Suppose your manager asks you next week to demonstrate that the AI coding tools your company signed up for are worth the subscription cost. Would you measure lines of code generated, or tickets closed? Or would you send out a survey asking whether developers feel more productive? Each of those approaches is flawed in a different way; the sections below explain why.
↫ Greg Wilson
Every single study that claims to prove “AI” has a positive effect on productivity falls into one or more of these categories.
Again, nobody has ever used “AI” to produce anything of value.
“Every single study that claims to prove “AI” has a positive effect on productivity falls into one or more of these categories.”
I do **not** accept the implicit premise in that statement, namely that “productivity” is a good measure to decide implementation of the technology on.
FlyingBoats,
They make two assumptions devoid of facts.
1 – The companies that promote AI usage have no actual way to measure its success
2 – The developers are somehow coerced into doing something they do not like.
I can’t speak to the first one. I’m not a C-suite executive.
However, for (2), my anecdotal experience is… every developer I talked to love those tools. Friends from Google, Amazon, and yes even Meta all mentioned more positives than negatives.
After all, the job of a developer is automating away his duties. We write scripts, tools, and entire frameworks to automate stuff.
If someone gives us a “generic automator, even if somewhat unreliable”, it is the best thing after sliced bread. Why should I spend time refactoring a large codebase if I can just have my agent orchestrate it?
“But aren’t you afraid for your job?”
That is the trick, we developers have always been afraid. This is nothing new.
I distinctly remember listening to a talk from a high level Microsoft VP. He was mentioning “I’m grateful everyday my badge is working”.
So AI cannot bring any more fears.
It might be worth talking to developers who don’t get paid by peddlers of snake oil.
Oh hey, I’m one of those! Here, let me help: AI sucks.
Aankhen,
I don’t want to dismiss out of hand. But which tools have you tried to use? And what was your workflow?
LLM is just a tool. It all depends on how you hold it.
I am programmer and I’ve been coding for more than 40 years now (30 of them in the gaming industry). For me AI is a tremendous productivity boost. Things like a one-shot tool to help with me with a specific task, chores and technical debt sitting for years in the backlog, grepping server logs to troubleshoot an incident – the AI has been a huge time saver for me and I refuse to go back to the dark ages and dig tranches with a shovel 🙂
Having said that, you really have to know what you re doing. Vibe coding in an area you know nothing about is probably going to end up a mess, especially if it’s something more ambitious. I also agree that AI slop is a big problem, and that AI should be used reposnibly. However, I think that a statement like “nobody has ever used “AI” to produce anything of value”. is a bit too much 🙂
Early doors I leaned more towards AI skeptic too. I would give an AI chat a problem in my domain, and watch it hallucinate something that looked plausible but that 100% didn’t work or didn’t exist.
However, more recently, and much more to do with coding, I have had AI take existing code and suggest and implement coding improvements that have increased the performance of the code by an order of magnitude. My programming skills are pretty average, and AI definitely makes me better.
If your job is coding, and in particular if you work on pretty esoteric problems that have no reference, then AI may not be as much help to you. But if the programming patterns you are likely to be implementing have been implemented elsewhere that AI could have learned from, then AI can 100% improve your productivity while reducing bugs.
And if there are not enough reviewers to keep up with this deluge of code and increased programmer productivity, then maybe companies need to change the balance between those who primarily code and those who do the reviews. Or hire more reviewers and do things faster all round.
mkone,
This is true at individual level. Most of the criticism to “slop” goes away when developers actually read and most importantly understand what the LLM has generated before sending it for code review.
“Again, nobody has ever used “AI” to produce anything of value.”
that’s a bit to much of a broad statement.
“LLMs make code generation faster, and that half is easy to measure. The other half is harder: time spent reviewing LLM-generated code for correctness, time lost debugging confidently wrong suggestions, security vulnerabilities introduced by plausible-looking but insecure code, and technical debt from suggestions that solved the immediate problem while ignoring the surrounding design” – we’ve never had statistics of this, not or humans and not for LLMs.
