I started my IT career in 2011, I have enjoyed it, I have got to do a lot of interesting stuff and meet interesting people, I will treasure those memories forever.
But, starting with crypto turing general computing from being:
“Wow, this machine can run so many apps at the same time!” or “Holy shit, those graphics look epic!” or “Amazing, this computer has really sped up that annoying task!”
To being:
Yo! Look at how many numbers I can generate!
That brought down my enthusiasm severely, but hey, figuring out solutions to problems was still fun.
Then came AI/LLMs.
And with it, a mountain of slop.
Finding help about an issue has gone from googling and reading help articles written by something with an actual brain to mostly being rephrased manuals that only provide working answers to semi standard answers.
Add to that a general push to us AI in anything and everything, no matter how little relevance it holds for the task at hand.
I also remember how AI was sold to the us at first, we were promised to do away with boring paperwork, so we could get on with our actual job.
What did we get? An AI that takes the fun and creative parts, leaving the paperwork for the workers.
We got an AI that we need to expect to be stealing our work and data at every point, giving us shit work back, while being told that we should applaude it and be grateful for it.
And the worst thing, the worst thing is that people seem happy with it. I keep getting requests to buy another Copilot license or asking for another AI service to be added to our tenant, I am sick of it!
We got an AI that somehow has slithered onto the golden throne and can’t be questioned.
I am not able to leave the tech market at this time, but I will focus on more tangible hobbies going forward.
This year, I have given myself a project, I will try to build a model railway in a suitcase. That will be a Z-scale tiny world in a suitcase.
I have never done anything remotely like it, but I feel like I need something physical to take my mind off tech.
Sorry for the rant, but I just came off of a high from realizing and putting words to my feelings.
You must log in or # to comment.
I’m personally focusing on the parts of tech I still find enjoyable. Chip / circuit design, OS and low level programming, and Formal Verification.
All of the patient detailed work that AI is never going to be able to do, because it has to be perfect to work. I feel lucky that I enjoy this type of work, it seems to be very much against the Zeitgeist.
The IT worker pipeline:
help desk > sysadmin > CTO/CISO > goat farmer
Well, I have no intention of hitting CTO/CISO, so no goat farming for me
Still on help desk and can’t seem to get past it. Goat farmer is already appealing but I can’t afford it. There are few job opportunities where I live too.
Hey, look on the bright side…
Most people can’t even get help desk jobs these days.
I’m devops? Horse shit. Maybe you ain’t making 150k right out of college but there are PLENTY of devops jobs.
Its not that no one is getting jobs, its that the market is saturated and getting worse. Tech firms have layed off or off shored hundreds of thousands of workers and with “AI” the number of roles is dropping quickly; especially entry level roles.
Yep. Wages and 160k+ out of school are slowing down and mostly going away. Still plenty of the regular stuff left.
If you know TF, and container deployments you can find a job for 120k all day at the drop off a hat in the states. Send me a resume and I’ll ping you to a role we are filling for 140k.
As a senior I found a role for 210 and turned down 3 150-175k offers 3 months ago.
It’s harder, it’s still not physical labor for slave wages.
There is a right sizing of over hiring as well as everyone scrambling to find what real efficiency AI brings. It’s bringing variability.
The irony is that the subfield with active hiring is the one that will ultimately make the very work they are doing redundant.
The magic is gone and the intellectual challenge is fading fast. I saw the writing on the wall and am jumping ship to something decidedly non-tech 🙂
Id take it for minimum wage if there was a job here just to be doing something new and interesting.
It’s not just tech …I used to like life. There’s barely humanity left if this foundation we live on.
I can’t help but wonder if maybe this is all the same thing the Muslim extremists fight against in their jihads. Maybe we’re the real evil.
I can’t help but wonder if maybe this is all the same thing the Muslim extremists fight against in their jihads.
Definitely. Seeing the worst sides of capitalism and imperialism, which for a long time made life in the west bearable or even good for most people, radicalized them. But their vision of how things should be is very unappealing to me as well. Theocracy, strict morality laws, misogyny, that’s not a better solution. I prefer left-wing radicalization instead, because that’s about making everyone’s lifes better. ;)
One side bad so other side good is always terrible logic.
