Countering AI with human creativity nonetheless! I love it!
I like that depiction they used of artistic tools attached to a gear.
Me too. It’s exactly the kind of clean metaphor that an AI generated image would never be able to understand.
So I am so ewhat pro AI. But hear me out. I sometimes refer to myself as an automation engineer. I spend a lot of my time automating the set up and use of various software tools. For those who know the term Infrastructure As Code is a part of my job too. And soo many tools have shitty UIs and even shittier apis. The rise of AI is going to add pressure to have better apis because that is what the AI uses. So even if AI falls flat on it’s face in a few years, any improvements in apis is a vig win for me. And since the automation I write is for my coworkers, not external customers, anyone in tech benefits from this.
Now for me personally, I work ina lot of different languages and DSLs. I rarely spend enough time in any one of them to really memorize the syntax. I pretty much can’t write a working program without some sort of reference. So, I can tell AI exactly what I want it to do, and it can code and test until it runs. Then I can use that as my syntax reference and make it do what it is supposed to do. That ends up being much faster than me having to google various syntaxes to see where I need a semicolon vs a comma, or where I need to use [] instead of {}. So it helps me.
And I do love using AI to file my jira tickets. Works great for those of us who’s work is interrupt driven. We often file the ticket after we’ve solved the problem.
Fuck me if I want to have sessions with my psychiatrist that aren’t recorded and transcribed by AI. Tech companies lie all the time about what they promise, like encrypting their data, and then just pay the fines after the fact, because scraping our data to train models made more profit than the fine. Yes, my doctor rolled out a system for telehealth that prompted me to sign documents agreeing to the use of AI. A procedure code and diagnosis code is the only thing that’s required for medical billing in instances where prior approval for services isn’t required, there’s no need for a transcription of our entire fucking session word for word for my DR to bill and get paid; especially when there is plenty of transcription software he could purchase to operate in house, WITHOUT AI.
AI is being forced down our throats in so many industries that have NOTHING to do with writing code, for the purposes of reducing headcount because the line must go up.
Problem with the theory is that people believe in LLM strongly enough that whatever pressure there is within a market to be vaguely similar evaporates. SQL certainly has dialects, but at least the basics are vaguely similar, as an example.
Working with a vendor that is oddly different from every other vendor in the space and we applied pressure to implement more typical interfaces. Their answer was “just have an LLM translate for you and use our different and frankly much weirder interface”. When we did cave and use it and demonstrated the biggest LLMs failed, they said at least they give you the idea. Zero interest in consistent API with LLM as an excuse.
On the write your code for you, it has to be kept on a short leash and can be a nightmare if not overseen, though it can accelerate some chore work. But I just spent a lot of time last week trying to fix up someone’s vibe coded migration, because it looked right and it passed the test cases, but it was actually a gigantic failure. Another vibe coded thing took 3 minutes to run and it was supposed to be an interactive process. The vibe coded said that’s just how long it takes, if it could be faster, the AI would have done it and none of the AI suggestions are viable in the use case. So I spent a day reworking their code to do exactly the same thing, but do it in under a second.
For the jira ticket scenario, I had already written a command line utility to take care of that for me. Same ease of use instead of using jira GUI and my works torturous workflows, but with a very predictable result.
So LLM codegen a few lines at a time with competent human oversight, ok and useful, depending on context. But we have the similar downside as AI video/image/text creative content: People without something substantial to contribute flood the field with low quality slop, bugs and slow performance and the most painful stuff to try to fix since not even the person that had it generated understood it.
There certainly is a group of people who believe in AI strongly. One part of them is just listening to the hype and jumpping on the wagon. Another part however is investing real time to understand it. They work to give it structure and guardrails so that it does what they want it to. And they help others do the same. But currently it still takes a lot of time investment to get good at using. And most people aren’t expecting that.
