@HardwareGeek That isn't happening to me, yet, at least. Or I'd have a stiff increase in my insurance.
And once my current phone breaks, it's getting replaced with a linux phone.
Posts made by Carnage
-
RE: Driving Anti-Patterns - Necro Edition
-
RE: Nope, you eat it
@Benjamin-Hall said in Nope, you eat it:
@Carnage said in Nope, you eat it:
But in all honesty, the taste is far less bad than the smell
Because by the time you get it in your mouth, your sense of smell (which handles most taste) is already offline from the reek. So the delta between (smell, taste +smell) is basically the texture, which I've been told isn't exactly great either.
Yeah, it's pickled fish, so the texture is kinda snot-like.
-
RE: Nope, you eat it
@HardwareGeek said in Nope, you eat it:
@Carnage said in Nope, you eat it:
the taste is far less bad than the smell.
That is said of Limburger cheese, too, but I have no intention of getting close enough to either to find out for myself. "Less bad" is a relative description, and "not bad enough to make you puke" is not an adequate level of goodness.
I dislike all forms of pickled fish, they all kinda taste like snot with some flavoring. This isn't much different. It just kicks you in the nose first.
-
RE: Nope, you eat it
@HardwareGeek Yeah, you don't open those indoors. Or if you do, you do so underwater. The cans tend to spew stinky fluid when opened and whatever that gets on smells for a long time. You can always rub it down with tomato sauce to make it smell less though.
And, you should keep the cans around for a year extra so they swell nicely. If the can isn't bulging, they aren't "ripe" yet.
But in all honesty, the taste is far less bad than the smell. -
RE: Driving Anti-Patterns - Necro Edition
@blek said in Driving Anti-Patterns - Necro Edition:
@Carnage My main requirement for a car is enough trunk space to carry two 50l beer kegs plus the tap, so the BRZ/86 is right out. If it wasn't for that I'd be driving something like a Peugeot 106 S16.
That's what the back "seat" is for in the BRZ. I've actually packed a surprising amount of shit into mine. But when I want to haul shit I have a Transit, the extra long and tall version. But my post wasn't really serious, while I love driving my BRZ it's probably a bad idea to get one if what you want is comfortable transport.
My wife that always complained about how I was driving it took it for a drive since her car was down at the time, and when she came home she told me my car is evil because she started driving like me, the car made her do it!Although, she is Mexican, and I have driven in Mexico. They all drive like madmen on crack. I suppose all she needed to connect with the spirit of her people was a car that made it happen.
-
RE: Driving Anti-Patterns - Necro Edition
@blek said in Driving Anti-Patterns - Necro Edition:
I sorta kinda need a new car and I'm thinking of a Subaru XV (Crosstrek in the US). Someone please talk me out of it, I know it's a horrible idea.
It is a horrible idea, you should get yourself a BRZ instead!
-
RE: I, ChatGPT
@DogsB said in I, ChatGPT:
@sockpuppet7 said in I, ChatGPT:
the copyright was always an excuse, these are common luddites
I want to see an unfiltered AI with everything in it to see what it creates, but the reality is that if you're not going to enforce copyright and have a way for creatives to make rent with their work, you're not going to have anything worth consuming.
It's kinda fun playing with AIs that have the safety rails removed. The chatbots are also all kinds of amusing and will gladly tell you how to build bombs, albeit I wouldn't trust them to actually get it right.
-
RE: Linux on the Desktop? A long way off...
@dcon said in Linux on the Desktop? A long way off... :
@Carnage said in Linux on the Desktop? A long way off... :
@Parody said in Linux on the Desktop? A long way off... :
@DogsB said in Linux on the Desktop? A long way off... :
Not a huge sample but still interesting. I did like that he was shocked by most people having a good experience when interacting with the Linux community.
Today we have way higher expectations than a few years ago.
2015-2020: "Look! My favorite game now works on Linux. That's amazing!"
2020-Now: "Ohh, this game with anti-cheat and two very specific apps don't work on Linux at all. Linux is not so good."Expectations were pretty low for a long time. Back when we had the LAN pit one of my buddies would bring over a bootable Linux CD every once in a while to show me how the hardware support had advanced. "Look, it's a year later and it can play sounds now! On this particular Sound Blaster, anyway."
