changes either. It had a very plausible explanation why it doesn't work and instructed me to go to the original version. When confronted with the fact none of its contradictory suggestions actually works, it said "something else must be wrong" and essentially gave up. Later I found by myself it used the API in a completely wrong way and confused two different versions of the APIs even though I gave it exact versions I am using. That's what you get for using a chinese-room machine that does not contain any facilities to actually understand what it is talking about.
Summarily, I think for doing "solved tasks" - which is a lot of programming done nowdays, like creating yet another webpage that displays a result of a query from the database, etc. - it is already a decent tool and will likely become better. Probably sometime soon creating a simple app along known interaction scenarios would be something LLM would allow any computer user to do, not requiring any coding experience at all. However, doing more advanced tasks that require deep understanding, foresight and judgement is going to be completely beyond it. What is worrying me here is that human programmers learn on simple tasks to gain advanced understanding. If all the low-effort segment will be occupied by LLMs, how do we get new advanced programmers? Right now the pipeline is still functional, but if the things will go as they go, in 5-10 years it will dry up. It will likely be good for me personally, because if I would want to keep working into my advanced age, I will have less competition from the younger generations, but then in another 10 years current advanced programmers will start retiring, and the ones who grew up on "who cares, LLM will do it for me" will be completely incapable of replacing them. I hope by then the Congress will ban usage of LLM-generated code in medical, transportation and other high-danger devices, otherwise I'd be in trouble.
JarJarJedi
3d ago
Anybody remembers "elevatorgate" yet?
JarJarJedi
3d ago
I am an adult and I don't use discord a lot but I have at least two groups I participate with which are on discord. I don't know why discord specifically, it could be anywhere as far as I am concerned, but the fact is they are on discord, and so there are a bunch of adults there. That's just how it happens - a group decides "why don't we set up online presence?" and one of the people says "I know discord, I can set us up there" and voila - the whole group is now on discord. Could end up on Telegram or Slack or whatever - it's kinda random.
JarJarJedi
4d ago
Tell me you don't actually know any introverts without telling me you don't actually know any introverts.
Everybody has their blessings and their curses. I am sure your blessings are plentiful, but there are people for whom the chance of success cold-asking a woman out in person are exactly zero. It's just a fact. There are things that some people can do and some can't. Some can win a match with a professional boxer and some can't. Some can run a marathon and some can't. Some can bench press 2x bodyweight and some can't. Some can successfully charm women in person from cold start and some can't. Giving them advice "don't do shit you have - maybe small, but non-zero - chance with, and instead do shit which you have zero chance with" is strictly harmful.
JarJarJedi
4d ago
They just wouldn't release this proof, if it ever existed (which it likely does not), but instead destroy it (which they likely did long ago, if it ever existed, which it probably didn't from the start). We know for a fact about many recent cases where evidence of crimes was destroyed (e.g. Clinton records, or IRS records of prosecution of Tea Party NGOs, many such cases) and absolutely nothing happened (except for some noise in the press, but there's always noise in the press). In fact, just recently we learned the videos from Jan 6 about pipe bombs just "disappeared" and... nothing. It is trivial and safe to disappear any evidence if you are in control of the government. Moreover, there had been ample evidence of both Clinton committing perjury, and Clintons having deep - much deeper than Trump ever had - links with Russia (among many other foreign interests, if anything, they are very equal opportunity corruptionist) and exactly nothing happened.
And this is not unique and not specific for Clintons alone. In fact, we just this week learned Fulton county illegally certified 315K votes - something that people were prosecuted for trying to look into - and mark my words, exactly nothing will happen to people who had done it. In Minnesota, billions were stolen under Waltz watch (and likely with his active enablement) - do you think anybody from his team will suffer any serious consequences (like jail and expulsion from politics)? Neither do I. I could add examples here ad nauseam, but I think my point is clear - it's only in movies once you publish something that looks incriminating a top political figure, they are instantly overthrown and the closing credits roll in. In reality, in most cases very little happens, especially when we talk about somebody of Clinton's caliber. There are too many people invested in it.
So summarily we would have some murky indications (because anything else had already been long destroyed) against the massive coverup machine which had already achieved numerous coverups. The premise that they did not release supposed dirt on Trump that they supposedly had, because they feared - while holding absolute control over the government - that this machine will fail to protect somebody as unassailable as Clinton - is completely laughable.
It's a big club and you ain't in it.
And you think this big club can't manage to release dirt on Trump without bringing Clinton down? After all they have done so far?
JarJarJedi
4d ago
Oh, they will try to make hay out of anything. If Trump would fund research that cures cancer, they would declare that he finances dangerous experimentation with poisonous chemicals that kill living tissue and prevent growth, and plans to inject millions of people with them to enrich his friends. There's always a way to frame something in a bad light. The antidote for it is pulling back the curtain and expose the game.
JarJarJedi
4d ago
would have brought down a bunch of bigshot Democratic politicians and donors too like Bill Clinton
Bill Clinton hasn't been a "politician" by that and pretty much nothing short notarially authenticated tape of him committing a felony could "bring him down" in any way. That's the person who fucked interns in White House and lied under oath and got away with it without any "bringing down" happening. Imagining that some words from Trump - which would automatically be dismissed by 100% of the Left as lies no matter what he says - could "bring down" anybody on the left, let alone a figure like Bill Clinton, is complete nonsense.