But as the second group grows, and the methods for them to share the structure they have set up for AI mature, more people will be anle to use it without all the upfront time investment. That is when the pressure on tool vendors to improve their api interfaces will really heat up. AI compliant or whatever buzz word shows up will be a near requirment for a tool to get investor dollars. MCPs were an attempt to put a layer between the apis and the AI. But if the underlieing api sucks, MCP can’t do much. I am not sure what will come next, but something more about the apis themselves is bound to spring up. Maybe even several standards. Thats ok, there can be several because AI can handle the context switching better than humans can.
It will end up as an assistant rather than take over the world at least while it is affordable
I truly believe an AI winter is coming
There are absolutely some economic factors that can have serious impact on it. And they are impossible to predict. If you really could, you would be rich. But, I don’t see it likely to be an assistant. It’s actually pretty terrible at that. My thinking is that it is a tool like any other. It will take a person significant investment to get proficient with it. Down the line, hopefully it will be more streamlined to distribute learnings and such that make it more accessible to those who haven’t invested the time. There is lots of work happening in that area now, but much more needs doing.
Or they’ll make apis shittier because they don’t want AI using it.
However, Copilot has made it a lot easier to navigate through Azure’s incomprehensible menu structure.
Well, grafana is an example. They want their own AI agent that you can pay for. So they still need the apis to be good. But they don’t make it easy to get your AI it own api token. Each user would essentially have to have two accounts. Which they probably charge for too. It’s not impossible to work around, but it’s a barrier. I would expect more of that kind of thing. Any tool that doesn’t have a way for AI to work with it is going to be selected against for a while. So there is pressure for them to be accessible.
I am in the same boat, long time infrastructure automation engineer as well. Sometimes it’s faster to explain how terraform or whatever needs to act and then fix the issues rather than having to sift through the docs for every provider.
I also do a similar thing to you with code, I also have to read a lot of other people’s code in languages I don’t know to help troubleshoot things and while I can usually follow the logic it is such a time saver to have AI to read the docs for the libraries and languages for me to at least find the part of the docs I need to read faster than searching myself.
Overall, I also agree with the sentiment on AI most of the time and all of its criticisms are definitely valid but I think too many people try to use AI to do their work for them instead of using it more like a rubber duck you can program with normal language.
My new (to me) “revelation” is that AI needs a ton of structure. It’s like a child who when presented with too many options stops thinking and just randomly chooses one to do so they can be done with whatever it is. From what I can tell, the people who make the most use of AI have it tightly controlled. Rules, hooks, and various other tricks to essentially herd AI into doing what it should do. Kinda like herding cats.
Right now the tools and such for setting up that structure are immature, and best practices are hard to define when the base AI is still changing a lot. For people who are just trying to use AI casually, they have heard the hype, and they think they should expect it to work like a person. When it doesn’t, they just say it sucks. And as a person, it does suck. It’s a tool. And a complex one at that. Seems it requires significant investment to get the most out of it.
The closest facsimile I have in my work is occasionally running an Excel formula I’ve written through Copilot in order to find a formatting error or to help fix an Access query, but If fundamentally understand what I’m doing, can validate that the produced result is correct, and can fix it if I have to somewhere down the line.
It’s good you’ve found some simple ways to use it, but in the vast majority of work I do, it would take longer if I used AI because everything produced using an LLM has to be human-validated regardless, so I might as well not skip the important step of learning and understanding it.
I never use it to ideate and never use it for anything that isn’t eminently simple, like creating a sheet with x number of columns and rows or something like that. I hate the idea of the environmental impact and that helps me avoid it.
And outside coding its like modest productivity improvments is the best we’ve done in the 4 years we’ve had these models.
I just can’t see it not being a bubble
Yeah, it’s nothing particularly special IMHO. The best feature I’ve found in using it is that, for Microsoft products in particular, it can tell me capabilities of certain things I didn’t know previously when I present it with a problem.
Search engines used to do that before they got enshittified.
Or we can refuse to listen to anti-“AI” crazies.