On my first linux install (some ancient RedHat version in 95 or 96?) I got the sound working by editing the sound driver and fixing a bug. But ever since then, I've had working sound on Linux on all hardware I've tried. Some of them took a bit of fucking about, but not to the point of fixing drivers at least.
Well, my work laptops still can't handle the lid being closed while running without requiring a hard boot (ubuntu 18 and 22).
The thing that's annoying with my work laptop is that it draws power when sleeping so it'll drain the battery in a day or two. Other than that everything works. I used to have a work laptop with a touch screen, and that was... Not entirely implemented. It worked as a huge touchpad, and if you had more screens plugged in, it went all sorts of wonky, because it tried mapping the display surface to the touch surface sooo... Not very good. I also turned off the touchscreen after a short while because even when it was working correctly with the single screen it was annoying as all hells.
-
RE: Linux on the Desktop? A long way off...
@Parody said in Linux on the Desktop? A long way off... :
@DogsB said in Linux on the Desktop? A long way off... :
Not a huge sample but still interesting. I did like that he was shocked by most people having a good experience when interacting with the Linux community.
Today we have way higher expectations than a few years ago.
2015-2020: "Look! My favorite game now works on Linux. That's amazing!"
2020-Now: "Ohh, this game with anti-cheat and two very specific apps don't work on Linux at all. Linux is not so good."Expectations were pretty low for a long time. Back when we had the LAN pit one of my buddies would bring over a bootable Linux CD every once in a while to show me how the hardware support had advanced. "Look, it's a year later and it can play sounds now! On this particular Sound Blaster, anyway."
On my first linux install (some ancient RedHat version in 95 or 96?) I got the sound working by editing the sound driver and fixing a bug. But ever since then, I've had working sound on Linux on all hardware I've tried. Some of them took a bit of fucking about, but not to the point of fixing drivers at least.
-
RE: United Airlines: the airline we love to hate, but we can't agree on why
@Polygeekery
Don't forget the fatties that don't want anyone to know that they are fat. (Even though we can fucking see they are.)
-
RE: I, ChatGPT
@Arantor said in I, ChatGPT:
@Carnage I was just impressed it produced actual working links. Even that’s not a given.
Yeah... I suppose they may have added post processing to remove any dead links.
I didn't read much of the WoT, but it didn't seem to have much in common with the study it references though, since that seems mostly about using AI to find the people that abuse alcohol prior to surgery to reduce post surgery complications. -
RE: I, ChatGPT
@Arantor said in I, ChatGPT:
@remi possibly, maybe even probably. I didn’t read them, but it wouldn’t be surprising.
4 of them seems to be the same thing, and one of the links looks like a reference to the actual report. 2 of them are the same article on two different servers with different CSS, and both of those are pages of the university that published the study. sooo.... Good references?
Mind, I didn't dig any further than reading a few sentences... -
RE: I, ChatGPT
@topspin said in I, ChatGPT:
@error said in I, ChatGPT:
Week 3 of Salesforce training. We're talking about chatbots. It's being aggressively pitched.
I just heard GenAI referred to as "true" AI, which to me further muddies the concept of AI vs AGI.
Whenever some new weak-AI thing comes out, it’s being marketed as the new big thing and so you need new words to describe the original concept of AI. AGI is just that. Once ChatGPT is marketed as AGI, you’ll need yet another term to talk about actual AI.
Even the people that coined the term "AI" say that it was a mistake, and they should have named it differently. There are plenty of things that are AI, but not what people expect to be AI, such as pathfinders. AGI is not gonna happen anytime soon, unless there are a bunch of crazies in a garage that solved the problem and are just working on the implementation.
LLMs and ChatGPT and it's ilk are still just "AI" just like A* and will not be anything else, no matter what the salespeople tell you. -
RE: United Airlines: the airline we love to hate, but we can't agree on why
@BernieTheBernie said in United Airlines: the airline we love to hate, but we can't agree on why:
@Carnage said in United Airlines: the airline we love to hate, but we can't agree on why:
And if you were fat, it might kick you back out because it'd think you were two people.
That's racist discrimination of 'muricans!
