I, ChatGPT
-
@ixvedeusi said in I, ChatGPT:
@LaoC That thing is gold.
I had my fun with Copilot before I decided that it was making me stupider - it's impressive, but not actually suitable for anything more than churning out boilerplate.
Pretty much sums up my impression on the LLM / GenAI craze in general. Sure it's fun and does some impressive things when you're lucky and poke it the right way, but it just doesn't seem to produce anything much that would be of actual use to any specific problem.
I use GitHub copilot a lot, have fun with generated images, get a lot of useful information on it, etc
haters gonna hate I guess, and luddites gonna whatever luddites do
-
@sockpuppet7 said in I, ChatGPT:
luddites gonna whatever luddites do
probably whatever they where doing before the new shining time can around
-
@sockpuppet7 said in I, ChatGPT:
I use GitHub copilot a lot
I tried it for a while in the beginning, now I've turned it off because it couldn't just shut the fuck up and kept spouting random-ass made-up code whenever I pressed tab to get to the right indent level. Also I hate those "ghost code blocks" that constantly shift around all the text and make the actual existing code impossible to follow, it's all way too much useless distraction from just getting things done.
I suspect it could be useful if there was an actual "now go have a try at writing that boilerplate for me" key combo and it wouldn't mess with my code unless I explicitly told it to. IDK, maybe that exists now and I just haven't found it yet?
But apart from the crappy UI, the main problem is that it's just waaaay too unreliable, and reviewing code it much more tedious, boring, and error-prone than just writing the damn thing myself.
@sockpuppet7 said in I, ChatGPT:
have fun with generated images
Have been wanting to give that a try when I have some time to waste on my hands, it does look like there's some fun to be had there and it might genuinely be useful for e. g. D&D campaign artwork and such things. Still not what I'd call "of actual use to a specific problem" of any significance though, it's mostly fluff.
-
@LaoC said in I, ChatGPT:
9/10 good rant would read again
Most organizations cannot ship the most basic applications imaginable with any consistency, and you're out here saying that the best way to remain competitive is to roll out experimental technology that is an order of magnitude more sophisticated than anything else your I.T department runs, which you have no experience hiring for, when the organization has never used a GPU for anything other than junior engineers playing video games with their camera off during standup, and even if you do that all right there is a chance that the problem is simply unsolvable due to the characteristics of your data and business? This isn't a recipe for disaster, it's a cookbook for someone looking to prepare a twelve course fucking catastrophe.
If another stupid motherfucker asks me to try and implement LLM-based code review to "raise standards" instead of actually teaching people a shred of discipline, I am going to study enough judo to throw them into the goddamn sun.
Things. that remind you...
Pretty wise for one so young:
An executive at an institution that provides students with important credentials, used to verify suitability for potentially lifesaving work and immigration law, asked me if I could detect students cheating. I was going to say "No, probably not"... but I had a suspicion, so I instead said "I might be able to, but I'd estimate that upwards of 50% of the students are currently cheating which would have some serious impacts on the bottom line as we'd have to suspend them. Should I still investigate?"
We haven't spoken about it since.
They know exactly what their target market is - people who have been given power of other people's money because they've learned how to smile at everything, and know that you can print money by hitching yourself to the next speculative bandwagon.
Yup. That's the most succinct summary of the current AI craze.
-
@ixvedeusi said in I, ChatGPT:
Still not what I'd call "of actual use to a specific problem" of any significance though, it's mostly fluff.
Entertaining is important
-
@sockpuppet7 Lack of entertainment is not a problem that our current society has.
-
@ixvedeusi said in I, ChatGPT:
I tried it for a while in the beginning, now I've turned it off because it couldn't just shut the fuck up and kept spouting random-ass made-up code whenever I pressed tab to get to the right indent level. Also I hate those "ghost code blocks" that constantly shift around all the text and make the actual existing code impossible to follow, it's all way too much useless distraction from just getting things done.
That depends on the IDE you use, I think. I've been using it on intellij IDEs and it's not annoying me.