That’s the issue, and my guess is: they won’t come either.
Here is an article making the case for AI more eloquently than I could – https://fly.io/blog/youre-all-nuts/
mkone,
While it makes logical sense to be so critical of LLMs that threaten other people’s work, the fact that it can be threatening to other people’s work kind of undermines the sentiment that it’s not useful. If it truly weren’t useful, there wouldn’t be a real threat. It is useful, and that’s the problem for anyone who doesn’t want AI technology to succeed. Someone might say it provides nothing of value, but that really doesn’t land right. A strong case against LLMs & AI has to contend with the fact that these are useful for people choosing to use these tools.
AI could end up hurting more people than it helps, I don’t really know. However I do know that telling people their AI tools aren’t useful is not a winning argument.
The criticism may well be sincere, but I agree that it is a strategic blunder. Real people, and plenty of them, ARE finding value in the output from these LLMs. Claiming they are of no use whatsoever only diminishes the speaker’s credibility at this point. While not all anti-LLM folk take this hard stance, it is rapidly becoming a stereotype and a strawman for LLM advocates to foist. Speaking as someone opposed to the technology for other reasons, I must admit that my usual reaction to the LLM opposition I see is: please… don’t help.
Baylan Tano,
They sold it wrong.
They tried to reach out to CTOs “here is a virtual programmer than can replace your meat puppets” letting them think they can fire all the workforce
But that of course caused backlash
When they could easily say: “Here is a productivity tool that will help make your existing workforce much more effective”
Which is not only better PR, but actually closer to truth.
(Is there ANY organization that removed all software engineers, and replaced them with vibe coding?)
sukru,
Yeah. It’s poor PR to frame investment in AI as a way to replace labor, although executives may be thinking exactly that.
I’d say the truth is fuzzy because of several factors.
Mass layoffs have become common and the company doesn’t need to say “we’ve done this because of AI”. “Normal” attrition can also perform this function without putting a label on it: employees retiring, changing jobs, restructuring, etc. Even if AI only fills the positions of employees leaving naturally, I’d say that still counts as replacement because AI takes a job that would have gone to a new hire.
The go to argument in favor of automation despite inevitable human redundancies is the potential for job creation in new areas. This is a valid point, however it can be debatable whether the job creation is sufficient to make up for the job losses. When it comes to tech in particular, job losses may be disproportionately suffered by many people in the lower classes while job gains may be disproportionately go to fewer people in the upper classes. Tech giants boast about creating thousands of jobs, but they typically gloss over many thousands, possibly millions of jobs being displaced macro-economically. Of course companies don’t care about external factors, but it would be socially irresponsible not to consider those effects because they have real consequences for a healthy society.
Alfman,
Yes, my friend was just laid off from Meta this week. That famous “10%” cull. But AI is a scapegoat. It is plain old mismanagement.
They over hired during the pandemic, and don’t know how to use all these engineers.
(Many teams were moved to AI roles even if they survived the cull. Unfortunately, I don’t think this was a good choice either)
At the end of the day, it is a net win for humanity. But the displacement affects individuals, while the wins are spread to the community.
Fortunately we have some of the best social safety nets in the entire human history (even if it does not “feel” like it). We can have large groups of people not work for years and still be able to feed themselves (which was never the cast until mid last century).
But the displacement is real. If someone held a high paying middle management job that no longer is there, offering them to “start over” at Starbucks… is demoralizing at best.
sukru,
Multiple reasons can be simultaneously true. These are easy to hand-wave, yet hard to actually prove. At the end of the day after all the job shuffling is done, if you have AI doing tasks that would have been done by humans prior, that is replacement by definition.
This is nothing new, automation has been replacing human jobs for generations. I have a hard time concluding this is either good or bad because I think it depends on circumstances.