Yeah, isn't it awesome?
They should have them in airports too! -
RE: Before AI....Before Crypto...There Was...BIG DATA
@sockpuppet7 said in Before AI....Before Crypto...There Was...BIG DATA:
@Carnage said in Before AI....Before Crypto...There Was...BIG DATA:
@sockpuppet7 said in Before AI....Before Crypto...There Was...BIG DATA:
@dkf said in Before AI....Before Crypto...There Was...BIG DATA:
@Arantor said in Before AI....Before Crypto...There Was...BIG DATA:
@topspin we solved the webscale problem with microservices.
solved
But seriously, I've suspected for a long time that much webscale processing/problems was really only caused by putting too much data in one place to start with. Microservices help because they don't require making a huge data dumpster as a first step of achieving anything at all, instead focusing on services that have only part of the picture.
I think microservices are probably the right size if you have as many teams than services. If you have a small team maintaining 10, 100, 1000 services you're doing nanoservices, or picoservices or whatever is even smaller than that
FaaS.
fuck as a service?
Nah, that's a really old profession.
FaaS in IT is usually Functions as a Service. So a swarm of functions doing stuff™. -
RE: Before AI....Before Crypto...There Was...BIG DATA
@sockpuppet7 said in Before AI....Before Crypto...There Was...BIG DATA:
@dkf said in Before AI....Before Crypto...There Was...BIG DATA:
@Arantor said in Before AI....Before Crypto...There Was...BIG DATA:
@topspin we solved the webscale problem with microservices.
solved
But seriously, I've suspected for a long time that much webscale processing/problems was really only caused by putting too much data in one place to start with. Microservices help because they don't require making a huge data dumpster as a first step of achieving anything at all, instead focusing on services that have only part of the picture.
I think microservices are probably the right size if you have as many teams than services. If you have a small team maintaining 10, 100, 1000 services you're doing nanoservices, or picoservices or whatever is even smaller than that
FaaS.
-
RE: WTF is happening with Windows 11? And nothing else
@Arantor said in WTF is happening with Windows 11? And nothing else:
@Carnage said in WTF is happening with Windows 11? And nothing else:
@Arantor said in WTF is happening with Windows 11? And nothing else:
@Carnage rumour has been doing the rounds that MS wants to buy Valve.
Of course they want to. They've been on a bit of a purchase and extinguish spree lately.
More worryingly are the people convinced it'll make Xbox better.
People are idiots. The XBox would possibly improve if MS sold it to Valve, because the problem is MS, not the tech.
-
RE: WTF is happening with Windows 11? And nothing else
@Arantor said in WTF is happening with Windows 11? And nothing else:
@Carnage rumour has been doing the rounds that MS wants to buy Valve.
Of course they want to. They've been on a bit of a purchase and extinguish spree lately.
-
RE: WTF is happening with Windows 11? And nothing else
@Arantor They really want people to abandon Windows completely. If Valve gets their shit together and release general use Steam OS sometime soon, there might be a mass exodus of gamers.
-
RE: United Airlines: the airline we love to hate, but we can't agree on why
@Polygeekery said in United Airlines: the airline we love to hate, but we can't agree on why:
@remi it is all vaguely similar to the old saying about how "You would be surprised the places you are allowed to go, if you just act like you are supposed to be there". Assumed authority and people's trust and lack of cynicism coupled with a deep aversion to conflict or making waves.
Yep, the best way to get in somewhere is to dress for it. Either a cleaner or facility service worker works a peach, but requires to get the right names. Otherwise just copy the look of the workers where you're going. Just act and look like you belong and you can tailgate past most door checkpoints.
Though none of that worked at one place I worked that had security classification. Rotating single doors with scanners that made sure that there was just a single person in each door section. And if you were fat, it might kick you back out because it'd think you were two people. There were also some on point security guards on both sides of the first set of rotating doors.
The only way into the secure area was an air bridge between buildings, and the secure building was more or less a concrete bunker with no windows on the first two stories from ground lever up. Under ground there was also no windows, for obvious reasons.