I won't say you've doing something wrong, it depends a lot on what you're coding. Maybe you don't have much boilerplate. Maybe your code is particularly hard for copilot to get right.
I won't be suprised if we eventually have programming languages adapted to be easier to LLMs
-
@LaoC said in I, ChatGPT:
9/10 good rant would read again
Coincidentally I learned today that our company has opened a "brainstorming initiative" where we can submit ideas about how we can make use of AI in our business.
Must resist the temptation to send in that link!
-
When any new tool is available there is a good chance of something really profitable and disruptive to appear, and people looking for ways to use this new tool makes total sense.
When phones with GPS and maps arrived, there was a lot of applications using your location, mostly disappeared. Uber remained and it's a major thing
-
@ixvedeusi said in I, ChatGPT:
@sockpuppet7 Lack of entertainment is not a problem that our current society has.
: Good entertainment though!
-
@sockpuppet7 said in I, ChatGPT:
When any new tool is available there is a good chance of something really profitable and disruptive to appear, and people looking for ways to use this new tool makes total sense.
When phones with GPS and maps arrived, there was a lot of applications using your location, mostly disappeared. Uber remained and it's a major thing
Uber is a terrible example. They still haven't made back the money they pissed up against the wall, and they're been legislated out of existence in Europe because they're so god awful...
yeah, probably a shining beacon of where this AI iteration will end up.
-
@sockpuppet7 said in I, ChatGPT:
I use GitHub copilot a lot
It's banned here at WTFCorp, which seems strangely at odds with the way they're aggressively pursuing GenAI for our customer support needs. They say they're worried about proprietary data leakage.
-
@LaoC said in I, ChatGPT:
9/10 good rant would read again
Most organizations cannot ship the most basic applications imaginable with any consistency, and you're out here saying that the best way to remain competitive is to roll out experimental technology that is an order of magnitude more sophisticated than anything else your I.T department runs, which you have no experience hiring for, when the organization has never used a GPU for anything other than junior engineers playing video games with their camera off during standup, and even if you do that all right there is a chance that the problem is simply unsolvable due to the characteristics of your data and business? This isn't a recipe for disaster, it's a cookbook for someone looking to prepare a twelve course fucking catastrophe.
This starts too well to be read on a computer screen. I'll shall put it on my kindle and read it in the sun.
Most of the market was simply grifters and incompetents (sometimes both!)
*edit
When I was younger, I read R.A Salvatore's classic fantasy novel, The Crystal Shard.
This is your yearly warning that R A Salvatore really has written a lot of crap.
However, I do have the technical background to understand the core tenets of the technology, and it seems that we are heading in one of three directions.
The first is that we have some sort of intelligence explosion, where AI recursively self-improves itself
I'm going to have to create a sign.
A second outcome is that it turns out that the current approach does not scale in the way that we would hope, for myriad reasons. There isn't enough data on the planet
A stunning return to form.
Having your team type in import openai does not mean that you are at the cutting-edge of artificial intelligence no matter how desperately you embarrass yourself on LinkedIn and at pathetic borderline-bribe award ceremonies from the malign Warp entities that sell you enterprise software.
I'm going to steal that one.
I shall answer this as politically as I can... there are those that have drunk the kool-aid. There are those that have not. And then there are those are that are trying to mix up as much kool-aid as possible. I shall let you decide who sits in which basket.
I have to admit I love that piece of writing. He has managed to take a well understood analogy and bent it to sound very insightful but has actually said nothing of substance. And then the writer gets stuck right into it. I would be proud to steal it and pass it as my own. I need to know who came up with it.
No, what I hate is the people who have latched onto it, like so many trailing leeches, bloated with blood and wriggling blindly.
Such a beautiful way with words.
This entire class of person is, to put it simply, abhorrent to right-thinking people. They're an embarrassment to people that are actually making advances in the field, a disgrace to people that know how to sensibly use technology to improve the world, and are also a bunch of tedious know-nothing bastards that should be thrown into Thought Leader Jail until they've learned their lesson, a prison I'm fundraising for
-
@DogsB Who watches the statistical model that watches your statistical model?