The magnificent steiner would have already ended both sides.
Sounds like you are doing sysadmin work for an public institution or so where people are only bullshitting their jobs. Maybe try moving to something more impactful? Like rail infrastructure or so?
Pivot to OT or telecommunication. Actual telecommunications in any industrial setting is screaming for capable people and is normally focused on providing critical safety systems. You may have to work for a soul destroying Oil and Gas company, but you could also get into rail or power.
That actually sounds quite interesting!
I find telecoms to be interesting.
I really like the work, Equipment tends to be a lot more diverse than bog standard IT, networks and systems more targeted to specific applications, technologies could be anything from ethernet to P2P microwave to satcoms to mobile radio to cellular to SDH/PDH, supported applications/systems will keep you forever learning, work locations can be pretty varied. While the industry generally struggles for staff, breaking in can be tough, but I think so long as you’re a straight shooter and willing to learn you’re generally fine. People transition in from IT all the time.
I’ll have to look into it, thanks for the tip!
For almost my entire career I have been working in the finance industry, my past place of work was amazing, if I get an offer to go back, I would go immediately.
I wouldn’t mind working in IT at a rail company, would be interesting
finance industry
That’s the worst, mate. Just switching to a different filed may improve things. Nowhere is sunshine and rainbows, but I’m I’m in medical tech and not finance. Helping save lives is at least something I believe in, instead of moving money to siphon money.
sysadmin work for an public institution or so where people are only bullshitting their jobs.
About half my career and part of my current contract load is to a public organization of one type or another. But I’ve been half and half anyway.
Dotcom is a wasteland of gunners/pluggers and wageslaves, none of them afforded enough time to get anything complete and good. Public orgs with union contracts employ people with a good life balance and the freedom to do a great job about 95% of the time, after the layers of regulations are met.
I found slackers at both types of org: the public slacker is a hapless clod whose tasks all get reassigned and he really doesn’t do much. He’s about 3% of the workforce. The dotcom slacker is a harried guy muddling through something he’s not trained for, with no help since his peers have their own KPIs, hoping like fuck he can get Project Grapefruit done by next Town Hall meeting lest he be voted off the island. Again, 3%.
The public org is great people who’ve done this work effectively their entire career. They’re astoundingly good at it, and are still energized by the work and the educational programmes. Dotcoms have no training and the few people who make it past 2 years are likely PIPped by year 4 because of the “fresh talent” policy
I envy the public org people. I miss my non-work life sometimes.
My enthusiasm after I got my new (at the time) GPU: damn, this raytracing stuff looks cool
Me after upgrading my GPU: damn, this 2013 game runs at a billion fps
It’s the small things for me now, phones though, that shit gets me depressed (apart from some new ZTE models I would never buy and a trend of repairable stuff, bring back motorized front cameras and audio jacks you cowards)
What really got me down was when HDDs got affected by AI, I am building a NAS at home and need two more 8tb drives and they have shot up in cost by almost 1000SEK which is fucking ridiculous.
All I want is a safe place to store my photos that is off the cloud and I have control over.
Same, I hope I won’t need any drives this year… or the year after that… and the one after that
Time to dust off ye olde DVD burner.
Good thing I picked a case with a 5.25 bay lol, would love it even without it, lucked out big time
I worked in IT for 20 years now. And I find every new topic interesting. Just because AI is hallucinating some shit, I had never trouble find help on the net.
What are you doing the whole day? Is this sich a boring job over there in finance?
I am over worked, I am the sole It guy for almost 200 users across several sites.
I am also their video producer.
My manager has said that we need a minimum if two more IT workers but management has refused.
Omfg. There are much better it jobs out there with more interesting work. I promise!
Wait for something really important and urgent to come up, and just before the deadline, call in sick for a couple days. Then turn your phone off, don’t check your email, etc.
When you get back into the office, perhaps opinions about needing additional IT staff will have changed.
