If in 2009 you claimed that the dominance of the smartphone was inevitable, it would have been because you were using one and understood its power, not because you were reframing away our free choice for some agenda. In 2025 I don't think you can really be taking advantage of AI to do real work and still see its mass adaptation as evitable. It's coming faster and harder than any tech in history. As scary as that is we can't wish it away.
…is exactly inevitablist framing. This claims perfect knowledge of the future based on previous uncertain knowledge of the future (which is now certain). You could have been making the same claims about the inevitability of sporks in the late 19th century and how cutlery drawers should adapt to the inevitable single-utensil future.
I read a story about 14 year olds that are adopting AI boyfriends. They spend 18 hours a day in conversation with chatbots. Their parents are worried because they are withdrawing from school and losing their friends.
I hate second guessing emails that I've read, wondering if my colleagues are even talking to me or if they are using AI. I hate the idea that AI will replace my job.
Even if it unlocks "economic value" -- what does that even mean? We'll live in fucking blade runner but at least we'll all have a ton of money?
I agree, nobody asked what I wanted. But if they did I'd tell them, I don't want it, I don't want any of it.
Excuse me, I'll go outside now and play with my dogs and stare at a tree.
If someone invested a lot of money in something, they probably are convinced that something is inevitable. Otherwise they would not invest their money. However, sometimes they may be a little bit helping their luck
There's plenty of examples where important people framed an inevitable future and then it didn't pan out.
Somewhat objective proof of "progress" will inevitably win out, yes inevitable framing might help sell the vision a bit, for now, but it won't be the inevitabism that causes it to succeed but its inherit value towards "progress".
The definition of "progress" being endlessly more productive humans at the cost of everything else.
The majority of the comments here reflect an acceptance of or even an enthusiasm for an LLM-using future. An embracing of the technology regardless of its downsides. A disregard of those who question whether it’s all a desirable future.
I’d have thought perhaps we’d learn the lessons of eg. smart phones, social media, cloud, VR, crypto, NFTs, etc, and think a little more deeply about where and how we want to go as a society and species beyond just adopting the latest hype.
> I’m certainly not convinced that they’re the future I want. But what I’m most certain of is that we have choices about what our future should look like, and how we choose to use machines to build it.
While I must admit we have some choice here, it is limited. No matter what, there will be models of language, we know how they work, there is no turning back from it.
We might wish many things but one thing we can't do is to revert time to a moment when these discoveries did not exist.
We have no idea how they work. We know the training making a model and generating things from it, but the thing we want from it being an Oracle or something, we have no idea how that works at the level of the specific knowledge it generates. And when we dig into it doesn't reveal anything very interesting. The premise of put everything digital in a pile and see if makes something like a god is a neat idea, but the god is literally playing dice.
This is the same strategy Hillary Clinton supporters tried to use too. The author is right, it's just a framing technique. We can choose the future we want.
It’s also possible for LLMs to be inevitable, generate massive amounts of wealth and still be mostly fluff in terms of objective human progress.
The major change from my perspective is new consumer behavior: people simply enjoy talking to and building with LLMs. This fact alone is generating a lot (1) new spend and (2) content to consume.
The most disappointing outcome of the LLM era would be increasing the amount of fake, meaningless busywork humans have to do just to sift through LLM generated noise just to find signal. And indeed there are probably great products to be built that help you do just that; and there is probably a lot of great signal to be found! But the motion to progress ratio concerns me.
For example, I love Cursor. Especially for boilerplating. But SOTA models with tons of guidance can still not reliably implement features in my larger codebases within the timeframe it would take me to do it myself. Test-time compute and reasoning makes things even slower.
There may be an "LLM Winter" as people discover that LLMs can't be trusted to do anything. Look for frantic efforts by companies to offload responsibility for LLM mistakes onto consumers. We've got to have something that has solid "I don't know" and "I don't know how to do this" outputs. We're starting to see reports of LLM usage having negative value for programmers, even though they think it's helping.
Too much effort goes into cleaning up LLM messes.
I find these LLM doomer takes as silly as LLM maximalist takes.
LLMs are literally performing useful functions today and they're not going away. Are they AGI? No, but so what?