-
@error said in I, ChatGPT:
@sockpuppet7 said in I, ChatGPT:
I use GitHub copilot a lot
It's banned here at WTFCorp, which seems strangely at odds with the way they're aggressively pursuing GenAI for our customer support needs. They say they're worried about proprietary data leakage.
No proprietary data was harmedutilized in any way in the closing of this support ticket.
-
@error said in I, ChatGPT:
@sockpuppet7 said in I, ChatGPT:
I use GitHub copilot a lot
It's banned here at WTFCorp, which seems strangely at odds with the way they're aggressively pursuing GenAI for our customer support needs. They say they're worried about proprietary data leakage.
We recently got an email saying "GitHub Copilot Enterprise Service is now available..."
I've ignored it.
-
@DogsB said in I, ChatGPT:
Uber is a terrible example. They still haven't made back the money they pissed up against the wall, and they're been legislated out of existence in Europe because they're so god awful...
Uber got a lot wrong, but Bolt and Liftago seem to be well and alive here. They work more like traditional taxi services except through app …
.That has nothing to do with GPS though, but rather with the shared whatever commie wet dream. GPS was overall pretty successful.
-
@ixvedeusi said in I, ChatGPT:
@sockpuppet7 said in I, ChatGPT:
I use GitHub copilot a lot
I sometimes use Codeium instead, because it has a more practical free tier.
I tried it for a while in the beginning, now I've turned it off because it couldn't just shut the fuck up and kept spouting random-ass made-up code whenever I pressed tab to get to the right indent level.
I don't do that much, so I don't mind as much either. The auto-indent is usually fine and then I use format-current-file instead.
Also I hate those "ghost code blocks" that constantly shift around all the text and make the actual existing code impossible to follow, it's all way too much useless distraction from just getting things done.
I suspect it could be useful if there was an actual "now go have a try at writing that boilerplate for me" key combo and it wouldn't mess with my code unless I explicitly told it to. IDK, maybe that exists now and I just haven't found it yet?
But apart from the crappy UI, the main problem is that it's just waaaay too unreliable, and reviewing code it much more tedious, boring, and error-prone than just writing the damn thing myself.
I was pleasantly surprised most of the time the code it spouts makes it past the borrow checker. That's quite a feat.
But for longer snippets it does not really have either enough information nor sufficient reliability to get them right. It does save some looking up of correct methods though—if it gets it right, go on, if it gets it wrong, I go to the documentation as I'd have to anyway.
@sockpuppet7 said in I, ChatGPT:
have fun with generated images
Have been wanting to give that a try when I have some time to waste on my hands, it does look like there's some fun to be had there and it might genuinely be useful for e. g. D&D campaign artwork and such things. Still not what I'd call "of actual use to a specific problem" of any significance though, it's mostly fluff.
I have used it (specifically craion) to generate pictures on Christmas and birthday present packages (the presents are home-made too, but that process does not involve computers beyond electronic scales) for a few years now, so that's an actual use for a specific problem though very minor one.
-
@topspin said in I, ChatGPT:
@ixvedeusi said in I, ChatGPT:
@sockpuppet7 Lack of entertainment is not a problem that our current society has.
: Good entertainment though!I don't understand the emoji here, asked the GPT, it said:
"Good entertainment, though it can sometimes be noteworthy (🏴), require a drink (🥃), feel like it's bad or going down the drain (🚽), and often be absurd or random (🦆)."
-
@cvi said in I, ChatGPT:
@DogsB Who watches the statistical model that watches your statistical model?
∞
-
@error said in I, ChatGPT:
@sockpuppet7 said in I, ChatGPT:
I use GitHub copilot a lot
It's banned here at WTFCorp, which seems strangely at odds with the way they're aggressively pursuing GenAI for our customer support needs. They say they're worried about proprietary data leakage.