And eat the dick billionaire shoving at your face willingly? Because that’s what AI are, people are forced to used it because their boss demand it.
So maybe you should fight against dick-shoving billionaires, not against useful tools?
AI is not a useful tool, it’s a lock-in subscription that chains you to those billionaires. Fighting against AI is fighting against control by billionaires.
AI is the Loom.
You are the Luddite.
Big tech is the Workhouse Owner.
I mean…looms actually seem useful. My experience with large language models is that they’re only useful when the output doesn’t really matter. Like…they’re fine if you’re “searching” for things that aren’t really defined and you don’t really care about the answer (i.e. “what are the five trendiest coffeeshops in Barcelona that are likely to have english speaking staff?” it can’t actually know any of that…what’s “trendy” even mean? Whatever, who cares, go to a coffee shop on your vacation, have a nice time).
But when it matters you just cannot rely on them…They can’t be relied on to use the correct words when precision of language matters, they can’t do “research” or “analysis” in any meaningful sense…like maybe better than a sharp middle-schooler? But not as well as a dumb undergrad.
And I don’t see any reason, understanding what the technology is to think they’ll get better at those things. It’s predictive in nature. You know…like maybe it’ll go from 60% reliable to 90% reliable over the next hundred years because they’ll find some way to focus on high-quality and relevant training data, while still using gigantic training data to get the model up and running…? But since it’s fundamentally a predictive model (trying to predict what a good answer would look like), it’s never going to be able to actually be relied upon for answers to questions when it matters.
And idk what the cost would be when factoring in all the externalities…environmental destruction, energy consumption…hell, even the infrasound from data centers fucking up everyone’s brain…like…there’s just no way this makes any economic sense. Right now it’s all mega-subsidized, but when that comes to an end…is it gonna cost $10 per prompt on average? $50? Idk, but I know everyone using it now will not want to pay for it.
Looms are bad. They don’t do silk. They can only do big blocks of material. Even so, lots of people want to use them.
But my point is that going around attacking AI will be as effective as the luddites destroying machinery.
The villains are the same in both cases. Capitalists.
AI is a broad term; of course neural networks and machine learning have been important in a lot of research etc. That’s all great. LLMs…it’s all anyone wants to talk about (maybe image generation too) and it’s junk for any application that matters.
If looms could only make burlap, and the capitalists tried to make burlap underwear a thing, I think the luddites would be wise to say to the public “hey, don’t buy this crap…it’s uncomfortable!” Of course, in reality, auto-looms did a lot of the same stuff traditional weavers could do. I think pointing out that when techbros say LLMs output is great, pointing out that LLMs output is generally garbage is effective. Luddites couldn’t really say that the output was significantly inferior (or maybe it was and people didn’t notice…jesus I hope that’s not the case with this garbage!).
Maybe that’s what we disagree about. To me, the auto-looms are only making burlap and I don’t see any reason to think they’re going to get much better. And they’re lighting the planet on fire :P
I am not willing to capitulate to this kind of BS: “LLMs are very useful and they’re clearly here to stay.” I just think that’s horseshit. That’s what the capitalists who are selling them want you to think, but I genuinely believe if you ever look at it in a critical context you’ll see.
Correct. Similar problem, different era. OSS collectivized the means of production for software dev to some extent. AI will make that meaningless and chain software development to capital again. That alone is a reason to fight it.
If they’re so useful, why are they being forced on everyone, including by making them part of performance reviews?
If they’re useful people will naturally use them.
And people do use them. Naturally.
What about that “forcing” thing you’re talking about? Look around. You’re being forced with everything by corporations. Why would this new cool technology should be an exception? You’re forced to watch sport events, listen to modern music, wear some vogue clothes, kiss your beloved leader’s ass, hate those evil Cubans or Ukrainians (depending on who your owner is).
temu tyler durden
Why would this new cool technology should be an exception?
