not at all, it’s like all the harm social media algorhythms have done but on steroids. I love the sign your parents had, and I love your whole post.
hijack question for the group: how is “being a jerk” different from this type of snottiness?
not at all, it’s like all the harm social media algorhythms have done but on steroids. I love the sign your parents had, and I love your whole post.
hijack question for the group: how is “being a jerk” different from this type of snottiness?
Moderating:
If you know it is a hijack and you’re in Great Debates or Politics & Elections please don’t post it. As to the post in question, just flag it if you think it is a problem, like someone flagged yours.
Moderating:
Hard not to take this as insulting another poster. Please do not insult other posters outside of the pit.
Further, both you and @Dr.Drake have gone off-topic with these comments.
Please don’t continue the sidebar about Musk’s wealth.
But AIs are getting better all the time, and quickly.
I’m not a programmer, strictly speaking, but I need to create JavaScript scripts occasionally as part of my job. A year ago, maybe 6 months ago, the scripts AI produced were garbage. Now, I just recently built a data merge solution using an app I wasn’t very familiar with, that accepted its own proprietary version of JS scripts to automate things. I needed to input hundreds of names into a template, organize the names and a logo in specific ways, and export the results as images. If I had had to figure it out myself it might have taken me weeks to get the whole process just right. Now, granted, I’m not good at coding, but it would have probably taken a talented coder several work days, at least. Using Claude 4.7 Opus, I had a working process in less than an hour. Sure, it didn’t work right away, it took a few iterations, but I just told the bot what was wrong and it identified and fixed the issue in seconds. The experience was both exhilarating and terrifying.
I’m seeing the same thing with image creation and videos. Sure there are still glitches, and a lingering hint of ‘uncanny valley’, but I bet in another 6 months I bet we won’t be able to tell the difference between reality and AI..
Same with design, writing, etc., etc. I think we like to pounce on the instances when AI gets it wrong and say “see? We humans are still better!” But it seems kind of like whistling past the graveyard, at this point.
I guess I don’t see how using AI to write scripts is different from my experience on a project about ten years ago. The project involved taking in these elaborate Excel workbooks containing derivatives trading data from something like 15 different banks, running it through this model, then using the output to generate 15 Excel workbooks of reports, unique to each bank. Most of the process was very manual and we had to do this twice a week an discuss the results on a big conference call with all the banks (pre Teams/Zoom mind you). Which meant we were in this constant cycle of being up until midnight every night gathering data and crunching numbers to produce what was ultimately very error-prone reports.
Well after about a week of this bullshit, another guy (who knew Python) and I (who is really good with SQL and Excel) decided to just automate as much of this shit as possible. I used Google to find some scripts which I could modify for some of our calculations. Basically getting a process that took half a dozen people two days to create shit reports down to about an hour for one or two of us to generate all the reports error-free.
That allowed us to free up most of the team to work on more interesting stuff as well as run different scenarios with the data. So the client could be like “what if we want to tweak Gamma a bit here” and we can be like, sure, let me rerun the numbers and get back to you in a few hours (instead of next week, or next Neveruary)
I look like a freakin hero because our contract was fixed so I just increased our margins massively.
I would say more people need to read Who Moved My Cheese and be more like the Mice scurrying through the maze looking for food instead of the two Littlepeople starving to death waiting to see what happens next. But the reality is most people are conditioned to be Hem and Haw, content to do their assigned tasks and complain once their cheese is moved on them.
The reality is that AI is here. So unless you are close to retirement, it’s probably a good idea to figure out what it can and can’t do and how to best use it to advance your career (or pivot careers if you think it’s a threat).
People have been doing stuff like this for decades. I’ve done my share. I’ve seen it go from coding stuff to finding a pre-existing package to do what I want. When I was a faculty member for a term I taught Data Structures, which included sorting algorithms. I don’t know if they even teach this anymore, but who has written a sorting routine in decades?
The difference with AI is that you describe what you want done, not the details of how to do it, so you don’t need someone who knows Python or even Excel. We might not be quite at the point where anyone can do it, but we’re close.
I don’t think you’re specifically referring to me here (at least, I hope not), but I have adapted to changing technologies over the past 40-some years in my work, as I am now adapting to using AI.