I think most companies overvalue their code. There isn't much one could do with the code I write at work without all the rest of the company
If code was that valuable there would sometimes be the odd programmer moving to another country and getting rich with a copy of the code. That I saw happening about 0 times
-
@sockpuppet7 just like some locals here, the Critical Drinker sometimes talks about drinking Toilet Duck to drain away the shitty entertainment.
-
@topspin said in I, ChatGPT:
@sockpuppet7 just like some locals here, the Critical Drinker sometimes talks about drinking Toilet Duck to drain away the shitty entertainment.
hanzoed before I got to complain that I've been saying that shit since 04. I think it comes from Fr Ted but I can't think of anything before that.
-
@sockpuppet7 said in I, ChatGPT:
@error said in I, ChatGPT:
@sockpuppet7 said in I, ChatGPT:
I use GitHub copilot a lot
It's banned here at WTFCorp, which seems strangely at odds with the way they're aggressively pursuing GenAI for our customer support needs. They say they're worried about proprietary data leakage.
I think most companies overvalue their code. There isn't much one could do with the code I write at work without all the rest of the company
If code was that valuable there would sometimes be the odd programmer moving to another country and getting rich with a copy of the code. That I saw happening about 0 times
Code isn't so much valuable as much as what it does and learning from it. If you see someone moving onto a better-paying job, it's usually because they learned what's wrong with that code in the vain hope of doing better elsewhere.
-
@DogsB said in I, ChatGPT:
This entire class of person is, to put it simply, abhorrent to right-thinking people. They're an embarrassment to people that are actually making advances in the field, a disgrace to people that know how to sensibly use technology to improve the world, and are also a bunch of tedious know-nothing bastards that should be thrown into Thought Leader Jail until they've learned their lesson, a prison I'm fundraising for
I’m going to piss you off with another quote that goes well with this:
a bunch of mindless jerks who were the first against the wall when the revolution came.
ETA: fuck it, the whole thing belongs in this thread:
The Encyclopedia Galactica defines a robot as a mechanical apparatus designed to do the work of a man. The marketing division of the Sirius Cybernetics Corporation defines a robot as “Your Plastic Pal Who’s Fun to Be With.” The Hitchhiker’s Guide to the Galaxy defines the marketing division of the Sirius Cybernetics Corporation as “a bunch of mindless jerks who’ll be the first against the wall when the revolution comes,” with a footnote to the effect that the editors would welcome applications from anyone interested in taking over the post of robotics correspondent. Curiously enough, an edition of the Encyclopedia Galactica that had the good fortune to fall through a time warp from a thousand years in the future defined the marketing division of the Sirius Cybernetics Corporation as “a bunch of mindless jerks who were the first against the wall when the revolution came.
-
@DogsB said in I, ChatGPT:
@sockpuppet7 said in I, ChatGPT:
@error said in I, ChatGPT:
@sockpuppet7 said in I, ChatGPT:
I use GitHub copilot a lot
It's banned here at WTFCorp, which seems strangely at odds with the way they're aggressively pursuing GenAI for our customer support needs. They say they're worried about proprietary data leakage.
I think most companies overvalue their code. There isn't much one could do with the code I write at work without all the rest of the company
If code was that valuable there would sometimes be the odd programmer moving to another country and getting rich with a copy of the code. That I saw happening about 0 times
Code isn't so much valuable as much as what it does and learning from it. If you see someone moving onto a better-paying job, it's usually because they learned what's wrong with that code in the vain hope of doing better elsewhere.
My point is that a leak of a snippet of code would be a non-issue for 99.9% of corpos, yet some protect it like mad
-
@sockpuppet7 said in I, ChatGPT:
If code was that valuable there would sometimes be the odd programmer moving to another country and getting rich with a copy of the code. That I saw happening about 0 times
I do tend to share code I write for my hobby projects with work and vice versa, which is probably dicey from a legal perspective.
-
@sockpuppet7 said in I, ChatGPT:
My point is that a leak of a snippet of code would be a non-issue for 99.9% of corpos, yet some protect it like mad
That's because it falls under the heading of IP. They're very protective of IP because they can sue people over it. I.E. Oracle sueing Alphabet over for loops.
