Great news everyone! I finally talk about AI hype. Someone finally mentioned LLMs one time too many, and the reckoning is upon us:
https://ludic.mataroa.blog/blog/i-will-fucking-piledrive-you-if-you-mention-ai-again/
Great news everyone! I finally talk about AI hype. Someone finally mentioned LLMs one time too many, and the reckoning is upon us: https://ludic.mataroa.blog/blog/i-will-fucking-piledrive-you-if-you-mention-ai-again/ 163 comments
12
@ludicity I miss the old days when it was decided an app needed AI. So a data scientist made a simple linear regression, and we would add it to the app. Then, BOOM, singularity achieved. At least from an executive perspective. I always had a good laugh at them, but at least the ML did something. The only good thing I'll say about LLMs is copilot will write my unit tests. Which is why I am now arming myself for the robot revolution. As this will be considered a capital crime by them. @ludicity Very enjoyable read, thank you. I wish there was some mention of the energy costs / climate impact of the tech tho, which TBH is what worries me the most. @xarvh @rhempel True, though I should point out the LLM revolution is slightly unique. They don't need engineering discipline - you just make an OpenAI request. If they move on to something else, it might suck, but it is actually harder to damage the environment through something like a data governance hypefest (unless it's GPU or blockchain powered). These are idle thoughts though, it could still go terribly and I haven't thought on what's next deeply. @rhempel@0mstdn.ca @xarvh @ludicity > Then the question becomes what to do with all the large, lightless, but well cooled empty buildings. Not the actual point of the original post or your comment, I realize, but my first thought was that if you can add effective air ventilation and filtration (as well as lighting, presumably), these may repurpose as public cooling centers in extreme heat events, which we will have increasing call for from climate change. @ShaulaEvans @rhempel @xarvh @ludicity half of them into non-AI datacentres, and half of them into house or hostel space @ludicity this was a fun read. Thank you for articulating many of the thoughts I was struggling to convey. It’s a weird time to see all of this happening @skinnylatte Hah, I was mostly ranting, but hopefully some of the thoughts might be useful the next time the topic comes up. @ludicity I don't know that unhinged rants really help convince anyone who's not already sick and tired of that shit, but as an avid reader of yours, I have to admit it does wonders as a mean of venting frustration by proxy (and it's always good to feel like you're not alone...) @ludicity @fthevenet I think it helps a little in disinclining people from throwing out the ML baby with the gen-AI bathwater... Though that might still happen, it has in the past... @underlap @ludicity "If you liked that, you'll love this...": Why ChatGPT is literally a bullshitter https://link.springer.com/article/10.1007/s10676-024-09775-5 @ludicity Good stuff and pretty much where I’ve been for a while now. The grift around LLMs is infuriating. With a few exceptions, I immediately distrust anyone who talks about “AI”. @bjn @ludicity The grift-hopping is what gets me. Hardly anyone notices these quantum blockchain AI grifts collapse, because the new grift is already there to distract people. It's a bit like a virus that mutates, and occasionally it mutates to something that grows big, sucks in billions and billions of investor money and just gives it to somebody else without creating anything of worth. What a weird system. Too bad the last two iterations are literally burning the planet in the process. @ludicity My only quibble with it is that I’m quite sure you don’t need the last sentence in the opening section. Righteousness needs no forgiveness. @ludicity Thanks for a refreshing & entertaining reality check on the AI grift. I'm a grumpy old semi-competent Scrum-averse Database Guy (mmm, lovely schemas ...) grinding through the twilight years of my career in the bowels of the data mines, keeping Broken Old Shit running because nobody can afford to fix things properly, while waves of functionally pointless & financially/environmentally disastrous LLM slurry flood in around us. So good to know we're not alone in our weary scepticism! @ludicity Excellent post :ablobcatattention: I completely get your point - as someone who spent over a decade in data warehousing, seeing "enablers" suddenly pushing data lakes as a solution to everything (completely without ACID in the beginning) gives me similar feelings :ablobcatcoffee: Industry is poisoned with Gartner's blabber and various kinds of techbros catch up quickly :blobcatdunno: Someone get the burn cream, stat, ‘cause AI bros got roasted in the hot and spicy (well, unsure on hot but definitely ghost pepper spicy) take on the generative #AI hype train by @ludicity 🔥🌶️👏😂 https://ludic.mataroa.blog/blog/i-will-fucking-piledrive-you-if-you-mention-ai-again/ @ludicity pure brilliance, thanks. I’m going to paraphrase your student cheating idea next time some upper management ignoramus suggests it in a faculty meeting. @ludicity Got here from the LLM post but stayed here for the agile one, and holy crap I get such clear flashbacks to a company I attempted to work at that was such a travesty that I GTFO'd in 3 months. I asked them to try async Slack-thread stand-ups because the ones on Zoom were such a mind-numbing waste of time...and multiple of the "engineers" seriously said they preferred the Zoom ones. Incidentally, that team had years of tech debt, and the company had numerous pro-RTO chucklefucks, too. @TindrasGrove Good to know! It's interesting because I really am not a sophisticated actor in the security space, but it's still quite obvious when some people are full of it. Although, of course, I'm sure slightly more savvy grifters sneak past my detectors. @ludicity I think there’s some significant overlap