But I maintain that the rise of AI is not simply just another new technology that we will adapt to, and everything will then be fine. I think it has the potential to permanently displace the earning potential and usefulness of many, many people. I don’t worry about this for myself, but for my kids, and their future kids.
Heh, I actually had a job interview recently where I had to write a few different sort routines in the recruiter’s own very limited programming language. I hadn’t done that sort of thing since the 90s. It was easy then, but it hurt my brain this time. Every language having their own sort libraries available had atrophied that skill.
It does make me wonder if I’ll lose basic troubleshooting skills through using AI in my job in the future. Since I don’t even trust the humans that bring their problems to me to get that right, I doubt I’ll trust the AI more than them. But only time will tell if I will start to depend on it like I now seem to depend on the people who write sort libraries.
I don’t believe LLMs have unlimited capacity to improve. It might be like human height, where exponential growth maxes out right around adulthood. I don’t know what they are going to be like in six months or five years. Nobody knows. The current business model looks unsustainable to me, even just in terms of sheer infrastructure.
I use Claude a lot to make Excel spreadsheets for me. Claude does it a hell of a lot better than I can, but it sounds like @msmith537 could do better than Claude. Maybe not faster. But better.
Maybe having the Best Spreadsheet Possible doesn’t matter for my purposes, but I’ll bet there are plenty of purposes where it matters quite a bit.
That’s the thing; pivot to what? For previous new technologies, there’s been sectors destroyed and sectors created, the impact of AI looks like it may cause a near across-the-board reduction in need for human employees, especially in entry level positions.
There’s very few roles with good career progression that aren’t likely to be impacted to some degree, except those with legal barriers to AI doing the work. There’s no obvious areas of growth; OK, maybe building data centres…
Code in libraries do very specific and well defined things. My guess is that your troubleshooting skills would get better with AI, since it will make a whole new set of mistakes.
Most problems with code don’t come from misplaced commas or end statements, but from not understanding the environment or requirements. I guess AI will be great in writing something that is well understood - like a sort - but what about something where the requirements are real tough to get down?
All the stuff I wrote was first time - often first time anywhere stuff - so while AI probably could have done well understood chunks it would never have been able to do all of it.
Oh, I’ve seen that in action. In my previous role I was tasked with vibe coding a log parser to give newbie support techs an inkling on what to look at when examining a particular type of problem. The highly variable and poorly documented data that could appear in the log files was bewildering to it, and it couldn’t establish what is and what wasn’t an important error without me providing a lot of guidance. That’s on top of it making other bewildering changes it wasn’t asked to do and it didn’t bother mentioning it was doing.
But that was six months ago, you kind of had to keep the model on a short leash at the time. It may be better at it now.
True. But also consider that out of maybe a dozen people working on this project, I was the only one to actually think of doing it and there were maybe 3 or 4 people in a company of 40 people who could actually do it.
In face, as I progressed in my career, I’ve found that in a typical company, very few people seem to have actual hard skills (like coding or fixing a copier). Probably a big reason so many people promote “soft skills”.
Unfortunately, I’m finding a lot of people in the workforce seem to have no skills at all.
A few years back we took our kids on a tour of the Lackawanna Coal Mine in Scranton, PA. I imagine myself in heaven explaining to a mine worker whose family was evicted after he was crushed or blown up trying to meet his 16 ton quota that I also had a “hard day” because the power went out while I was working from home and I have to dial in by phone to give a big presentation.
I imagine in 5-10 years the concept of “work” will look very different, perhaps even bizarre and nonsensical to us. And to young people in the workforce, armies of people toiling away grinding through paperwork in cubicles in a big glass building will seem as archaic as a hand writing entries in a ledger with a quill and ink pen.
So you think there may be a physical limit to the capacity of AI to improve, similar to Moore’s Law? Interesting.
Whether or not that’s true, I agree that the current business model seems unsustainable. In addition to the sheer enormity of infrastructure, AI companies are burning through enormous amounts of cash subsidizing the cost of user subscriptions to their AI programs, in order to get and keep users. I wouldn’t be surprised if the current AI bubble pops soon, temporarily tanking the economy, similar to the Web 1.0 tech bubble popping at the end of the 90s.