@error said in I, ChatGPT:
I do tend to share code I write for my hobby projects with work and vice versa, which is probably dicey from a legal perspective.
especially if they think its on their time.
-
@DogsB said in I, ChatGPT:
They're very protective of IP because they can sue people over it. I.E. Oracle sueing Alphabet over for loops.
We (WTFCorp) got sued for a really bullshit submarine patent, that basically covered embedding an application (Flash player in our case) on a website. I believe they conceded a
ransomlarge cash settlement.
-
@topspin said in I, ChatGPT:
@DogsB said in I, ChatGPT:
This entire class of person is, to put it simply, abhorrent to right-thinking people. They're an embarrassment to people that are actually making advances in the field, a disgrace to people that know how to sensibly use technology to improve the world, and are also a bunch of tedious know-nothing bastards that should be thrown into Thought Leader Jail until they've learned their lesson, a prison I'm fundraising for
I’m going to piss you off with another quote that goes well with this:
a bunch of mindless jerks who were the first against the wall when the revolution came.
What he said about that in the rant was quite interesting. The devs on the project were basically making themselves unemployable by working on it for five or six years because they'd become a useless niche. The con artist at the top, on the other hand, would move to the next bubble. Their skills would be more transferable.
ETA: fuck it, the whole thing belongs in this thread:
The Encyclopedia Galactica defines a robot as a mechanical apparatus designed to do the work of a man. The marketing division of the Sirius Cybernetics Corporation defines a robot as “Your Plastic Pal Who’s Fun to Be With.” The Hitchhiker’s Guide to the Galaxy defines the marketing division of the Sirius Cybernetics Corporation as “a bunch of mindless jerks who’ll be the first against the wall when the revolution comes,” with a footnote to the effect that the editors would welcome applications from anyone interested in taking over the post of robotics correspondent. Curiously enough, an edition of the Encyclopedia Galactica that had the good fortune to fall through a time warp from a thousand years in the future defined the marketing division of the Sirius Cybernetics Corporation as “a bunch of mindless jerks who were the first against the wall when the revolution came.
READ ANOTHER BOOK!
-
@error said in I, ChatGPT:
@DogsB said in I, ChatGPT:
They're very protective of IP because they can sue people over it. I.E. Oracle sueing Alphabet over for loops.
We (WTFCorp) got sued for a really bullshit submarine patent, that basically covered embedding an application (Flash player in our case) on a website. I believe they conceded a
ransomlarge cash settlement.I sometimes wonder if there would be more money in moving into law but equilivant salaries would require north of 39 hour weeks. I would probably have to work it too.
-
@error so, exactly the main way flash player was intended to be used and was used by countless pages. Great thing to have a patent on.
I wonder if I should file a patent for insulting someone over the internet. Prior art doesn’t seem to be a problem.
-
@topspin said in I, ChatGPT:
@error so, exactly the main way flash player was intended to be used and was used by countless pages. Great thing to have a patent on.
US Patent No. 7,599,985, "Distributed hypermedia method and system for automatically invoking external application providing interaction and display of embedded objects within a hypermedia document."
-
@topspin said in I, ChatGPT:
Prior art doesn’t seem to be a problem.
At least not in the US patent system. USPTO pretty much rubber-stamps patent applications, even with obvious BS. They may still throw out applications that claim time travel or violations of thermodynamics, but prior art? Issue the patent and let the courts sort it out.
-
@DogsB said in I, ChatGPT:
@error said in I, ChatGPT:
@DogsB said in I, ChatGPT:
They're very protective of IP because they can sue people over it. I.E. Oracle sueing Alphabet over for loops.
We (WTFCorp) got sued for a really bullshit submarine patent, that basically covered embedding an application (Flash player in our case) on a website. I believe they conceded a
ransomlarge cash settlement.I sometimes wonder if there would be more money in moving into law but equilivant salaries would require north of 39 hour weeks. I would probably have to work it too.