That’s very presumptuous though. Automation concentrates wealth and power. This is problematic on it’s own but thanks to the nearly unlimited scalability of software, the impacts of automation get multiplied by orders of magnitude. A tech monopoly leaning on automation might replace millions of jobs throughout the country. Even though tech giant employees will have higher wages than the jobs replaced, the total aggregate wages for everyone will go down. This is a financial necessity for the tech giant’s services to be more competitive than the jobs it is replacing. And then there’s the issue of gentrification, which has become a huge problem in every city where these tech giant’s reside.
Granted I’m only focusing on the negatives here, but I’m doing it to prove a point: it’s not that straightforward or clear to say it is only a win for the community… there are systemic cons for the community as well that we shouldn’t be ignoring.
If you’re comparing to mid-evil times then sure, but compared to our parents and grandparents we are loosing safety nets. Meanwhile the (US) government keeps reducing social programs in order to cut corporate taxes. The trillions they are making comes at the cost of the rest of society. This is so regressive to society and pisses me off so much, but it’s tangential to the topic here so I’ll try to drop it.
That’s true, although I suspect the most impacted victims are going to be the entry level applicants who have never had a good job.
Alfman,
No, I specifically mean mid last century.
Yes, because, they are part of the problem. Sorry, they are the cause of the problem.
A certain generation lavished themselves with “unfunded liabilities”. In other words generous benefits at others’ expense.
Take the infamous Medicare Part-D. From inception they knew they did not pay for it. It was always “future generations will figure out how to pay for our unlimited pharmacy bills”
And today Medicare is about to go bankrupt because of it. Not because, we, the current working generation is not paying enough, but because the previous generations did not, and using our money that should be accumulated for our retirement to pay off theirs.
In a just world, all “unpaid pensions” (which they call underfunded) would be retroactively fixes, and any extra benefits they collected would be clawed back.
But they also happen to be the most powerful voting group that trancends both parties, and control the media and large companies.
Our best bet is somehow funding our retirement privately, while still having 15% of or wages garnished to pay off theirs.
(No, our kids will never pay ours, that would actually require 30%+ tax rates, if you do the math honestly)
sukru,
Last century contained both the great depression as well as a golden era for the middle class – it’s a huge economic range. Social benefits are on a downwards trajectory now. It’s affecting my family personally with employer health care benefits getting cut along with increasing fees.. It’s rare for retirement plans to be fully employer funded anymore, etc. The US is actually filthy rich, but the problem is that the distribution keeps getting more and more skewed to the very top. Practically every economic study shows the same thing: the ultra wealthy are getting hugely disproportionate shares of the pie at the expense of everyone else. It’s because of them we’re loosing social benefits and national debt that is out of control, even wages are nowhere close to keeping up with worker productivity.
Just look at historical tax rates on corporations and it’s no wonder government programs that used to work are becoming insolvent.
https://pennycalc.com/tax-rates/corporate/
Alas, they’ve got an army of lobbyists corrupting the government to make sure corporate interests come first. So it’s no surprise that things are getting worse. The question becomes how much money can elites steal from society before it all collapses. I’m sure there’s a science to it.
Really both are true, in the US we pay way too much compared to the rest of the developed world….medicaid doesn’t make money, a lot of that money actually ends up in the pockets of private companies where only a fraction ends up paying for patient healthcare and they keep the rest as profit. I agree we need to fix that. But it’s also true that we’ve lost trillions in corporate taxes that 1) balloons the national debt and 2) has created massive shortfalls for social programs.
If that’s the way you want to go, fine, but then you should also be for retroactively clawing back trillions of dollars that corporations effectively stole through government corruption. The corporations didn’t earn that money legitimately.
Alfman,
I really wish there was a magic solution, but retirement systems are all math. And math does not care what we want. It has pure logic.
You can buy a annuity today on the open market. It is essentially what Social Security is… or at least claims to be.
The highest “contribution” is roughly $900 (+$900 from your employer)
That will buy you a $4,500 fixed payment if you retire at 65 and work for 20 years
Social security offers $4,000 per month for working 40 years
Do the math…
(And not even going into a 401k, which will pay you even more, but at a higher risk)
That is undoubtably true.