I kind of think we’re in the ‘incandescent light bulb’ stage of AI development right now. Incandescent lighting was fine for a long time, but a lot of energy was lost to heat. Now we have LED lighting, which is much more efficient; the same amount of light output requiring a fraction of the energy. I think eventually we’ll reach a similar ‘LED light’ stage in AI development where much less energy and enormity of infrastructure is necessary to produce highly advanced AI output. That’s probably the point at which AI begins to improve itself, becomes self-aware, and wipes out humanity ![]()
Everyone’s YMMV, but I feel that constant use of AI has not dulled my brain because it is doing my thinking for me, I feel the opposite - that it’s sharpened how I think. Years of chatting with various AIs have gotten me to see things from viewpoints I would never have thought of before and challenged a lot of my previous assumptions. I feel that AI will ultimately make people better thinkers, not worse. And by freeing up people of some burdens, it lets people think more about other things that matter. It’s like how the calculator makes mathematicians better, not worse, because it now spares them the labor of cheap low calculations and lets them spend more time thinking about the deeper stuff.
This is such a perfect statement. It’s what I used to run into daily waaaaaay back in the day. People could use a tool or write “perfect” code, but misunderstanding the why of what they were doing would foil them every time.
Isn’t AI having these problems, too?
What we System Engineers used to call “building the right thing as well as building the thing right.”
I suspect AI is having these problems if anyone tried to get it to write an entire system. I’m not sure anyone would admit it.
Before I wrote any code for my system I’d type in exactly what I needed to do in pseudo-code into a big gmacs file I had open. More often than not I’d run into a logic problem or a big hole in my understanding, which I needed to fix. And I found that when someone asked me to write something, and I asked for informal specs, most never considered corner cases - and these were very good engineers. Is AI going to ask you what to do if it gets this or that input, or just make up its mind and write something.
So, where is her manager? I was on a performance review committee, and we talked to lots of people about the problems, and one of the biggest was managers not having a clue about what their reports did. Good managers make their reports better - or get rid of them, if that is impossible.
I get that a lot in the Bay Area, and I’m betting that this is from people working for real estate offices without enough to do. In any case they don’t want to buy your condo, they want to list it. I’ve gotten letters with offers from the “we’ll buy your house” people. I’m staying away from them.
Some people may not have enough work to do, whether from being fast or from a slow period. They could ask for more, or they could do some reading and self-improvement. But if the management who are too stupid to measure total output and measure how much time people spend typing instead mouse jiggling is justified.
I had someone in my group who took a nap for an hour most afternoons. He did world class stuff, so I sure as hell wasn’t going to object. When I was working on code, I’d often describe the problem to myself, read the Dope for half an hour, and the code would write itself in my head while I browsed.
That’s the worse example yet. You think maybe these people were in fear that Musk and his DOGE scum were just looking for an excuse to fire them, and the five bullets had nothing to do with what they really did but were an attempt to tell people who had no clue about what they did something that would keep them employed.
There’s also the ridiculous assumption that all bullets are of equal importance. As Dilbert’s boss explained when telling why he wasn’t getting promoted: “Last quarter, you only designed one microchip. While I attended dozens of meetings “.
And the assumption that whatever 20-year old DOGE/Musk toady received the email would have any understanding of the subject.
You all kind of made my my point for me. Corporate America is full of make-work bullshit jobs that do nothing and are overseen by management who does even less. So many jobs probably don’t need to exist now, let alone be replaced by AI.
Of course management is “too stupid”. From what I can tell most of them have no idea what goes on in their group or on their projects at all. They spend most of their time reading abstracted management reports to each other.
In my current job, every month we cobble together a 75 page management report for some senior executives. Or did, now some other group does that. It’s full of status updates and KPIs and whatnot, but as far as I can tell, no one makes any real decisions off any of that information.
I might even argue that is the point. Middle managers don’t exist to provide any sort of leadership or guidance or whatever. They exist to serve as a sort of figurehead and this bullshit paperwork and updates and status reports that have to be generated serves as sort of virtual “panopticon” to make you unsure whether the organization is actually watching you.
The exception is perhaps IT where if a system goes down, it becomes pretty apparent that something is not working.