I very often do push back against things that they try to force down everyone’s throat. AI is not exceptional in this regard.
Wow, you must be the weakest and the easiest to manipulate and brainwash human being on the planet.
The projection is strong in this one
No, I do none of those things. You just sound like a resentful conformist. Maybe try thinking and doing for yourself, instead of thinking and doing as you’re told.
So cool…
How about allowing AI to be rolled out and adopted organically instead of trying to gavage it down everyone’s throats?
That’s the question to corporations, not to LLMs.
That would be true of any technology. Technology doesn’t have agency; my comment is clearly directed at corpos that are pushing it.
“AI Tools” describes both the product and the people who use them.
not against useful tools
Nobody’s fighting against you guys. Why do sloperators have to take everything personally? Smh my head.
What “usefulness” do you get out of them?
They save my time tremendously while searching for something in documentation. Especially if I don’t know if it is actually there.
Are you not familiar with “ctrl+f”?
Didn’t know ctrl-f could parse natural language and not only rely on knowing the correct keyword. When did it gain that functionality?
Better question is, when did you lose basic keyword-based searching skills? I know you may want your answers on a platter but realize that there’s value in manual searches. Searching for something with LLMs on the page that you’re on is questionable on so many levels.
They’re not useful tools.
I’d that tool didn’t come at the destructive costs involved, AI would be a lot more palatable.
destructive costs
YouTube storing shitillions of dickabytes of cat videos “costs” much more while being completely useless. But those are funny cat videos. Hands off of those videos. Yes?
You may think you’re being clever, but that is hardly a reasonable comparison while also ignoring the glaring corporate irresponsibility underlying both.
If that’s what it takes to stop the excessive destruction caused by unregulated data center construction and operation, yes.
yes.
What “yes”?
Do you need to reread my comment instead of reacting to a single word of it?
If AI is just like a video data center, why did data center energy usage stay stable before AI? Why has data center energy usage doubled since 2010 if videos and AI are equivalent in energy usage?
I don’t care what you do but keep your hands off those videos. I need them for things.
You’re screaming into the echo chamber, mate. Unless you’re so rabidly anti-AI you believe and spread one of a few comforting, imaginary narratives, you’ll be dog piled.
I’m staunchly critical of AI, but won’t pretend that it only consists of generative AI, that it still operates as poorly as it did years ago, nor that a disturbing percentage of the population either doesn’t care about or actually supports that shit, so I get my share of insults. Being pro-AI won’t get you much civility so set your expectations low. Unless you’re trolling. Then you’ve nailed it.
I’ve never seen this many downvotes on a comment. 👏
Exactly, it’s weird the anti movement, AI is exceptional in so many ways. Maybe the ruling class don’t want the general population to have access to and like something so fundamentally useful and have created a smear campaign
Guys trying to shove AI into everything: oh nooo, pleeeaaase don’t use our product, we would haaaate it if you did that!
Worst conspiracy ever.
It’s not the product makers that are creating the smear campaign, it’s the people who don’t want the general populous to access power AI.
So they have the power to run a smear campaign that would alter society’s opinion on AI—which is so very useful and everyone would love it otherwise—but they don’t have the power to influence the building of billions of dollars of infrastructure to power these data centers?
The people pushing this are the people that benefit from the general populace being dumb as shit. That’s why they’re pushing it. AI doesn’t make anyone more informed or intelligent, it’s an echo chamber.
Yes, they have the power to smear, it’s not that hard.
They can’t stop then building why would them being able to smear mean they can’t build data centers? That doesn’t make any sense.
The anti AI groups can smear, I don’t see what that has to do with AI companies building data centers.
AI gives informed and data resourced answers. It’s better than just reading Facebook comments.
It increases intelligence because it shows the best way to do something, and everyone can learn from that.
It’s not an echo chamber, it provides answers for almost everything.