Yes, the healthcare system is a mess. When I interviewed with them back in the day Microsoft offered me fully paid, zero contribution zero co-pay health plan. Today it is basically illegal for them to offer that anymore.
But this is entirely incorrect. Cannot be further from truth.
Because they neither administer our social programs (the government does), nor print money for crony projects (again the government does. and only the government can)
Yes, Medicare Part-D basically ensured that.
Neither of them are true.
But let’s focus on social programs.
Social Security
Is an excellent idea that came out of great depression. It is a government mandate for people to buy retirement plans.
But is is also a terrible execution, since the government also chose to be the sole executor of that plan. Do not get me wrong, but, they are fully in charge from start to finish.
And Social Security was specifically designed to be a closed program. It is fully funded with FICA taxes, and it is illegal for them to receive money from general budget. In other words, if there is a shortfall it is entirely on them.
Other countries, like Nordic social programs have seen them, and have fully or partially privatized their pensions to avoid the same pitfall.
Since government forcing buying a plan might be good
Them administering it is clearly an abject failure.
I really wish our politicians stop giving lip service to “Swedish model” but actually take up on it. For example, if your generation lives longer, your monthly payout mathematically drops unless you work longer. If the system’s liabilities exceed its assets, the math takes over automatically.
They don’t burden future generations for the faulty math of previous ones. They should have separate “accounts”
sukru,
You obviously don’t want it to be true but it is….follow the money. They are guilty on both fronts, not only did they siphon large sums of money away, they’re also the ones who lobbied for the mess of government policies that we have. People who hate the government always want to pretend it isn’t corporations that got us here, but it is.
I’m with you on that, we need to make the math work. However it’s still true that an enormous reason the math doesn’t work is because corporations have been exploiting worker productivity gains for their own financial gains and not compensating workers. You can blame older generations for what happened in the past, but that doesn’t mean we should ignore or excuse corporations who are guilty today! If workers had been fairly compensated and corporations paid taxes at historic rates, the budget would be balanced and we wouldn’t be living in an era of shortfalls. By and large corporations have set up a system whereby they get to keep the king’s ransom and the rest of society is left paying for it.
Alfman,
Don’t get me wrong, I don’t like corporatatecracy (is that even a term)? but one issue at a time.
Specifically for systems government themselves fully manage, we don’t need to look for scapegoats outside.
I don’t disagree, I have only been getting meager pay raises, some years below inflation (which is essentially a pay cut). But the entire industry moved this way post pandemic. Just 5 years ago, things were extremely well.
Corporations are actually paying more taxes than before, especially compared their brethren on the “other side of the pond”.
Giants like Mercedes pay zero taxes many years even though they don’t have the excuse of growth (at least Amazon for example started paying after they stopped growing). And don’t get me started on non profits like IKEA, which is supposedly some interior design research foundation or something.
That being said, the ratio of tax burden is now less on corporations directly but more on workers. Money is fungible though, but that is a very long rabbit hole.
(Ultimately only and solely workers can pay taxes. Anything else is an illusion, including corporations, which are legal fictions, again a long rabbit hole).
sukru,
I’m not using corporations as a scapegoat, it’s just that I don’t think you realize why blaming the government alone doesn’t make sense given how much is privatized and how much weight corporations carry in passing legislation.
https://www.nber.org/reporter/2025number4/economics-medicaid-new-evidence-privatization-competition-and-program-design
So if your going to fault government, I don’t take issue with that. But I do take issue with pretending that corporations don’t play a significant part. When it comes to US healthcare, government and corporations are more of a Gordian knot. It’s illogical not to blame corporations too.
The data I provided was for US corporations, the top tax rate is about half what it was 50 years ago, although many large corporations don’t pay that because they’ve found ways to avoid it leading to corporations paying MUCH less than before into the tax system.
https://en.macromicro.me/series/7202/us-effective-corporate-tax-rate
?
