I am surprised and disheartened to see how "positive" software engineers are around the technology. Above 50%, to me, seems already high, as I feel I am very much in a minority. Based on Nick Suresh's take, I think I need to find another place to work but honestly.. I am wondering where is it :)
BTW: any chance you might get Nick on the pod? I would imagine it would be a great fit.
It's not as black-and-white as journalists are portraying. Talk to rank-and-file software developers off the record you'll find a far more nuanced perspective. We had code-generators before this "AI" boom in the form of IDEs that would spit out templates. But that was predictable and reliable. GenAI code can be unreliable, you don't know where it's unreliable, you have to spend time babysitting the code. And what about ownership & patents? Does the code-generator corporation own the software patents your small startup would normally get for themselves? Are they witnessing the code you're developing in their master "trending" dashboard?
I still consider it a craft and I still consider it something that can be done shittily or well (I guess it is comparable to sewing.. people did it as a craft, then machines made it quicker but crappier). I would hope we will figure that this is actually engineering though... at some point
Hey, I'm a "software architect" - I usually say "developer", but I design entire systems - and I've worked in "data analytics/engineering" too - I was the house "data scientist" at a start-up-ish company for a couple of years - so let me push back on that first set of numbers: I detest what currently passes for AI. It's the most overhyped technology I've ever seen, its social, environmental, and probable eventual economic effects are hideous, and many of the people promoting it are atrocious.
"Those who described AI in the most positive terms were, overwhelmingly, in leadership roles": In other words, Business Idiots, as Ed Zitron calls them. They do little or no real work themselves and can't tell whether other people's (or bots') work is good or bad. So naturally, they love chatbots, the ultimate crutch for the mentally lame and lazy. Besides, they're hoping to fire most of their underlings and collect more money themselves. (I'm caricaturing a bit, but really not very much.)
"A possibility here is that engineers whose work involves generating new code are enthusiastic about new AI automation tools because they make their jobs easier and faster": Another possibility is that many of us who know better have voted with our feet. I quit that "data scientist" job to become a freelance software developer, partly because I didn't enjoy the environment at that company, whose CEO made stupid decisions that led to tanking revenue and firing about a third of my colleagues. And that was before the slopnami hit. I expect the environment at many companies is much worse now. In general, good engineers don't need to put up with that.
I don't expect it will, but this growing resentment of "the undemocratic imposition of commercial technologies into our shared spaces" should lead to overthrowing the Business Idiots and the politico-economic system that exists mainly to serve them at everyone else's expense. Occasionally and idly, I wonder how bad life has to get for most ordinary people before it dawns on them that many things routinely reviled as "unrealistic" or "socialism" are not only practical but essential for preserving any possibility of a decent life for most people. (I'm not holding my breath.)
Ironically Glassdoor, and its new owner Indeed, has gone “all in” on AI and is mandating its usage among all its software engineers. Many of whom are against it, but they can’t push back without risking their job or compensation level..
Moreover, I wonder how much longer we can trust the data coming out of a company like Glassdoor that is rebranding as “AI first”.
Yes yes YES--power and democracy are central to all these fights. I am involved in fighting the law that removes all decision-making power and 70% of the property taxes paid by centers in West Virginia from counties and towns, and a proposed high-voltage power line to run power across most of West Virginia to power data centers in Virginia and Maryland. If they get this, they will use eminent domain--and by the rules, WV ratepayers will have to foot a lot of the cost for this line which delivers much harm and no benefits. What all these fights have in common is a political, business and tech elite empowered to do as they please on our land and we get no more say than the mice get. This aspect doesn't get enough mention when people talk about why they don't want a data center or whatever.