JarJarJedi
4d ago
That's some messed up thinking right here. Of course he knew Epstein, nobody ever tried to prove he didn't - he himself admitted many times he knew Epstein, and they had interacted socially many times up to 2000s, and then he personally banned Epstein from his clubs after learning about him recruiting there, which is pretty hard to do if you don't know him. Pretending as anybody is arguing that "he did not know Epstein" is just insane. There's a big distance between not knowing somebody at all and being best friends for life. And there's a very long timeline here, spanning decades, for the duration of which Epstein met very many people - pretty much everybody there is to meet. And of course there would be photos, that's the whole point of it. The whole business of "being connected" is having photos of you with celebrities, and being in the same parties as important people are. Implying that this means everybody who ever been in the same room as Epstein is now complicit in his crimes is bullshit.
JarJarJedi
4d ago
She looked me dead in the eye and said "don't talk to me".
That is what I call immature and asshole behavior.
JarJarJedi
4d ago
People meet in all kinds of weird places. But discord probably not ideal - you'd at least want to know how your potential mate looks like, not?
JarJarJedi
4d ago
That's always one of the possible outcomes. You take the risk, and often it doesn't pay off. And the girl doesn't owe anyone her acceptance of their advances, if she feels uncomfortable being propositioned, she can refuse, and she can leave. There's nothing immature about it. One has to decide if it's better for him to always leave in a state of vague hope but never know, or to risk it and know either way.
JarJarJedi
4d ago
I've only been to France a couple of times for work, and of course IT professionals all spoke very decent English, but from the people in service (waiters, ticket clerks, etc.) it was very much hit or miss. Some spoke absolutely zero English and basically refused any attempt to bridging the language gap. Once I had to use ticket machine instead of the ticket booth because the clerk refused to understand anything (which is kinda weird in a railway station - they must have significant tourist traffic there) and once I had to go to another restaurant because they didn't have any non-French menu and refused any attempt to discuss the order in English (it was before Google Translate time of course). It wasn't too bad in general, I found my way around to food and transport fine eventually, but I got the impression that some people in France definitely don't want anything to do with English. In Germany I had an opposite experience - we once got lost on a train station and asked some German ladies (in English) about it and they gave us super-detailed explanations and refused to part with us until making sure we made to our train, and their English wasn't that great (better than my German for sure!) but they were extremely enthusiastic about using it. We felt a bit uncomfortable that we took much more of their time than we planned to.
JarJarJedi
4d ago
That's my problem too - I can squeeze out some basic Spanish phrases (donde esta la biblioteca?) but my capability of understanding the answers goes down quickly. Especially if the environment is noisy, as my hearing is also not what it used to be (it's mostly ok but I could miss some details and when trying to parse a language I am not well versed with, it starts showing).
JarJarJedi
7d ago
(Uncle Sam knows your date of birth and contribution record)
Yes, and you would think they would have some provisions so that the database wouldn't have records showing 100+ year old people being paid, and other invalid cases being identified, but that's not the case. They can make it clean, they just didn't bother to.
There is significant retail rorting of SSDI, but mostly by Red Tribers so not something a Trump-led DOGE would want to go after.
What this is based on?
JarJarJedi
7d ago
I am going to be in Madrid for a week or so next summer. Mostly work, but probably also some free time. I have been studying Spanish for a while now, and I got to the level when I am mostly understanding written text, at least up to the newspaper level (though some rarer words are missing, but mostly I get through), but when I listen to the radio it's about 50/50 chance I can understand. I know most people these - especially in tourist places and hotels and so on - would speak English likely, but I think it'd be cool if I could speak Spanish to them - and especially understand them speaking Spanish to me. Any suggestions about how could I increase the likelihood of this happening?
JarJarJedi
7d ago
Medicare, medicaid and SS also give some opportunity for grift, and so does defense budget (I mean, if climate change is a threat to national security, we can finance climate change projects through defense budget, right? And if racism is a public health issue, we can finance DEI programs via healthcare budgets as well) we still have almost a trillion dollars in discretionary non-defense spending. It is true that solving the budget balance without addressing mandatory spending is not possible. But I am not talking about solving the budget yet, I am just talking about cutting off the most aggressive leeches, and thus forcing them to at least play on equal footing. My point is not about solving the budget - that can come later - but about denying the enemy the resources which should be either deployed to more worthy causes or returned to (or not taken from) the taxpayers. If the Left wants to donate to their favorite causes, they are welcome to, but without the help of the IRS.
And even Krugman (who one can usually rely on distorting the reality as much as possible to benefit The Party) admits this:
Oh, and the federal government has been providing aid to state and local governments, largely to limit layoffs of schoolteachers.
The schoolteachers part is most likely a lie (I didn't check but I know who Krugman is) but the preceding part is true - significant chunk of federal money goes as "aid" to local budgets, where it is rerouted - either directly, or through a basic fungibility trick - to various pet causes. Establishing transparency and control over this would do the conservative cause a lot of good - but they are doing virtually nothing about it.
JarJarJedi
8d ago
in the sense that the money goes to the thing it's labelled for,
Only in a very broad sense, e.g. if the money is labeled "covid subsidies", it is going to somebody who claimed they need a subsidy because of COVID. But whether they actually need that subsidy, whether they should be in the front of the line for that subsidy, and whether their claim has any relation to reality, and whether they are actually going to spent the money to the cause they promised to spend - all this is controlled very weakly. And the leech networks have long adapted to the weak controls and learned to extract money by saying the correct "open Sesame" phrases, after which they get access to streams of money.
That said, I absolutely agree that state and local money need the same treatment.
Ah great, so I am not the only one annoyed by this. I had been looking for a decent small phone for years, and looks like they just died out. Really, if I want to enjoy iMAX-sized cinematic productions, that extra inch is not going to help me. But they make it heavier and less convenient to carry in my pocket. I'd gladly support manufacturing smaller ones with my money, but there seems to be no market at all there. Maybe those newer flippers though, have to check them out. I used to have a flipper phone for years, before modern mega-bricks became the standard.
More options
Context Copy link