If I’m understanding you correctly, you don’t believe corporations should exist? If so that’s an interesting take. IMHO governments and corporations should only exist under our authority and not above us. If it ends up being the case that we serve them rather than the other way around, I think it’s fair to question why we should have them at all.
Alfman,
No, I mean corporations only exist on paper, and our imagination. They are abstractions of human labor and services. What was the saying “corporations are people”. But more like “corporations are groups of people”
Being abstractions and second level numbers, they don’t mean much without the larger picture.
Remember, any tax on corporations is by definition double taxation.
But to topic on hand, they produce tax value directly (surplus) and indirectly (via wages paid, which lead to income and FICA taxes, and dividents which lead to capital gains taxes)
We really want the first one to be as small as possible, there should be little surplus in the virtual account that is not used to investment or wages.
That is why their so called “effective” tax rate goes down. They are actually using the money, and the “corporate tax” which is a secondary taxation drops.
And except a few sectors, like finance, which you might guess I don’t have fondness of, the “cost of labor” or in other words the share of profits by workers have been steadily increasing. (Why do you think all C-levels were immediately sold on AI, even though it eventually will help workers instead?)
Again financialization of the economy” is an illness that hides all the good parts, since they are essentially parasites sucking out the blood of the system.
sukru,
I don’t really care if you want to look at corporation as paper entities, corporations are social constructs, obviously. But that doesn’t invalidate criticism of them.
Even in the context of paper entities, the original statement “Ultimately only and solely workers can pay taxes. Anything else is an illusion, including corporations, which are legal fictions, again a long rabbit hole” doesn’t really hold up. Corporations DO get income and they DO pay taxes. Even social constructs impact real people’s lives. They can divert wealth from real wages, social benefits and even paying off the national debt, all of these have significant macroeconomic impacts. And on top of this, not only isn’t the wealth being fairly distributed working class incomes/benefits, but it’s actually being spent to lobby against workers and create more corruption. So even though they are “paper entities”, the impact on people is still real.
Yes money that flows through corporations is taxed as income to the corporation and then income to the shareholders. Although thanks to massive tax cuts for corporations, tax avoidance and owners paying capital gains taxes rather than normal income, the amount they pay is surprisingly low. Billionaire executives can pay less than their own employees. I know you are dead set on not blaming corporations, but the taxes that corporations haven’t paying (from back when these social programs were originally created) directly impacts the country’s ability to provide those benefits without going into debt. The pro-corporate crowd always want to blame somebody else, but dropping the corporate taxes is catastrophically maligned with a country that’s able to provide social benefits.
To use your own point from before: do the math.
Wages have notoriously not kept up with GDP.
https://fred.stlouisfed.org/graph/?g=2Xa
I agree with you that there are a lot of parasites. Where we disagree seems to be in acknowledging that corporations themselves are high on that list.
Alfman,
My friend, I knew this was a rabbit hole. I even wrote it was too deep. But it was probably a mistake to enter it. Now we cannot come out.
Yes, that is true. But they are just a facade for the people behind them.
Those are only the responsibilities and frankly the ability of national governments. Everything else is an illusion.
So, two facts needs to be understood.
1 – There is no real “unbalanced budget”, “surplus” or “deficit”
2 – We are at the highest tax levels in this country except the World War 2. Both pre-and post pandemic.
These two topics are related. And is an important concept.
But you have to “pierce the veil”
“There is no unbalanced budget,”
… because all spending is funded by direct or indirect taxes. …all the time.
(Those indirect taxes include inflation (which is a tax by another name), and borrowing among other things)
So, the actual tax rate is the total government spending.
And it currently sits 37.92%
(And this does not include state level spending, or indirect ones like medical mandates and contractors)
How this is distibuted, personal income, investment, corporate, inflation, … does not matter. Those are a game of musical chairs.
The question one should ask is:
“If we are spending the highest amounts since the birth of this nation, and comparable to welfare states in Europe…. why is our pensions are failing? why are our schools producing the worst outcomes for our kids? why is our infrastructure crumbling?…”
Without answering those everything else is secondary, and illusion.