I think is people don’t like it they don’t need to use it. I’m ok with rules stating ai generated content must be labeled though. I wouldn’t even mind a toggle so it could just be turned off.
But it is tech that is here to stay. It is useful to many people.
I think you might be surprised. Generative AI has limited utility and costs a lot to operate; so much, in fact that t does not appear there are enough natural resources on the planet we’re on to ramp it up to the scale that is intended. Soon, the hype-based funding will dry up, and the free and subsidized generative AI tokens will all disappear. Only then will we see the true cost of using it and if users will bear that cost. If it costs a lot of money to ask it to do things, people will go back to doing a lot of those things themselves.
It is allowing me to do things that I could not do on my own due to time constraints. It also allows me to do stuff that I could not afford to do (generate my company logo and variations of the mascot for social media posts.)
I know a lot of people think I am rich, but I live in a third world country and hardly have any extra to save, so I truly can’t afford to pay a graphic designer to do that work. I actually did once, on upwork, but the vector work she did was terrible and she couldn’t figure out how to remove these thick black lines. ChatGPT was able to fix it though.
It’d be fine if that was the case. Right now if you don’t like it you’re still forced to read (and often review) AI generated rumblings, communicate with LLMs instead of humans when contacting support, accept AI-specific terms even you won’t use the AI part of a product, have data centers pollute your city and pay ridiculous amount of money for a stick of ram.
I think is people don’t like it they don’t need to use it.
Tell that to the people living near new data centers who can’t get clean water and are being charged exorbitant rates for electricity. They have no say in the matter.
- Lincoln Institute of Land Policy | Data Drain: The Land and Water Impacts of the AI Boom
- Consumer Reports | AI Data Centers: Big Tech’s Impact on Electric Bills, Water, and More
- Forbes | America’s AI Boom Is Running Into An Unplanned Water Problem
- CNBC | AI data center ‘frenzy’ is pushing up your electric bill — here’s why
This is occuring all over the US, these issues are far from isolated incidents.
Useful to who? If you need an LLM to write just a basic e-mail/comment/caption you’re maybe … how do I say this nicely? Not that smart …
If you use an LLM as a search engine, same thing.
If you use an LLM as a psychologist, same damn thing.
And the majority of people are using it for those things. It’s just plain stupidity. I’m not saying there’s no use to AI, but right now it’s being used in a terrible way by people that have no use for it at all.
That would be wonderful if those anti-AI folk would stop using LLMs and switch their attention to something more constructive. But they can’t.
Imagine letting an entertainment product write your code for you. Why the fuck are you doing this if you don’t even like the act of programming?
No idea why people do that. I suppose some people are too dumb to write code and some other people are too dumb to understand what programmers use LLMs for. What dumb people do best? Attentionwhoring, screaming and throwing hysterical tantrums.
How do dumb people get dumber? By letting AI do their work and thinking they’re saving time, while in reality, proven by multiple studies, they are not. I would recommend you do some research and read the studies, but you will probably just go to your sycophant AI agent for the research.
There’s basically no time saved, but people do get dumber because they’re not thinking critically themselves anymore.
This dumbness and lack of critical thinking is very apparent in your case.
I get writing boilerplate and unit tests can probably be done by software well enough, at least when supervised.
Ill be honest, that’s not even my real issue.
My real issue is that programming, devops, systems administration. All of these things are art forms, every bit of them. From high-level application architecture down to the tiniest details of implementation.
Like how much of a library you choose to include, what you name your variables, what type of loops you use to iterate through data. How you choose to format and comment your code.
Giving these choices to the machine is like the painter giving their brush to it.
Just like images generated by stable diffusion will never be worth their fully-human painted equivalent. So too will LLM-developed programs fail to hold that value.
For what its worth, this isnt new. I’ve held contempt for VC-worshipping developers who see programming as a means to an end far longer than LLMs have been used for serious work.
🙄
The thing is people using this stuff are doing harm to the planet and society. So leaving them alone is not going to happen.