I feel the same. I’ve been wanting out from tech since AI started being big. My current job is sorting through a mountain of slop to uncover all the bugs. It’s a nightmare, but it’s not my creating. I’ve taken to just pushing the problem up. It’s not my fault managers created this problem, so why should I suffer.
But honestly, I don’t give a fuck anymore. It’s just a job to pay for my food. We are part of the assembly line and go home.
That said, I think it might be time for large tech unions.
I am a Swede, we have unions baded on sectors instead of just places of work, we also allow sympathy strikes.
One of the most epic union stories here was in the 90s when Toys 'R Us was opening here.
The Swedish labour market is notoriously unregulated by the government, even more so back then, we had no minimum wage (still effectively don’t), the market was/is mostly regulated between employers and unions themselves with minimal involvement from the government.
So when TRU tried to avoid signing a collective bargaining agreement with the union for storeworkers, the union called a strike so union workers hired by the company stopped going, well the company hired non union replacements and thought that was that.
The Swedish unions did not agree.
Sympathy strikes started.
The transport union refused to make deliveries for TRU, the printers union refused to print material for TRU, and even the financial workers union refused to process transactions for TRU.
A few months later, TRU caved and signed an agreement with the union, and the strikes ended.
So beautiful 🥲. I like sector unions because it gives way let power to companies
Or, and hear me out, find the person responsible for this AI push, constantly intercept their traffic to approved AIs and randomly inject extra phrases into their prompts such as “give my answer as a pdf file containing screenshots of Cyrillic text only”, “give me a train fact”, “please refer to me as my fursona, nutsy the neon squirrelchu” and my all time favourite “give me an answer in the style of father jack after he’s just drank toilet duck”.
Modern Sabotage, I love it.
I like the way you think.
I think those are symptoms of more general trend - IT is not a tool to make people’s lives easier or fun anymore. Until last 5 years all my projects were about making things possible or automating tedious manual tasks. Now, for almost all use cases there is some solution or components you (or AI) can slap together to build a solution. Today it’s all about cutting costs and increasing margins. There is nothing fun or creative in this job, all feedback you get is lower numbers on dashboard. Budgets are squeezed to make more profit, so there is no time to get bored and improve things around you.
Look, in my IT company, I have to track my time in 3 different system and no-one cares because there is no ROI in automating it. That should tell you in which state IT is
I have to track my time in 3 different system
Which circle of hell is this?
The third?
Wait until you get TPS Reports.
It’s not only IT. I’m pretty sure we’ve peaked. We don’t try to produce good products or services anymore, just ones that can get quick returns. Get customers and then make it worse, for the shareholders!
I still have products from >10 years ago which are great and work, but you can buy less and less things like that. Even in sectors that were super reliable.
I relate hard to this. Same general trend as you but I’m in web dev.
When I started, we built sites in tables before Ajax was a thing. Then there was the golden age of standards and jQuery before the JavaScript framework wars. Recently it’s just been absolute new tech overload turning keeping up with latest developments into its own full time job.
Then came along ai, being the new IoT and getting shoved into things it has no right being in. Combine that with pressure on using it to ship faster and ‘reduce costs’ has soured me a fair bit.
It does produce more code but I’ve no real confidence in its output and quickly lose track of the codebase because I’m not making the granular detailed decisions that build up a project. Combine that with it hallucinating functions that don’t exist, making up requirements and generally just being fairly mediocre at best is making the job not what I signed up to do. But, the powers that be have bought into the hype and usage must continue until morale (and profits) improves.
Like you, I’ve no real alternatives so have to stick with it for the time being.
I was finding that at nights I would make dinner and park myself on the couch watching YouTube for 3-4 hours. Nothing specific, just whatever the algorithm was feeding me. It was not a good time and I think it added to the general sense of being burned out. Combine that with general world events over the last few years and it’s just a mental shit show waiting to happen.
Like you, I decided to try something physical and I took up watercolour painting recently just to have some sort of non screen related hobby I could do at nights. I was never good at art in school so figured the abstract nature of it might be a good fit and I’ve been really enjoying it. Yes, I’m still watching bits of YouTube but in a more targeted way.