There is waaay too much projecting and philosophizing going on in these comments and not enough engineering-minded comments from objective observers.
Is AI hyped? Sure. Are LLMs overshadowing other approaches? Sure. Are LLMs inefficient? Somewhat. Do they have problems like hallucinations? Yes. Do they produce useful output? Yes.
We may be underestimating the effort that goes into cleaning up LLM messes. LLMs learn to program from code bases written by humans. Not just written by humans, maintained by humans. So the bugs that humans spot and remove are under-represented in the training data. Meanwhile, the bugs that evade human skill at debugging lurk indefinitely and are over-represented in the training data.
We have created tools to write code with bugs that humans have difficulty spotting. Worse, we estimate the quality of the code that our new tools produce on the basis that they are inhuman and have no special skill at writing bugs that we cannot spot, despite the nature of their training data.
I think this a big side effect of the field moving way too fast for it to be evaluated properly. I don't recall seeing such a big rally in the CS research like this. Nearly every group I know that tackled really totally different topics are converging to LLMs. Talk about eco-diversity in the CS reaearch, all that is reduced now, LLMs are the palm trees of this field now.
This concept is closely reated to politics of inevitability coined by Timothy Snyder.
"...the politics of inevitability – a sense that the future is just more of the present, that the laws of progress are known, that there are no alternatives, and therefore nothing really to be done."[0]
1. LLMs are a new technology and it's hard to put the genie back in the bottle with that. It's difficult to imagine a future where they don't continue to exist in some form, with all the timesaving benefits and social issues that come with them.
2. Almost three years in, companies investing in LLMs have not yet discovered a business model that justifies the massive expenditure of training and hosting them, the majority of consumer usage is at the free tier, the industry is seeing the first signs of pulling back investments, and model capabilities are plateauing at a level where most people agree that the output is trite and unpleasant to consume.
There are many technologies that have seemed inevitable and seen retreats under the lack of commensurate business return (the supersonic jetliner), and several that seemed poised to displace both old tech and labor but have settled into specific use cases (the microwave oven). Given the lack of a sufficiently profitable business model, it feels as likely as not that LLMs settle somewhere a little less remarkable, and hopefully less annoying, than today's almost universally disliked attempts to cram it everywhere.
I do wonder where in the cycle this all is given that we've now seen yet another LLM/"Agentic" VSCode fork.
I'm genuinely surprised that Code forks and LLM cli things are seemingly the only use case that's approached viability. Even a year ago, I figured there'd be something else that's emerged by now.
The difference is that the future is now with LLMs. There is a microwave (some multiple) in almost every kitchen in the world. The Concord served a few hundred people a day. LLMs are already ingrained into hundreds of millions if not billions of people’s lives, directly and indirectly. My dad directly uses LLMs multiple times a week if not daily in an industry that still makes you rotate your password every 3 months. It’s not a question of whether the future will have them, it’s a question of whether the future will get tired of them.
Investments are mostly in model training. We have trained models now, we'll see a pullback in that regard as businesses will need to optimize to get the best model without spending billions in order to compete on price, but LLMs are here to stay.
(1) Model capabilities will plateau as training data is exhausted. Some additional gains will be possible by better training, better architectures, more compute, longer context windows or "infinite" context architectures, etc., but there are limits here.
(2) Training on synthetic data beyond a very limited amount will result in overfitting because there is no new information. To some extent you could train models on each other, but that's just an indirect way to consolidate models. Beyond consolidation you'll plateau.
(3) There will be no "takeoff" scenario -- this is sci-fi (in the pejorative sense) because you can't exceed available information. There is no magic way that a brain in a vat can innovate beyond available training data. This includes for humans -- a brain in a vat would quickly go mad and then spiral into a coma-like state. The idea of AI running away is the information-theoretic equivalent of a perpetual motion machine and is impossible. Yudkowski and the rest of the people afraid of this are crackpots, and so are the hype-mongers betting on it.
So I agree that LLMs are real and useful, but the hype and bubble are starting to plateau. The bubble is predicated on the idea that you can just keep going forever.