Should be a lot of difference from country to country, but I know lawyers that struggle to make ends meet and some that makes bank. I think the variance is higher than IT
-
@sockpuppet7 said in I, ChatGPT:
My point is that a leak of a snippet of code would be a non-issue for 99.9% of corpos, yet some protect it like mad
You don't get it. They don't protect it because it's highly valuable ; they protect it because they don't want everybody to know how
their code is. 
-
@Zerosquare said in I, ChatGPT:
@sockpuppet7 said in I, ChatGPT:
My point is that a leak of a snippet of code would be a non-issue for 99.9% of corpos, yet some protect it like mad
You don't get it. They don't protect it because it's highly valuable ; they protect it because they don't want everybody to know how
their code is. That's a reason a lot of people don't opensource some things, they don't want to make the code look good enough to be viewable to the world
-
-
@DogsB said in I, ChatGPT:
I sometimes wonder if there would be more money in moving into law but equilivant salaries would require north of 39 hour weeks. I would probably have to work it too.
You would also have to renounce your membership as a human and embrace being a slime mold.
-
@sockpuppet7 said in I, ChatGPT:
My point is that a leak of a snippet of code would be a non-issue for 99.9% of corpos, yet some protect it like mad
The problem is that most people don't understand "code". For bean counters, it has to be valuable, since they throw a ton of money at producing and maintaining it. So it has to be protected. The other option is that it's not valuable, i.e., worthless, ... so why are they paying people for it again?
OK, this is a bit exaggerated, but the point stands.
Yes, I went through this shit quite a few years ago when people insisted that we'd "encrypt" some shader code (nevermind that you have to decrypt it to hand it to underlying standard API, making it trivial to intercept). At the time, my immediate boss at the time slapped a ROT13 or something on it and told me to stop explaining why it's useless and shut up. (Apparently the fact that you can easily "steal" such code was deeply unsettling to management. People with the knowledge to do such things ... very suspicious and Not Good(tm).)
-
@cvi what kind of things you code that involve shaders? seems interesting (if it's pii you can rot13 your answer)
-
@cvi So that's why memfrob(3) exists.
-
@ixvedeusi said in I, ChatGPT:
@DogsB Great, another one of those articles where the entire content is just variations of the headline repeated over and over.
Well, like the guy quoted in the article said: "LLMs are highly capable of saying the same thing in many different ways".
-
-
@cvi ROT13 is useless, but if you XOR it with something slightly less trivial, it will at least not show up in
stringsoutput.
-
@topspin said in I, ChatGPT:
@LaoC said in I, ChatGPT:
@DogsB said in I, ChatGPT:
New study finds way to spot AI hallucinations - TechCentral.ie
Researchers have found a new method which could be used to help spot when generative AI is likely to ‘hallucinate’ – where it invents facts because it does not know the answer to a query – and help prevent such incidents occurring. A new study by a team from the University of Oxford developed a...
There's never a "you need a statistical model to look after your statistical model" meme when you need one.
Hey, that but with a duck-duck-goose theme!
-
@cvi said in I, ChatGPT:
@DogsB Who watches the statistical model that watches your statistical model?
If police police police police, then who police police police?
Police police, police police police, police police police.
-
@djls45 s/police/buffalo/g
-
@topspin said in I, ChatGPT:
@sockpuppet7 just like some locals here, the Critical Drinker sometimes talks about drinking Toilet Duck to drain away the shitty entertainment.
Oh, I thought it was something no-true-scotsman something that I somehow didn't fully catch.
-
@topspin Yeah, you could do that. This was in the OpenGL|ES era of Android. You pretty much hand over the shader source as a string to OpenGL via glShaderSource(). You can just intercept that call, assuming you don't have a graphics debugger that outright hands you the sources.
(There was stuff like ShaderBinary(), but since it was before SPIR/SPIR-V was standardized, it was kinda tricky to use in a portable way. Plus, GL|ES devices at that time sucked, so reverse engineering the shaders wouldn't have been that bad.)