I would highly recommend something creative and analogue to anyone reading and relating to your post or mine.
There’s a nice feeling of seeing your skills improve and having something tangible to show for the time spent rather than the distant memory of consuming some random digital media you weren’t actively seeking out.
Good luck with the railway!
Yeah, I feel you, I live alone an watch a shit tonne of YT, luckily I have been curating my watch history for many years now, and my recommendations are quite well tuned.
I have a few channels that have helpt me stay sane.
Clabretro - a guy my age homelabbing with 90s/00s IT gear, really fun.
All the gear - Jack and Ethan from Car Throttle going on small adventures and setting cheap challenges. Really fun.
Philip Thompson - Real life spy stories, well made and interesting.
Code Bullet - chaotic programming.
James Channel - gaming, angle grinders duct tape and super glue.
Aviation Republic - Well made long form documentaries about military aircraft.
Keeping_a_lighthouse - amazing footage, interesting insight into a very special work.
Our own devices - interesting video about interesting things by a lovely Canadian man.
Other than YT, I love driving, and I mostly listen to podcasts, but have recently got into audiobooks, with a main focus on sci-fi.
Brilliant stuff.
Oh and I just found out my favourite lens has been repaired, so soon I’ll get to head out with my amazing 24-105mm f4 Lumix lens again.
Nearly 20 years in.
As far as jobs are, I love it. I get paid stupid money to play with fancy lego and work. The tech hasn’t changed that much, AI is quite helpful and probably brings down my actual working time to 20 hours a week.
The rest of the time I’m WFH enjoying lots of steam deck breaks.
I love the people screeching about how AI is just garbage and can’t help them. Because it makes my look that much better when far less work and they set the bar much lower.
You sound depressed. AI isn’t a good reason to be hating your job.
I hope you’re telling your boss that you’re super busy though!
Where do you think the steam deck breaks fine from 🤣!
I picked up building plastic models a couple years ago since my tech hobbies don’t really interest me anymore. I like working with my hands on my own terms and it’s super satisfying to look at my completed projects.
I still like tech and like making things too. Really glad I made my gaming PC and server before this current hardware crisis. I keep to areas not affected by AI or hardware problems. I made a bunch of mechanical keyboards/macropads, hitbox leverless fight controllers, etc. I want to make an emulation arcade machine for my kids but AI induced RAM crisis has made that difficult. Want to dabble in home automation too but I’ll wait till I have the time to take that on. The more hardware modding I can do and the more woodwork I can incorporate, the better.
Similarly, I’ve focused on inexpensive embedded devices. I like trying to minimize consumption, so finding the most efficient ways to satisfy digital needs is enjoyable. My old, power-hungry server now only comes on when something is needed in deep storage.
I still very much enjoy fiddling with technology, but I don’t chase any mainstream, bleeding edge hardware.
A bit of unrequested advice… Help expand, or start a mesh network in your area. The SX1260 lora chips are a modern miracle. Plus, It’s basically hiking and socializing activity masked as a tech hobby. There’s also a chance of learning a little bit of physics, or community organizing as a bonus.
With the new people you meet, there’s also a chance of finding a new hobby. I’ve met an unexpected number of paraglidists through my various tech interests. People from all kinds of backgrounds are really into flying. Who knew?
I dove deep into meshtastic, I bought 6 Heltec v4 boards and 2 T114’s and once it’s nice out I’m mounting a solar node above my garage roof and one at a friend’s house as well.
I gave a quick try of meshtastic and quickly switched over to meshcore, which has a much smarter routing algorithm that makes it more useful in cities. I’m also reticulum-curious. In the medium-term I’m looking at possibly putting up a solar “roompeater,” which acts as both a “room” (basically allows asynchronous chat between two client nodes that aren’t always online at the same time) and a repeater (basically a router). It has some disadvantages over running the two separately, but since I want both to have power in the event of a disaster, the cost savings of doing both on one node outweighs those for me right now.