> 2. Almost three years in, companies investing in LLMs have not yet discovered a business model that justifies the massive expenditure of training and hosting them, the majority of consumer usage is at the free tier, the industry is seeing the first signs of pulling back investments, and model capabilities are plateauing at a level where most people agree that the output is trite and unpleasant to consume.
You hit the nail on why I say to much hatred from "AI Bros" as I call them, when I say it will not take off truly until it runs on your phone effortlessly, because nobody wants to foot a trillion dollar cloud bill.
Give me a fully offline LLM that fits in 2GB of VRAM and lets refine that so it can plug into external APIs and see how much farther we can take things without resorting to burning billions of dollars' worth of GPU compute. I don't care that my answer arrives instantly, if I'm doing the research myself, I want to take my time to get the correct answer anyway.
I don't really buy your point 2. Just the other day Meta announced hundreds of billions of dollars investment into more AI datacenters. Companies are bringing back nuclear power plants to support this stuff. Earlier this year OpenAI and Oracle announced their $500bn AI datacenter project, but admittedly in favor of your point have run into funding snags, though that's supposedly from tariff fears with foreign investors, not lack of confidence in AI. Meta can just finance everything from their own capital and Zuck's decree, like they did with VR (and it may very well turn out similarly).
Since you brought up supersonic jetliners you're probably aware of the startup Boom in Colorado trying to bring it back. We'll see if they succeed. But yes, it would be a strange path, but a possible one, that LLMs kind of go away for a while and try to come back later.
You're going to have to cite some surveys for the "most people agree that the output is trite and unpleasant" and "almost universally disliked attempts to cram it everywhere" claims. There are some very vocal people against LLM flavors of AI, but I don't think they even represent the biggest minority, let alone a majority or near universal opinions. (I personally was bugged by earlier attempts at cramming non-LLM AI into a lot of places, e.g. Salesforce Einstein appeared I think in 2016, and that was mostly just being put off by the cutesy Einstein characterization. I generally don't have the same feelings with LLMs in particular, in some cases they're small improvements to an already annoying process, e.g. non-human customer support that was previously done by a crude chatbot front-end to an expert system or knowledge base, the LLM version of that tends to be slightly less annoying.)
Sort of a followup to myself if I come back searching this comment or someone sees this thread later... here's a study that just came out on AI attitudes: https://report2025.seismic.org/
I don't think it supports the bits I quoted, but it does include more negativity than I would have predicted before seeing it.
> “most people agree that the output is trite and unpleasant to consume”
That is a such a wild claim. People like the output of LLMs so much that ChatGPT is the fastest growing app ever. It and other AI apps like Perplexity are now beginning to challenge Google’s search dominance.
Sure, probably not a lot of people would go out and buy a novel or collection of poetry written by ChatGPT. But that doesn’t mean the output is unpleasant to consume. It pretty undeniably produces clear and readable summaries and explanations.
Let's not ignore the technical aspects as well: LLMs are probably a local minima that we've gotten stuck in because of their rapid rise. Other areas in AI are being starved of investment because all of the capital is pouring into LLMs. We might have been better off in the long run if LLMs hadn't been so successful so fast.
> most people agree that the output is trite and unpleasant to consume
This is likely a selection bias: you only notice the obviously bad outputs. I have created plenty of outputs myself that are good/passable -- you are likely surrounded by these types of outputs without noticing.
There are pretty hidden assumption in this comment. First of all, not every business in the AI space is _training_ models, and the difference between training and inference is massive - i.e. most businesses can easily afford inference, perhaps depending on model, but they definitely can.
Another several unfounded claims were made here, but I just wanted to say LLMs with MCP are definitely good enough for almost every use case you can come up with as long as you can provide them with high quality context. LLMs are absolutely the future and they will take over massive parts of our workflow in many industries. Try MCP for yourself and see. There's just no going back.
> There are many technologies that have seemed inevitable and seen retreats under the lack of commensurate business return (the supersonic jetliner)
I think this is a great analogy, not just to the current state of AI, but maybe even computers and the internet in general.