in our fields (especially when it comes to who is actually using the not-snake-oil), so there’s some amount of transferability in BS detection skills. Last week I went to a local data analytics conference, and the talk I got the most out of was the one person who said “you don’t need AI for any of this!!” The people who try to sell zero trust as a product, not an architectural philosophy, seem to mean SSO, but ✨fancy✨ I think there’s another possibility that it seems you didn’t mention (I skimmed, sorry): namely, that generative models actually get worse over time as they consume more and more of their own excrement. The ensuing AI centipede of excremental progress will eventually collapse into a shit pile. @TonyVladusich @ludicity There is a boom in data annotation driven by demand for training data cleaning services I know most CIOs at Corporations of Significant Size are the same, but dang if you aren’t describing the one at the one I work at. Your podcast's name made me laugh harder than any joke I heard in the last year. My god. @ludicity The LLM output of the articles sounds a bit passive-aggressive, can you adjust a bit your prompt for the blog post? Ok, that joke was tasteless, what I literally hate is that EVERYTHING under the sun nowadays is called an AI. You know, things that 5–10 years would just go by “algorithm”. Or “heuristic”. Today for marketing purposes the same f%cking two if statements are already “AI”. @ludicity You might be on a good trail here. Management in your average company has been usually outsourced to the people just clever enough to feed themselves in public without embarrassing the company. @ludicity Sorry, I know that even a decade as an IT consultant gives too small a sample, but that's my conclusion. A little bit more than a decade ago, I just started at a company that makes toll systems, and as it happened, in the first week I had the privilege of listening to the 2 hour “state of the company” speech given by the C-idiots. @ludicity At the start the big picture: e.g. observations like video-based systems are the future, no complicated setup, no need for devices in the cars, bla bla. In the second hour, more concrete measures and how they apply to the company: downsize the video camera engineering group by 90% down to 2 engineers, we'll ask our competition to sell us their proprietary video tech on the cheap instead of continuing to develop it in-house. @ludicity I may not vibe with the style, but it’s a great article anyway. Thanks! @ludicity "I don't actually know what 'zero-trust' architecture means, but I've heard stupid people say it enough that it's probably also a term that means something in theory but has been sullied beyond all use in day-to-day life." Yeah, that's spot-on actually. Zero-trust used to mean "don't allow anybody to do something just because of their IP address" - i.e. place zero trust in the network. Now it somehow means more VPNs. No, I don' t know either. @ludicity Thank you. I am not a software person at all, but this answers all my suspicions about AI, and tells me I was right. @ludicity as the only local support for 150+ users I'm fed up with users and managers trying to sell me "AI".. and I'm getting angrier and more confrontational every time, now I just ask them "how much energy and water was wasted in what you just did?" But lovand behold! This is the future!! Fuck that noise @ludicity > if you continue to try { thisBullshit(); } you are going to catch (theseHands) My God. Yes. YEEESSSSSSSSSS "..spending half of the planet's engineering efforts to add chatbot support to every application under the sun when half of the industry hasn't worked out how to test database backups regularly." @ludicity While I haven't managed to turn my 35+ years in IT and software development into a credible (or marketable) body of experience, I do recognize the same rusty lamppost every time I make another revolution on the carousel. I suppose, one day, our corporate leaders will achieve their dream of eliminating workers, er, sorry, Operating Overhead, from their profit machines, er, Businesses. But I can't ignore the fact that I've heard this tune before. @ludicity @ludicity you’re talking about the #MBA class. They all need to be pile driven so deep into the earth they melt from the core temperature. And our universities are cranking them out faster than any other *profession* @ludicity Good read. "...these tools aren't good enough (yet) to assist me with this seriously." LLMs will never have the ability to design software or help designing it, as they cannot be creative (there's no thought process involved). Training them more and more or adding more computational resources will not change this. It's a bit like with Mr. Creosote in the Monty Python sketch. @thomasfuchs @ludicity They're great at "solving" (regurgitating answers to) problems that a human has already solved and that were scraped into the training data, but they're useless at problems not in their training dataset. So LLMs should be useful at designing software provided some data annotator in India has already done the work for the specific prompt you're requesting. @ludicity "average person speaks 10 phonemes per second" factoid actualy just statistical error. average person speaks 0 phonemes per second. Synergy Greg, who lives on blockchain & speaks over 10,000 phonemes per second, is an outlier adn should not have been counted @ludicity That point about organizations that i) cannot reliably manage a CRUD application thinking ii) that rolling a LLM into their stack is gonna revolutionize their warez is *fantastically* spot-on @ludicity LOL "spending half of the planet's engineering efforts to add chatbot support to every application under the sun when half of the industry hasn't worked out how to test database backups regularly." @ludicity I appreciate the segment about you visiting a less developed nation and how that made you feel about how funds in tech are distributed. @ludicity (also I can attest the “quietly understood” portion is often explicit. Depends on which company has more power, but “we’ll give you a deal on our software so you’ll pretend to use it, and you write about it on your blog five times and once in a YouTube video” isn’t uncommon, and in writing as a “marketing agreement”. If you’ve ever wondered why all the startups use each other, well, this makes them appear to have customers when it’s all VC money) @ludicity I love it. It's cathartic to read somebody's rant and think "Thank the Gods, it's not just ne" @ludicity Metafilter is having fun with you blog post. https://www.metafilter.com/204248/But-We-Will-Realize-Untold-Efficiencies-With-Machine-L @ludicity I recently read the first few paragraphs of a "UX expert" talking about how "the age of conscious AI is upon us" and how technology is no longer a "tool" but a "companion". It was all I could do to stop myself from rage-throwing my computer across the room. If people like this are truly the "UX experts" today, god help us all. @ludicity "Everyone is talking about Retrieval Augmented Generation, but most companies don't actually have any internal documentation worth retrieving" Any internal documentation they had was in the heads of people who walked out the door with the last RTO mandate, or were escorted out when management decided to replace them with "AI". @ludicity Tech lawyer here, one step removed from the hype frontline but we swim in the slip stream (“AI Regulation”). If you need someone to get you off the murder charge, a few guys in my team would probably work pro bono 😂 @ludicity nice callouts on blockchain and quantum computing ... and I’d like to add all the little darlings of the dot-com and post dot com (Java, Ruby on Rails etc) that seem related in that “grinning idiot avatar trying to make lots of money” thing you talk about Other than being a touch more positive on the machine singularity than I am personally, this was a good distillation of what I've been ranting about myself. @ludicity thank you for this, such an enjoyable read. So many truths business doesn't want to hear.👏 @ludicity As a survivor from the AI hype bubble of the 1980s, I feel your pain. Our company was producing tools which did something fairly modest but genuinely interesting; but the grifters promising things which could not then and still cannot now be delivered so poisoned the space that the whole market collapsed and we -- and a lot of other small companies producing modest but genuinely interesting tools -- were swept away. "And then some absolute son of a bitch created ChatGPT, and now look at us. Look at us, resplendent in our pauper's robes, stitched from corpulent greed and breathless credulity, spending half of the planet's engineering efforts to add chatbot support to every application under the sun when half of the industry hasn't worked out how to test database backups regularly." @ludicity Loved your article! You eloquently express many of the feelings I have about the whole LLM hype; I just feel that my managers are going to be less enthousiastic about your... colourful language. @ludicity I don't even know where to start here. Our org has absolutely power-shovelled well north of $100M into internal AI product development and exactly 0 of them are in any kind of operational state, let alone production. One team's been working on deploying a single linear regression for about 4 years, and are still trying to morse code "data fucked, product owner quit, bomb our location" back to the parent org. Also have about a 1:300 DE to DS ratio which tells you all you need to know. @ludicity I'm a smoking rabbit because I would literally be stomped to death by our exco for even exhaling slow enough for it to sound like a sigh when someone mentions a new AI opportunity, so strong is the fear of dissent. The only slightly lesser sin is suggesting things like product management and data fundamentals might be more critical to good outcomes over carpet bombing AI into everything. @ludicity thank you for providing a perfect response for every single topic that starts with "how about implementing AI" @ludicity and we shouldn't stop there! In fact, let's put an end to infinite meetings https://youtu.be/JbpkJKmpyJw @ludicity fantastic. I got sick of the entire money grub direction of the silicon crowd in the 90s and decamped to the Midwest to problem solve clothing. Pretty relaxing. Good luck with the theater adventure! Mmm costumes! @ludicity @adrienne > You can hear it too, on freezing nights under the pale moon, when the fire burns low and the trees loom like hands of sinister ghosts all around you - when the wind cuts through the howling of what you hope is a wolf and hair stands on end, you can strain your ears and barely make out: > "Just Use Postgres, You Nerd. You Dweeb." @ludicity "[It] turns out that the core competency of smiling and promising people things that you can't actually deliver is highly transferable." BOOM 💥 @ludicity This was hilarious and felt so good to read. Really hoping for another AI winter just so the hype train ends @ludicity The impression I've gotten of the state of the tech is that it's more like a spinal cord than a brain, but executives think so little of sapience itself that they can't tell the difference... and, since they resent the idea of needing employees, don't want to hear that there is one. @ludicity Lol dude I love your writing style, have been binge reading your blog since an hour and wow after years it feels coming across an actual human being in tech. Such a respite from the usual corporate "thought-leadership" @ludicity O thank you. Alas that the grifters haven't the wherewithal to understand their sin. @ludicity loved reading this, your writing styles really funny and engaging, thanks :ablobcatenjoy: |
@ludicity well said. Also, if I hear more about Gartner magic AI squares and positioning relative to them, I’ll be ill. Plus, I’m looking forward to the talk once/if it happens.