This is an extremely important, but long tangent. Do you really want to go in there?
(We’ll probably have to do another 10+ posts.)
I’m not saying otherwise. What I said is the ratio of cost of labor in corporate finances.
But again how can it be increasing?
Because they are hiring more people.
A corporation is an artificial construct to organize human labor and capital. They are on better or worse than the individual making them up.
(But, there are “emergent” properties of course. For instance a group of highly bright individual can act in stupid ways collectively.)
If they are able to “buy” politicians by lobbying, shouldn’t we ask why politicians were for sale in the first place?
The scary thing is: when it comes to art, lots of people have no eye for art and if the genAI creates something good enough, they’ll take it.
With coding it’s a bit different, but I would say: it’s good at a bunch of things, but it doesn’t fully overlap with what humans can do (it might overlap much more with junior developers, which means they need to be trained differently).
Lennie,
I’d say this is ironic. Since clearly the AI is unable to do proper art. At least in its current form. But high end models are perfectly capable of writing production quality code.
Yet, I see most of the backlash from artists, and very few from programmers. Most programmers I have seen are actually happy.
… except junior programmers. Yes.
That is a conundrum, since to have senior engineers, the pipeline requires them to be junior at first. And now there is much more to learn.
(A good programmer has to know the “whole” stack. The entirety of it. Not just HTML + JavaScript. Not just Java + C++, Not just C + Assembly, but things like computer engineering, CPU design, and low level logic and its electrical representation. Now… they also need to understand machine learning, linear algebra, statistics, and AI. Ouch…)
mkone,
Here is a misleading excerpt from the article:
If it is boilerplate code, then build a template or a framework that’ll reduce tediousness. You even have the opportunity to optimize it, run it through code reviews and reuse it as quality code. There is really no need to invent the wheel by copy pasting mediocre code over and over again, whether it is from some random site on the Internet or a LLM.
Sure, just because he says so. The author claims to be an expert due to his own seniority (also referring to literature on C++ design patterns as “ill advised”), does not instill confidence on his opinion piece.
Alfman,
The issue isn’t that it is not useful, but that is being advertised as capable of replacing actual skilled experts. The output from LLMs are often suspect and is being marketed as equivalent to any quality work produced by a human being. None of these claims have held up. Any rational individual that understands the limitation of the tools would understand its place and not lie about its usefulness. But that has not been the case now, has it? Furthermore, LLMs have been trained on works owned by others with no recognition or compensation. Not only it is undermining the value of human labor, it is also blatant theft.
Proponents of LLMs, that liken ‘AI’ advancement to the industrial revolution, would sell us the benefit of how it has made goods affordable and improved output. Little is often discussed about the great human costs in terms of the quality of life of workers/craftsmen, the exploitative nature and abuse of the industrial elites, lowering the quality of life for the average family. Many battles have been fought and won since then, for the rights of workers and ordinary folk to attain decent living standards. With ‘AI’ the elites are at it again. This time however, with attempts of diminishing a workers value with none of the hyped up benefits. We are living in an era where workers with slave wages in China, is something the elites want us to be enviable about.
adkilla,
Most of us here seem to be saying that LLMs are mostly coming for lower level work, at least for the time being. Work that would normally have been delegated to entry level employees can now be delegated to an LLM. I know people hate to hear it, but they are getting better quite quickly.
Technically it would be “copyright infringement” and not “theft”. Traditionally copyrights wouldn’t cover transformative uses, but we’ll ignore that here. So if you train an LLM on GPL code and use it to generate proprietary code, you can make the case that is infringing – it makes sense. However if you train an LLM on GPL code and use it to generate code that’s merged into a GPL project, there is no infringement since this use is allowed by the GPL!. In fact a new anti-LLM restriction would actually be in violation of the GPL. I understand this rubs many the wrong way, but it’s true and the GPL is quite clearly written on this point.