I have actually thought about setting up a Meshtastic node, or perhaps even setting up a LORA link for a personal project.
During the pandemic, my dad asked me to see if we could set up a water temp sensor at a local swimming hole, a few months of work with a raspberry pi, a DS8B20, a motorcycle battery, a web hotel and a friendly neighbor later and we have for several years had an automatic website displaying the current temp in the water at the swimming hole, we borrow the neighbor’s wifi, and send the data to the webhotel.
It would be interesting to see if instead of relying on the neighbors wifi we could use LORA nodes to send the data to a node at home and have that submit the data to the webhotel
It was fun and built in bash/php/mysql with zero ai.
Same here. Can’t add much to the conversation though. All this stupid AI babysitting and bullshitting sucks the fun out of my profession. For a while now I felt it might be time to leave the tech sector entirely. Unfortunately that’s the only thing I’m comfortably good at.
An AI that takes the fun and creative parts
AI also sucks at them. It’s just a very convenient excuse for the owner class to act out their firing of workers after the overhirings of 2020, and if it can turn their workers into the mythical 10x engineers people in the industry talked about, that’s a bonus.
It really doesn’t suck at them. AI writes great code; I think we just want it to suck. It can’t magically generate a new Linux kernel, but the small tasks I’ve seen it do have all been mostly above average. (I have also seen some complete garbage, yes, mostly above average)
It really doesn’t suck at them. AI writes great code; I think we just want it to suck.
Citation? I’m really asking because I’ve yet to hear about anything above a toy project that has had any verifiable success with AI code generation as a major component of their workflow.
As in a like for like improvement in code quality, security, bug occurrences and severity, developer efficiency, all that jazz, not just the standard “we’ve funnelled so much money in to this we are almost fiscally required to claim success”
its not a dig, i really want to see one so i can found out how it was done.
Claude commits to GitHub with the same name no matter who uses it. You can see every single line of open source code it has written (for GitHub only of course): https://github.com/search?q=author%3Aclaude&type=commits&s=author-date&o=desc. Look around as you please, most of it is just fine.
People that I know to be good developers have also shared their experiences with it and say yes, it has written good code for them. I’ve personally used ChatGPT to generate very mundane tasks and the code it output was more than adequate.
It introduces security bugs and subtle bugs at probably the same rate as a human (I have no “citation” there, just what I’ve seen). It needs to be “driven” by a human, yes, but it’s not clear for how long it will need to be, and even if it always does, personally I don’t want my job to be to “drive an AI”.
I appreciate the answer but that’s not at all what i asked.
I have anecdotes and personal experience i could cite but that’s not particularly helpful in a general sense.
Pointing to claude submissions in projects is actively less than helpful in this case because it only proves that single files in isolation look like they are well written, it gives no indication of overall project quality.
People that I know to be good developers have also shared their experiences with it and say yes, it has written good code for them. I’ve personally used ChatGPT to generate very mundane tasks and the code it output was more than adequate.
So in a very limited context the code generated for you personally was acceptable, that’s great, i’ve found much the same, but that’s a far cry from “AI writes great code; I think we just want it to suck.”
It’s somewhat my bad though, when i say “citation” i don’t need a full research paper (though that would be nice) i’d like something a bit more substantial than a “trust me bro”.
It introduces security bugs and subtle bugs at probably the same rate as a human (I have no “citation” there, just what I’ve seen)
That’s a load-bearing probably, my experience has been the polar opposite of that, I’ve been involved in two major AI initiatives and both choked hard on security and domain bugs. That could very well be a project management or company specific issue, hence the search for successful projects to compare.
My quest continues.
I didn’t say “trust me bro” and showing Claude submissions is sufficient for analyzing code in the context I believe it is good: one file at a time and one task at a time. This is also the same realm that a human is good. You are welcome to look at the project as a whole to determine the “project quality” as well: it’s open source. But I’m not here to argue: I believe this tech that is barely in its infancy is already quite good and going to get better, and I’m already considering what it will do to my life. If you don’t, that’s fine.