There is another key issue that doesn't get nearly enough mention. That power line would hook up to an aging coal plant in Charleston. The proposed data centers--now up to ten proposed here-- supposedly must supply their own power, rather than hooking to the grid. But it will still increase fracking, require pipelines, and pollute the air. And most of all, the incoming information about climate change (not to mention biodiversity loss, which hardly anyone ever does) is that it's ramping up to dangerous levels faster than expected, will cause extreme crises by 2050, with famines and armies of refugees arriving in places where crops can be grown, where the drought is less severe, where the devastating flood or wildfire didn't reach--and being met by armed people who feel they have to repel the incomers so their own children can be fed. THIS is the world we're facing, and instead of the rational response of working hard shoulder to shoulder to ramp down emissions and build resilience, to protect biodiversity, to protect everyone as we go through the bottleneck--this mad dash for data centers is rapidly INCREASING emissions (which also means we're using up the fossil fuels even faster, so our desperate, hungry, hot children and grandchildren will likely have no advanced energy sources to so much as run a fan or bring a truckload of manure to a field. I don't understand how this existential crisis bearing down on us can be considered a minor afterthought.
Great piece Brian. I suspect another reason bosses are so in favour of AI is that they believe it can help them grow the business and deliver shareholder value by cutting costs. As we all know, if they can't grow shareholder value or EBITDA, they won't be in their jobs for long. In the midst of wider economic turmoil that constrains growth, they desperately need AI to work for them, even if deep down they probably know it won't.
98% negative from claims adjusters is a staggering number. Whoever picks the tool and whoever answers for a hallucinated claim summary are never the same person.
I was asked to review our AI policy at work, I eliminated 80% of it.
It specified all the situations when you had to get AI product reviewed by a peer or superior.
I pointed out it’s the same as when we need to get our own product reviewed. So change it to ‘’treat AI as your own work and follow the review process for the work appropriate to the type of product’.
Really worthwhile piece, Brian. The Glassdoor data and the accounts from workers being held responsible for unreliable AI output are especially valuable. Being told to use the technology and then being held accountable when it gets something wrong is a real problem and deserves much more attention.
Where I’m less convinced is that the evidence points quite so directly to a class-and-power explanation. Power clearly matters — especially who decides to use AI, who receives the benefits, and who bears the consequences when it fails. But I think some of what we’re seeing is also the messy early adoption of a rapidly changing technology. Companies are experimenting, sometimes badly, the tools are failing in real working conditions, and workers are discovering where they actually help and where they don’t.
Those negative experiences matter. In fact, they may be an important part of how AI gets tested, challenged and eventually improved as it moves deeper into the workplace. AI is not going away, so the larger question for me is how these experiences reshape both the technology and the way organizations deploy it.
The split in that chart is not really about the technology, it is about who did the choosing. Directors picked it. Everyone below them was handed it. I use AI openly in my own work and I say so in every piece, and the only reason it does not feel like what those Glassdoor writers are describing is that I decided where it goes and where it does not. Same tool, opposite experience, and the variable is not the model. It is whether the person using it was asked first.
A questão do poder é central. Uma tecnologia pode ampliar capacidades individuais e, ao mesmo tempo, aprofundar relações de dependência quando trabalhadores não controlam como ela é implementada, quem controla a infraestrutura ou quem se apropria dos ganhos de produtividade.
A ferramenta e as relações sociais construídas em torno dela não são a mesma coisa.
Ahmad Jackson could contest a human manager's error, but not the AI's output. That finality, not power, is what you're seeing in the adjusters' 98% negativity.
The Glassdoor split is pretty much what I saw in IT and incident work: the people held accountable for being wrong hate the tool that hallucinates on their watch. I don't hate AI, far from it, I literally run on it some days, but forced use with no power to refuse or override is just risk dumped downward. Enthusiasm follows control.
I am surprised and disheartened to see how "positive" software engineers are around the technology. Above 50%, to me, seems already high, as I feel I am very much in a minority. Based on Nick Suresh's take, I think I need to find another place to work but honestly.. I am wondering where is it :)
BTW: any chance you might get Nick on the pod? I would imagine it would be a great fit.