I find that the bigger problem is when an LLM is built with incompatible FOSS licenses. IMHO the FOSS community should promote / endorse FOSS friendly LLMs that are designed to explicitly respect source licenses. This way we’d be able to use LLMs and know that the code they produce explicitly complies with the sources used to train it.
To be fair though I have been quite vocal on the impact AI can have on workers, haven’t I? I agree it’s very disruptive. And while advocates of automation would generally say “just get a better job”, that presumes there are enough new better jobs attainable to offset the lost jobs, but this isn’t a given. It’s completely possible that the workers loosing their jobs to automation end up being forced to take worse jobs.
One of the largest growing sectors today is eldercare work. Plenty of openings to absorb unemployed workers, but also very low pay.
https://www.nbcnews.com/business/economy/eldercare-workers-job-growth-pressure-rcna258820
I believe that AI job displacement could make these kinds of job migrations become far more common over the coming years.
Alfman,
This is true. But I’m afraid it is worse than that.
A senior engineer was a junior back in his day.
Without a healthy pipeline that trains juniors, we won’t have enough seniors to “babysit” AI.
I’m not sure how to fix this. Our kids are going to be really affected by it. (Adding to the usual problems like screen time and social media)
“nobody has ever used “AI” to produce anything of value.”
Probably one of the dumbest things I’ve ever seen posted here.
A bunch of Linux GUI backend libraries are now getting AI contributions that actually fix things. Yes you have to be careful with pull requests to make sure that features are properly gated/separated so they can be tracked and merged, but we’re seeing things get fixed that would have taken years taking days now. The best use case I’ve seen is getting AI to analyse an X11 feature and re-implement it in Wayland.
there has been a multitude of studies that shows that people gets LESS productive when using AI, not more.
My biggest fear is the future. If the new generation of IT/coders are all growing up straight into AI, then they may not learn thoroughly the basics needed to have the mindset to be in the industry AND may end up not being able to troubleshoot the edge cases.
Case in point, AI hardly ever helps me, because when I am stuck, it’s because I am doing something out of the beaten path and sometimes does me more harm than good (as I get distracted while I figure out that I am dealing with something more unique).
It helps a lot with boilerplate but I have to watch out to not let myself forget the basics.
If by “AI” (there is no such thing) you mean “LLMs”, then I personally just produced two things of value using them. One thing at work, and one thing at home. Both things also took some of my own effort, and a non-programmer wouldn’t have been able to produce these two things, even with an LLM. Am I a monster for having used the LLMs because of how much compute power sits behind them and how much waste heat is generated by it? Possibly. Are these systems still useless? No, not anymore. Are people using them in absolutely horrible ways, hurtful to society as a whole? Definitely. Should we not be talking about how to regulate the COMPANIES training and hosting LLMs? And should we not be pouring money (in spades!) into ways of generating electricity in environmentally friendly ways?
If your job is writing very derivative code, you love LLMs. That’s it. You also have a VERY precarious job.
CaptainN-,
The problem is that very little code our industry creates is truly original or innovative. When I entered CS, I loved the idea of breaking ground on cutting edge projects. But upon joining the workforce I was sorely disappointed with just how much work is unoriginal and derivative in nature. Sometimes I challenge myself to try new ideas within the constraints of what companies are paying me to do, but unfortunately in practice one of the biggest constraints has been the budget itself, which disincentivizes experimentation and originality and strongly favors using existing off the shelf products and copy/paste solutions.
It’s not my preference at all, but it’s the reason off the shelf packages (ie wordpress) thrive: cost savings matters more to the average business than originality. As much as I enjoy working on technically challenging cutting edge work, there’s not nearly as much of that work to be had and I count myself lucky whenever I can land it. The vast majority of work doesn’t fit the original/innovative mold.
Getting back to the point, even if we assume you are completely right about what LLMs are good at, I still think there’s reason to be concerned that LLMs could take over a significant number of jobs in the tech industry and the question is always “where will the workers who formally did those jobs go next”? Are there going to be enough new opportunities to counteract displacements? At the moment there isn’t a clear answer.