Best gig I ever had, since not even the bean counters were allowed to know what I worked on. Need stuff? They couldn't know why or what, just sign the dotted line for the money. -
RE: Before AI....Before Crypto...There Was...BIG DATA
@boomzilla said in Before AI....Before Crypto...There Was...BIG DATA:
This article is over a year old now, but it's interesting to see how an earlier bubble / craze has died out and never lived up to its promises. It's written by one of the guys who arguably started it all, since he was a "founding engineer on Google BigQuery" and the guy who got stuck with going around promoting it.
I questioned one of the evangelists about it when I had to work towards a data lake, and they couldn't understand why I thought that having at least some idea of structure world be good, but no. It was all sock gnomes.
Speaking of, that's the best measure of faddery in it. If there is a magic step in the plan that will just have usefulness emerge, then it's all high caliber bullshit.
This has worked every time a new dumb dad showed up during my time as a code jockey. -
RE: I, ChatGPT
@Zecc said in I, ChatGPT:
The trolling shitposts of yore comes back to haunt Google. It's beautiful!
-
RE: I, ChatGPT
@Zerosquare said in I, ChatGPT:
@Gern_Blaanston said in I, ChatGPT:
So, AI is going to replace consultants.
Eh, they are free to try. I'll just double my rate if there has been AI there first.
-
RE: The Official Funny Stuff Thread™
@HardwareGeek said in The Official Funny Stuff Thread™:
@Zerosquare said in The Official Funny Stuff Thread™:
That "made of EU and non-EU ingredients" mention is quite common here. I assume the phrasing is legally mandated, but it still sounds idiotic.
Even more so when the phrase is "Milk originc".
Why would you put gin in milk?
-
RE: Nope, you eat it
@boomzilla said in Nope, you eat it:
Aaaw, now I am mildly upset at myself that I didn't think of making baku for my friends wedding, who I'm the best man for. I used to make that for parties in my teens. Goes great with beer!
-
RE: Spyware as a service
@boomzilla Supposedly, files were reappearing on devices that had been factory reset and were running with new users as well.
I would not trust anything apple again after this. They either store shit on their servers that shouldn't be, or they are leaking shit from user accounts that shouldn't leak, or the secure delete isn't secure. None of the options is at all in any way good.
Not that I'd trust MS anywhere near as much as I'd trust Apple anyway. Not even the Kola borehole has a shot at going below the low bar MS sets in trustworthyness. -
RE: Scientific Science
@topspin said in Scientific Science:
‘bug bounty’ programmes
In our industry, too, the costs of undetected errors are staggering.That sounds like a good idea on the face of it. I'll call it Paid Post-Publication Peer-Review, P4R for short.
Will it work to improve peer review? Probably not. Is it worth trying? Anything, at this point.Estimating the Reliability and Robustness of Research (ERROR), pays specialists to check highly cited published papers, starting with the social and behavioural sciences
I'm going to order a new Ferrari right now ...Scientists that are discovered to publish shitty or outright doctored and false papers should be sent on a world tour and put in stocks at the largest universities where people can throw rotten eggs at them, ending with a month in stocks at the university where they published the most falsehoods. And all the universities should have a small town square with a collection of shame stocks for this purpose.
And also fine the publications that publish the most shit. -
RE: Spyware as a service
@dcon said in Spyware as a service:
@boomzilla said in Spyware as a service:
AI has usurped the brain power previously reserved for crypto currency enthusiasm. There was lots of that going on, too.
That sentence just made me think. When do we get the AI-crypto rush?
AI powered crypto trading bots have been a thing for a few years now.
I suppose you could store your AI datablob on the blockchain for some extra luls. -
RE: I, ChatGPT
@boomzilla said in I, ChatGPT:
@Carnage I always ask for a link to the information from my AI so I can check it against, like, actual documentation. Sometimes you get lucky and it's correct.
It's been a while, but I asked for a response with reference links in it, and about 50% of the links were generated and never existed, and of the other half, a good quarter didn't even have anything to do with what the LLM said. The last quarter seemed at least relevant, even if they could say the opposite of what the LLM said.
Kinda like what you'd expect to get from statistically generated text with a sprinkling of random on top.
-
RE: I, ChatGPT
@boomzilla said in I, ChatGPT:
@DogsB it just highlights that it's not reasoning about the letters and reminds you to be more careful when you get less obvious hallucinations.