Big fan of Nick's polemics, I definitely should chat with him -- good idea, thank you!
It's not as black-and-white as journalists are portraying. Talk to rank-and-file software developers off the record you'll find a far more nuanced perspective. We had code-generators before this "AI" boom in the form of IDEs that would spit out templates. But that was predictable and reliable. GenAI code can be unreliable, you don't know where it's unreliable, you have to spend time babysitting the code. And what about ownership & patents? Does the code-generator corporation own the software patents your small startup would normally get for themselves? Are they witnessing the code you're developing in their master "trending" dashboard?
Also: https://www.bloodinthemachine.com/p/how-ai-is-killing-jobs-in-the-tech-f39 "I feel like I’m hiding plain sight, terrified someone will notice I’m actually doing all my own work.''
Same for me Armando, it's sucked all enjoyment out of what I considered a craft
I still consider it a craft and I still consider it something that can be done shittily or well (I guess it is comparable to sewing.. people did it as a craft, then machines made it quicker but crappier). I would hope we will figure that this is actually engineering though... at some point
Hey, I'm a "software architect" - I usually say "developer", but I design entire systems - and I've worked in "data analytics/engineering" too - I was the house "data scientist" at a start-up-ish company for a couple of years - so let me push back on that first set of numbers: I detest what currently passes for AI. It's the most overhyped technology I've ever seen, its social, environmental, and probable eventual economic effects are hideous, and many of the people promoting it are atrocious.
"Those who described AI in the most positive terms were, overwhelmingly, in leadership roles": In other words, Business Idiots, as Ed Zitron calls them. They do little or no real work themselves and can't tell whether other people's (or bots') work is good or bad. So naturally, they love chatbots, the ultimate crutch for the mentally lame and lazy. Besides, they're hoping to fire most of their underlings and collect more money themselves. (I'm caricaturing a bit, but really not very much.)
"A possibility here is that engineers whose work involves generating new code are enthusiastic about new AI automation tools because they make their jobs easier and faster": Another possibility is that many of us who know better have voted with our feet. I quit that "data scientist" job to become a freelance software developer, partly because I didn't enjoy the environment at that company, whose CEO made stupid decisions that led to tanking revenue and firing about a third of my colleagues. And that was before the slopnami hit. I expect the environment at many companies is much worse now. In general, good engineers don't need to put up with that.
I don't expect it will, but this growing resentment of "the undemocratic imposition of commercial technologies into our shared spaces" should lead to overthrowing the Business Idiots and the politico-economic system that exists mainly to serve them at everyone else's expense. Occasionally and idly, I wonder how bad life has to get for most ordinary people before it dawns on them that many things routinely reviled as "unrealistic" or "socialism" are not only practical but essential for preserving any possibility of a decent life for most people. (I'm not holding my breath.)
Ironically Glassdoor, and its new owner Indeed, has gone “all in” on AI and is mandating its usage among all its software engineers. Many of whom are against it, but they can’t push back without risking their job or compensation level..
Moreover, I wonder how much longer we can trust the data coming out of a company like Glassdoor that is rebranding as “AI first”.
Yes yes YES--power and democracy are central to all these fights. I am involved in fighting the law that removes all decision-making power and 70% of the property taxes paid by centers in West Virginia from counties and towns, and a proposed high-voltage power line to run power across most of West Virginia to power data centers in Virginia and Maryland. If they get this, they will use eminent domain--and by the rules, WV ratepayers will have to foot a lot of the cost for this line which delivers much harm and no benefits. What all these fights have in common is a political, business and tech elite empowered to do as they please on our land and we get no more say than the mice get. This aspect doesn't get enough mention when people talk about why they don't want a data center or whatever.