Supersonic transports must've seemed amazing, inevitable, and maybe even obvious to anyone alive at the time of their debut. But hiding under that amazing tech was a whole host of problems that were just not solvable with the technology of the era, let alone a profitable business model. I wonder if computers and the internet are following a similar trajectory to aerospace. Maybe we've basically peaked, and all that's left are optimizations around cost, efficiency, distribution, or convenience.
If you time traveled back to the 1970s and talked to most adults, they would have witnessed aerospace go from loud, smelly, and dangerous prop planes to the 707, 747 and Concorde. They would've witnessed the moon landings and were seeing the development of the Space Shuttle. I bet they would call you crazy if you told this person that 50 years later, in 2025, there would be no more supersonic commercial airliners, commercial aviation would basically look the same except more annoying, and also that we haven't been back to the moon. In the previous 50 years we went from the Wright Brothers to the 707! So maybe in 2075 we'll all be watching documentaries about LLMs (maybe even on our phones or laptops that look basically the same), and reminiscing about the mid-2020s and wondering why what seemed to be such a promising technology disappeared almost entirely.
slower, no fast option, no smoking in the cabins, less leg room, but with TVs plastered on the back of every chair, sometimes
its actually kind of scary to think of a world where generative AI in the cloud goes away due to costs, in favor of some other lesser chimera version that can't currently be predicted
but good news is that locally run generative AI is still getting better and better with fewer and fewer resources consumed to use
> 2. Almost three years in, companies investing in LLMs have not yet discovered a business model that justifies the massive expenditure of training and hosting them,
I always think back to how Bezos and Amazon were railed against for losing money for years. People thought that would never work. And then when he started selling stuff other than books? People I know were like: please, he's desperate.
Someone, somewhere will figure out how to make money off it - just not most people.
My guess is that LLM's are bridge technology, the equivalent of cassette tapes. A big step forward, allowing things that we couldn't before. But before long they'll be surpassed by much better technology, and future generations will look back on them as primitive.
You have top scientists like LeCun arguing this position. I'd imagine all of these companies are desperately searching for the next big paradigm shift, but no one knows when that will be, and until then they need to squeeze everything they can out of LLMs.
Oh wow I forgot that the microwave oven was once marketed as the total replacement of cooking chores and in futuristic life people can just press a button and have a delicious good meal ( well you can now but microwave meals are often seen as worse than fastfood ).
It seems to me from a cursory glance of the blog post that because certain notable humans / individuals are "framing" the modern AI/ML (LLM) era in a more inevitable way, which I totally get, but isn't that how human life works?
The majority of humans will almost always take the path of least resistance, whether it's cognition, work (physics definition), effort. LLMs are just another genie out of the bottle that will enable some certain subset of the population to use the least amount of energy to accomplish certain tasks, whether for good or bad.
Even if we put the original genie back in the bottle, someone else will copy/replicate/rediscover it. Take WhatsApp locked secret passphrase chats as an example - people (correctly) found that it would lead to enabling cheaters. Even if WhatsApp walked it back, someone else would create a new kind of app just for this particular functionality.
The quotes in the post are made by people in an attempt to sound profoundly predictive on some vague super-ai future. Its good to call out that bullshit.
On the other end of the spectrum is that people - demonstrably - like access to the ability to have a computer spew out a (somewhat coherent) relevant suggestion.
The distance between those is enormous. Without a vocabulary to distinguish between those two extremes people are just talking past each other. As demonstrated (again) in this thread.
Consequently one side has to pull out their "you're ignoring reality" card.
All because we currently lack shared ideas and words to express an opinion beyond "AI yes or no?"
In the 90s a friend told me about the internet. And that he knows someone who is in a university and has access to it and can show us. An hour later, we were sitting in front of a computer in that university and watched his friend surfing the web. Clicking on links, receiving pages of text. Faster than one could read. In a nice layout. Even with images. And links to other pages. We were shocked. No printing, no shipping, no waiting. This was the future. It was inevitable.
Yesterday I wanted to rewrite a program to use a large library that would have required me to dive deep down into the documentation or read its code to tackle my use case. As a first try, I just copy+pasted the whole library and my whole program into GPT 4.1 and told it to rewrite it using the library. It succeeded at the first attempt. The rewrite itself was small enough that I could read all code changes in 15 minutes and make a few stylistic changes. Done. Hours of time saved. This is the future. It is inevitable.