On that there biker forum I'm hanging out on, people are posting ChatGPT as a reliable source and as a way to prove someone wrong. I keep pointing out to them that it's wholly untrustworthy. Maybe I should start posting these things so people can understand that it can't be trusted, it's not data, it's a statistical approximation of data.
-
RE: I, ChatGPT
@Zecc said in I, ChatGPT:
The tweet from yesterday is now an article on NPR:
"We believe that AI voices should not deliberately mimic a celebrity's distinctive voice — Sky's voice is not an imitation of Scarlett Johansson but belongs to a different professional actress using her own natural speaking voice," the company wrote.
And yet you tried to get ScarJo on board.
They also backed down and removed the voice option when asked for proof for how they got it. I'm guessing they are getting hold of a voice actress that sounds like Red John's son right now to back this claim if it goes to court.
-
RE: The Official Funny Stuff Thread™
@dcon Imma do that next time. Though, funnily enough my wife very rarely gets angry at me.
-
RE: I, ChatGPT posted in Side Bar WTF
-
RE: Driving Anti-Patterns - Necro Edition
@HardwareGeek said in Driving Anti-Patterns - Necro Edition:
@boomzilla What was the BAC of the people involved in that activity?
Yes.
-
RE: Things that remind you of WDTWTF members
@Benjamin-Hall said in Things that remind you of WDTWTF members:
@Zerosquare said in Things that remind you of WDTWTF members:
@Benjamin-Hall said in Things that remind you of WDTWTF members:
Sometimes, thinking is overrated and I just want to turn my brain off.
Have you tried being a NodeJS developer?
Being an NodeJS developer requires lots of thought, at least if you want to do it well... Because you have to maneuver around an ecosystem that is all rusty nails and shards of broken glass.
Be a HPC, and just throw something together that barely works and move on to the next gig. Making an actually working product is the next idiots problem.
-
RE: I, ChatGPT
@boomzilla said in I, ChatGPT:
I'd also put the paying/product meme here, but slack charges for the privilege of being data mined.
-
RE: I, ChatGPT
@sockpuppet7 said in I, ChatGPT:
@Carnage said in I, ChatGPT:
@sockpuppet7 said in I, ChatGPT:
@Carnage said in I, ChatGPT:
The trick is the G in AGI. No flavor of NNs (or by extension LLMs) can do that. They have no intelligence, they have no reasoning. It's all mirrors and smoke.
intelligence isn't defined well enough for this argument to get anywhere
For no definition of intelligence, LLMs have any.
But that is irrelevant, LLMs still cant reason, they have no knowledge or awareness.I don't think anything short of human intelligence would convince you
You're one so feel free to convince me that LLMs are more than statistical analyses of word frequencies, or how they could become more than that.
-
RE: I, ChatGPT
@sockpuppet7 said in I, ChatGPT:
@Carnage said in I, ChatGPT:
The trick is the G in AGI. No flavor of NNs (or by extension LLMs) can do that. They have no intelligence, they have no reasoning. It's all mirrors and smoke.
intelligence isn't defined well enough for this argument to get anywhere
For no definition of intelligence, LLMs have any.
But that is irrelevant, LLMs still cant reason, they have no knowledge or awareness. -
RE: I, ChatGPT
@sockpuppet7 said in I, ChatGPT:
Reinforced learning with some random exploration can, and got better than humans on videogames
I built a NN/DT/perceptron game bot in 2005 that could beat games better than humans. That's not a particularly hard task.
The trick is the G in AGI. No flavor of NNs (or by extension LLMs) can do that. They have no intelligence, they have no reasoning. It's all mirrors and smoke.