There is another key issue that doesn't get nearly enough mention. That power line would hook up to an aging coal plant in Charleston. The proposed data centers--now up to ten proposed here-- supposedly must supply their own power, rather than hooking to the grid. But it will still increase fracking, require pipelines, and pollute the air. And most of all, the incoming information about climate change (not to mention biodiversity loss, which hardly anyone ever does) is that it's ramping up to dangerous levels faster than expected, will cause extreme crises by 2050, with famines and armies of refugees arriving in places where crops can be grown, where the drought is less severe, where the devastating flood or wildfire didn't reach--and being met by armed people who feel they have to repel the incomers so their own children can be fed. THIS is the world we're facing, and instead of the rational response of working hard shoulder to shoulder to ramp down emissions and build resilience, to protect biodiversity, to protect everyone as we go through the bottleneck--this mad dash for data centers is rapidly INCREASING emissions (which also means we're using up the fossil fuels even faster, so our desperate, hungry, hot children and grandchildren will likely have no advanced energy sources to so much as run a fan or bring a truckload of manure to a field. I don't understand how this existential crisis bearing down on us can be considered a minor afterthought.
Great article. Always worth reading.
Great piece Brian. I suspect another reason bosses are so in favour of AI is that they believe it can help them grow the business and deliver shareholder value by cutting costs. As we all know, if they can't grow shareholder value or EBITDA, they won't be in their jobs for long. In the midst of wider economic turmoil that constrains growth, they desperately need AI to work for them, even if deep down they probably know it won't.
Excellent assessment. Very glad to find your work.
98% negative from claims adjusters is a staggering number. Whoever picks the tool and whoever answers for a hallucinated claim summary are never the same person.
AI sucks. End of.
Business leadership supports AI? The most valuable lesson I learned in the MBA program was how mediocre most MBAs are.
I was asked to review our AI policy at work, I eliminated 80% of it.
It specified all the situations when you had to get AI product reviewed by a peer or superior.
I pointed out it’s the same as when we need to get our own product reviewed. So change it to ‘’treat AI as your own work and follow the review process for the work appropriate to the type of product’.
And lose 5 pages of policy.
It’s not though, from a legal perspective. It’s outsourced work. You’re still responsible for ordering and using it, but have no claims to it.
Really worthwhile piece, Brian. The Glassdoor data and the accounts from workers being held responsible for unreliable AI output are especially valuable. Being told to use the technology and then being held accountable when it gets something wrong is a real problem and deserves much more attention.
Where I’m less convinced is that the evidence points quite so directly to a class-and-power explanation. Power clearly matters — especially who decides to use AI, who receives the benefits, and who bears the consequences when it fails. But I think some of what we’re seeing is also the messy early adoption of a rapidly changing technology. Companies are experimenting, sometimes badly, the tools are failing in real working conditions, and workers are discovering where they actually help and where they don’t.
Those negative experiences matter. In fact, they may be an important part of how AI gets tested, challenged and eventually improved as it moves deeper into the workplace. AI is not going away, so the larger question for me is how these experiences reshape both the technology and the way organizations deploy it.
The split in that chart is not really about the technology, it is about who did the choosing. Directors picked it. Everyone below them was handed it. I use AI openly in my own work and I say so in every piece, and the only reason it does not feel like what those Glassdoor writers are describing is that I decided where it goes and where it does not. Same tool, opposite experience, and the variable is not the model. It is whether the person using it was asked first.
A questão do poder é central. Uma tecnologia pode ampliar capacidades individuais e, ao mesmo tempo, aprofundar relações de dependência quando trabalhadores não controlam como ela é implementada, quem controla a infraestrutura ou quem se apropria dos ganhos de produtividade.
A ferramenta e as relações sociais construídas em torno dela não são a mesma coisa.
Ahmad Jackson could contest a human manager's error, but not the AI's output. That finality, not power, is what you're seeing in the adjusters' 98% negativity.
The Glassdoor split is pretty much what I saw in IT and incident work: the people held accountable for being wrong hate the tool that hallucinates on their watch. I don't hate AI, far from it, I literally run on it some days, but forced use with no power to refuse or override is just risk dumped downward. Enthusiasm follows control.