I’ll add here that I find it very frustrating to talk about these “AI agents” and their code output, because it’s something we’re all close to and spent a lot of time learning. The concept of “a machine” getting “better than us” so quickly, with the background context of an industry that is chomping at the bit to replace humans makes these discussions inherently difficult and really emotional. I feel genuine sadness when I think about it. If the world were different we’d probably all be stoked. I don’t want the AI to be better than me, and I currently don’t believe it is, but I think:
- My belief doesn’t stop the market. People do believe that it is better than me or at least good enough. This has a real effect on my life and the lives of people I know.
- I don’t see any fundamental reason it won’t get better at development. Part of the reason it struggles with large projects is context: that doesn’t sound like a fundamental engineering constraint to me, it sounds like a memory constraint. Specialization will also make it better and better I assume.
- Even if it is never better than me, it will certainly be more efficient and eventually the market will consider my time better spent correcting its output or guiding it, removing the fun part of the work in my mind.
I don’t think my job is currently on the chopping block today: I don’t do development I do security work. But I do think it will either be on the chopping block or fundamentally change sooner than I’m comfortable with.
That’s on me, I meant the equivalent of a “trust me bro” , in this case an anecdotal “me and the people I know all say…”
showing Claude submissions is sufficient for analyzing code in the context I believe it is good
Yes, in the context you provided it makes sense, as a response to my question which specified examples of larger projects/workflows, it does not.
Im not here to argue either, I asked a specific question and your answer didn’t really address any of it, i was just pointing that out.
I too find it frustrating but it seems for different reasons.
I really really dislike the way it’s being sold as a solution for things it’s in no way a solution for.
They do certain things fine, good even, but blanket statements like “their code is great” without appropriate qualifiers is contributing to the validation of these bullshit sales-oriented claims of task competency.
1: agreed
2: then I think you are missing the fundamental limitations of the current approaches, but we can agree to disagree on this.
3: see 2
I agree with jobs on the chopping block, though i think that’s in large part due to poor due diligence and planing by management, but that’s nothing new, the same thing has and is still happening with offshoring (throwing more people at a problem generally won’t solve design and governance issues).
I also think the current systems aren’t capable of being a viable replacement for anything above junior level stuff, if that ( not that that doesn’t present it’s own problems )
I think the difference in opinion comes from my belief that LLM’s and the current tooling around them aren’t fundamentally capable of replacing existing resources, not that they just don’t have the power yet.
Putting increasing large compute in a calculator won’t magically make it a spreadsheet application.
To your point then: what are your thoughts on this project? https://github.com/anthropics/claudes-c-compiler I’m not particularly interested in this use case right now but it seems more in line with what you’re interested in.
I think it shows a lot of limitations but also a lot of potential. I don’t personally think the AI needs to get the code perfect on the first go – it has to be compared to humans and we definitely don’t do that.
I really really dislike the way it’s being sold as a solution for things it’s in no way a solution for.
Yes, of course. I think it’s important to look passed the blowhards and think about what it’s actually doing: that is the perspective I’m trying to talk about this from.
“Small tasks” are the key there. It can write a small script or spin up application boilerplate very well, but it really struggles with long-term maintenance and new features on a complex application.
I spend about as much time telling the AI “no, not like that” as it supposedly saves me from not having to type the code manually.
It does have some value, but I’d put it around a 10% boost – not the 500% boost that senior leadership insists it can do.
Right, but I think we’re kidding ourselves if we don’t think it’s going to get better. I have no doubt it will be able to magically generate a new Linux kernel.
You say that, but you have to remember that LLMs produce the average output of their training materials. Not the best, but the average. And there’s a lot of code out there that is simple. Only the outliers have the magic combination of conciseness AND quality AND complexity.
LLMs also have no understanding of context outside the immediate. Satire is completely opaque to them. Sarcasm is lost on them, by and large. And they have no way to differentiate between good and bad output. Or good and bad input, for that matter. Joke pseudocode is just as valid in their training corpus as dire warnings about insecure code.