And LLMs get a fun regression to the mean thing going when you start feeding LLM-generated content back in. That is something that doesn't happen with brains. History is a solid statement that we keep inventing and building more and more complex things. -
RE: I, ChatGPT
@sockpuppet7 said in I, ChatGPT:
@cvi said in I, ChatGPT:
@sockpuppet7 said in I, ChatGPT:
the standard you should compare it's output is the inner voice in your mind, not your final work on paper
one big difference is that LLMs don't learn, not even in the sense of neural networks. the training that updates the network (weights, but not structure) is done once. when interacting with an LLM, it is simply fed a context that contains all previous state. if it weren't for a bit of randomness in the prediction of whatever it should output next, it would be entirely deterministic (and it's in fact possible to run networks in fully deterministic modes).
the brain learns all the time. the weights and structure change. this is how you remember shit.
that's not a limit of LLMs, RLHF can be performed continuously
The training isn't the problem with LLMs, it's that they generate statistically seemingly relevant output from the input. There is no intent, intelligence or agency at all, just math operating on words according to weights in a transform cluster of nodes according to a large dataset. Howe often you update the net of nodes doesn't really matter for what types of problems they can be used to solve. What they produce isn't data, it's a statistical approximation of data.
-
RE: I, ChatGPT
@sockpuppet7 said in I, ChatGPT:
@Carnage said in I, ChatGPT:
@sockpuppet7 said in I, ChatGPT:
@dkf said in I, ChatGPT:
@sockpuppet7 said in I, ChatGPT:
they seem to train with everything they got, and later use fine tuning to shape it, and that appears to work better than training with less, selected data, if I understand any of what I read about it
The problem is that means the only answers it can ever generate are projections from its training set. If that's what you're looking for, that's great. If you need anything that isn't just a projection of what went before, LLMs are exactly the wrong tool.
you should know it's BS, neural networks generalize and create new things on the patterns it learned
now you have billions of parameters, with a ridiculous amount of data to form connections, and it definitely write things unlike anything before
put some reinforcement learning over it to make it smarter with it's experience and we'll soon have our terminators, if global warming doesn't finish us before it
This is the socks gnome business plan, except it's about collecting data and still a ? on the magic step.
you guys see gpt-4 generating impressive things and dismiss it because it's less than human intelligence, but it's damn close. dunno what kind of belief make you dismiss it like some dumb autocomplete
Because there is no intelligence, no reasoning, no magic. It will not make a "quantum leap" and suddenly become more than a statistical word analysis. It's just math and marketing, and there is nothing you can put in the ? step to make it so. It's been the same every "AI" hype cycle, and this cycle is not different from the preceding ones.
I said at the start of this hype cycle that LLMs will never be AGI, because they are not, and really cannot be. There will not be an exponential growth in capability with increase in data, and from what I've understood, they are already hitting the diminishing return part of the cycle pretty hard, where doubling data gives maybe a few percent in performance increase, and that will decrease pretty hard from here on.
And once VC figures out that it won't be a magical money printing machine this time either, this hype cycle will be over and we will have another AI winter.It needs to be a different tech to make the magical ? step happen, LLMs are unfit for anything beyond a really heavy autocomplete. It is a mathematical trick, a very fancy and expensive trick, but a trick none the less.
-
RE: I, ChatGPT
@sockpuppet7 said in I, ChatGPT:
@dkf said in I, ChatGPT:
@sockpuppet7 said in I, ChatGPT:
they seem to train with everything they got, and later use fine tuning to shape it, and that appears to work better than training with less, selected data, if I understand any of what I read about it
The problem is that means the only answers it can ever generate are projections from its training set. If that's what you're looking for, that's great. If you need anything that isn't just a projection of what went before, LLMs are exactly the wrong tool.
you should know it's BS, neural networks generalize and create new things on the patterns it learned
now you have billions of parameters, with a ridiculous amount of data to form connections, and it definitely write things unlike anything before
put some reinforcement learning over it to make it smarter with it's experience and we'll soon have our terminators, if global warming doesn't finish us before it
This is the socks gnome business plan, except it's about collecting data and still a ? on the magic step.
-
RE: The Official Status Thread
@topspin said in The Official Status Thread:
@sockpuppet7 most of the time I should be working, too. But honestly it feels like I stopped 4-5 years ago, or at least I’m as productive as if I had.
Ah, but you are intensely agile while you get nothing done!
-
RE: Sportsball WTF
@Dragoon I've been on a sail boat on the ocean during a thunder storm. That's quite the experience for all the senses.
-
RE: Nope, you eat it
Dr. James Udelson, chief of cardiology at Tufts Medical Center, confirmed to the AP that the chip could have played a role in the teen's death
The doctors are a lot less certain about the contribution than the article.