PS: Most replies seem to compare my experience to experiences that the responders have with agentic coding, where the developer is iteratively changing the code by chatting with an LLM. I am not doing that. I use a "One prompt one file. No code edits." approach, which I describe here:
For sure; similarly, when someone showed me Prettier many years ago, I immediately understood its value. This will save significant time every year I previously spent manually formatting my code and having arguments with other engineers about tabs versus spaces.
AI bros will probably feel I'm being sarcastic and facetious; but I'm genuinely not. LLMs are an awesome tool to have in the toolbelt. I use them every day. The question is simply on the scope of their capability.
Is this the future of how all code is written? Or is it just the future of how mostly-mechanical refactors happen? Can these systems take extremely abstract prompts and deliver adequate results? Or do they need to be communicated with in a way that so-closely-resembles computer code that one might as well just write the code themselves?
the internet got rid of a lot of waiting. AI gets rid of a lot of cognitive work. the resulting impact on peoples lives will be much more negative with AI. and we have a choice as to whether or not we allow AI to exist
LLM is an almost complete waste of time. Advocates of LLM are not accurately measuring their time and productivity, and comparing that to LLM-free alternative approaches.
207 comments
[ 3.0 ms ] story [ 114 ms ] thread…is exactly inevitablist framing. This claims perfect knowledge of the future based on previous uncertain knowledge of the future (which is now certain). You could have been making the same claims about the inevitability of sporks in the late 19th century and how cutlery drawers should adapt to the inevitable single-utensil future.
I read a story about 14 year olds that are adopting AI boyfriends. They spend 18 hours a day in conversation with chatbots. Their parents are worried because they are withdrawing from school and losing their friends.
I hate second guessing emails that I've read, wondering if my colleagues are even talking to me or if they are using AI. I hate the idea that AI will replace my job.
Even if it unlocks "economic value" -- what does that even mean? We'll live in fucking blade runner but at least we'll all have a ton of money?
I agree, nobody asked what I wanted. But if they did I'd tell them, I don't want it, I don't want any of it.
Excuse me, I'll go outside now and play with my dogs and stare at a tree.
Just because you don't like it, it doesn't mean it's not going to happen.
Observe the world without prejudice. Think rationally without prejudice.
Somewhat objective proof of "progress" will inevitably win out, yes inevitable framing might help sell the vision a bit, for now, but it won't be the inevitabism that causes it to succeed but its inherit value towards "progress".
The definition of "progress" being endlessly more productive humans at the cost of everything else.
I’d have thought perhaps we’d learn the lessons of eg. smart phones, social media, cloud, VR, crypto, NFTs, etc, and think a little more deeply about where and how we want to go as a society and species beyond just adopting the latest hype.
While I must admit we have some choice here, it is limited. No matter what, there will be models of language, we know how they work, there is no turning back from it.
We might wish many things but one thing we can't do is to revert time to a moment when these discoveries did not exist.
The major change from my perspective is new consumer behavior: people simply enjoy talking to and building with LLMs. This fact alone is generating a lot (1) new spend and (2) content to consume.
The most disappointing outcome of the LLM era would be increasing the amount of fake, meaningless busywork humans have to do just to sift through LLM generated noise just to find signal. And indeed there are probably great products to be built that help you do just that; and there is probably a lot of great signal to be found! But the motion to progress ratio concerns me.
For example, I love Cursor. Especially for boilerplating. But SOTA models with tons of guidance can still not reliably implement features in my larger codebases within the timeframe it would take me to do it myself. Test-time compute and reasoning makes things even slower.
LLMs are literally performing useful functions today and they're not going away. Are they AGI? No, but so what?
There is waaay too much projecting and philosophizing going on in these comments and not enough engineering-minded comments from objective observers.
Is AI hyped? Sure. Are LLMs overshadowing other approaches? Sure. Are LLMs inefficient? Somewhat. Do they have problems like hallucinations? Yes. Do they produce useful output? Yes.
We have created tools to write code with bugs that humans have difficulty spotting. Worse, we estimate the quality of the code that our new tools produce on the basis that they are inhuman and have no special skill at writing bugs that we cannot spot, despite the nature of their training data.