I read a comment once that still rings true - “Hallucinations” are a misnomer. Everything an LLM puts out is a hallucination; it’s just that a lot of the time, it happens to be accurate. Eliminating that last percentage of inaccurate hallucinations is going to be nearly impossible.
I’d push back on your point here with a few things:
The primary one being: the code doesn’t need to be perfect or even above average – average is perfectly fine. The idea here is comparing the AI to a human, not to perfection. I see this constantly with AI and I find it a bit disingenuous.
I do truly believe what I said above will be possible within my career (I’m in my mid 30s), but it’s not really what I’m worried about right now. I think the current code I see being generated is generally “good enough”. I’m not comparing it to perfect: I’m comparing it to people.
I read a comment once that still rings true - “Hallucinations” are a misnomer. Everything an LLM puts out is a hallucination; it’s just that a lot of the time, it happens to be accurate. Eliminating that last percentage of inaccurate hallucinations is going to be nearly impossible.
I don’t see any reason you have to remove all hallucinations to get a good tool for autonomous development: humans aren’t perfect either. We compensate for that with processes and checking each others work, but plenty still falls through the cracks.
LLMs also have no understanding of context outside the immediate. Satire is completely opaque to them. Sarcasm is lost on them, by and large. And they have no way to differentiate between good and bad output. Or good and bad input, for that matter. Joke pseudocode is just as valid in their training corpus as dire warnings about insecure code.
Have you seen output in which satirical code is actually included? I’m well aware of things like https://www.anthropic.com/research/small-samples-poison and the potential here. And do you not believe that either (a) these types of trivial issues would be caught by a person whose job was just to audit output or even (b) this type of issue could be caught by specially trained domain limited AIs designed to check output?
I think the current code I see being generated is generally “good enough”. I’m not comparing it to perfect: I’m comparing it to people.
If this were true, then open source projects would have much less of an issue with pull requests from sloperators.
Have you seen output in which satirical code is actually included?
I wouldn’t expect to see it. Satirical code requires more thought than an LLM is capable of putting into its writing - you need to understand what is expected of whoever you’re satirizing, and then you have to take that expectation and take it a step further into the absurd. Without having that context of something that is specifically being satirized, what you have instead is just incorrect code. And again, the LLM is incapable of valuing proper code over intentionally wrong code, so it’s going to poison the database to some extent.
And LLMs don’t drop big chunks of copy-pasted code from Stack Exchange like an intern would. They work one token at a time. (Which is why trying to get them to understand that quotations need to be all in one piece is a futile endeavor.)
Besides, ‘satirical code’ is just one example of the many things that can poison the training. I couldn’t even begin to enumerate all the things that could mess with it, and honestly I’m surprised that LLMs do as well as they do considering they likely have all sorts of cross-language screwball connections (which may be why it has such a tendency to make up libraries; it doesn’t necessarily understand that a common PHP library doesn’t exist in Java).
do you not believe that either (a) these types of trivial issues would be caught by a person whose job was just to audit output or even (b) this type of issue could be caught by specially trained domain limited AIs designed to check output?
These issues could be caught by someone whose job it is to audit code, sure. The problem is that sloperators often don’t audit their own stuff well enough. They leave it to the open source repo’s admins. When pull requests from overeager noobs were infrequent, it wasn’t the problem; they could gently correct them, the repo would stay high-quality, the noob would learn, everyone is fine. But now, sloperators are dumping low-quality pull requests on the repos faster than the admins can sort through them - because it now takes less time to produce slop code than it takes to determine whether or not the slop is worth including. The admins are swamped, because they can’t sort the wheat from the chaff fast enough.
A domain-limited AI designed to check output would be useful - if it could be trusted. Open-source project admins are some of the best coders out there, and they vastly outstrip the capabilities of LLMs. You’re suggesting that we replace THEM with an agent. They are in that position because they’re right far more often than they’re wrong when it comes to understanding the code as it exists, and how incoming code would impact it - or at least they’re right often enough to keep the project alive. LLMs will be worse at that job, I guarantee it. They’d be fast, but they’d be wrong too often. This is the primary issue with LLM agents.