"...the politics of inevitability – a sense that the future is just more of the present, that the laws of progress are known, that there are no alternatives, and therefore nothing really to be done."[0]
[0] https://www.theguardian.com/news/2018/mar/16/vladimir-putin-...
This article in question obviously applied it within the commercial world but at the end it has to do with language that takes away agency.
1. LLMs are a new technology and it's hard to put the genie back in the bottle with that. It's difficult to imagine a future where they don't continue to exist in some form, with all the timesaving benefits and social issues that come with them.
2. Almost three years in, companies investing in LLMs have not yet discovered a business model that justifies the massive expenditure of training and hosting them, the majority of consumer usage is at the free tier, the industry is seeing the first signs of pulling back investments, and model capabilities are plateauing at a level where most people agree that the output is trite and unpleasant to consume.
There are many technologies that have seemed inevitable and seen retreats under the lack of commensurate business return (the supersonic jetliner), and several that seemed poised to displace both old tech and labor but have settled into specific use cases (the microwave oven). Given the lack of a sufficiently profitable business model, it feels as likely as not that LLMs settle somewhere a little less remarkable, and hopefully less annoying, than today's almost universally disliked attempts to cram it everywhere.
I'm genuinely surprised that Code forks and LLM cli things are seemingly the only use case that's approached viability. Even a year ago, I figured there'd be something else that's emerged by now.
What are you basing this on? Personal feelings?
(1) Model capabilities will plateau as training data is exhausted. Some additional gains will be possible by better training, better architectures, more compute, longer context windows or "infinite" context architectures, etc., but there are limits here.
(2) Training on synthetic data beyond a very limited amount will result in overfitting because there is no new information. To some extent you could train models on each other, but that's just an indirect way to consolidate models. Beyond consolidation you'll plateau.
(3) There will be no "takeoff" scenario -- this is sci-fi (in the pejorative sense) because you can't exceed available information. There is no magic way that a brain in a vat can innovate beyond available training data. This includes for humans -- a brain in a vat would quickly go mad and then spiral into a coma-like state. The idea of AI running away is the information-theoretic equivalent of a perpetual motion machine and is impossible. Yudkowski and the rest of the people afraid of this are crackpots, and so are the hype-mongers betting on it.
So I agree that LLMs are real and useful, but the hype and bubble are starting to plateau. The bubble is predicated on the idea that you can just keep going forever.
You hit the nail on why I say to much hatred from "AI Bros" as I call them, when I say it will not take off truly until it runs on your phone effortlessly, because nobody wants to foot a trillion dollar cloud bill.
Give me a fully offline LLM that fits in 2GB of VRAM and lets refine that so it can plug into external APIs and see how much farther we can take things without resorting to burning billions of dollars' worth of GPU compute. I don't care that my answer arrives instantly, if I'm doing the research myself, I want to take my time to get the correct answer anyway.
Since you brought up supersonic jetliners you're probably aware of the startup Boom in Colorado trying to bring it back. We'll see if they succeed. But yes, it would be a strange path, but a possible one, that LLMs kind of go away for a while and try to come back later.
You're going to have to cite some surveys for the "most people agree that the output is trite and unpleasant" and "almost universally disliked attempts to cram it everywhere" claims. There are some very vocal people against LLM flavors of AI, but I don't think they even represent the biggest minority, let alone a majority or near universal opinions. (I personally was bugged by earlier attempts at cramming non-LLM AI into a lot of places, e.g. Salesforce Einstein appeared I think in 2016, and that was mostly just being put off by the cutesy Einstein characterization. I generally don't have the same feelings with LLMs in particular, in some cases they're small improvements to an already annoying process, e.g. non-human customer support that was previously done by a crude chatbot front-end to an expert system or knowledge base, the LLM version of that tends to be slightly less annoying.)
I don't think it supports the bits I quoted, but it does include more negativity than I would have predicted before seeing it.
That is a such a wild claim. People like the output of LLMs so much that ChatGPT is the fastest growing app ever. It and other AI apps like Perplexity are now beginning to challenge Google’s search dominance.
Sure, probably not a lot of people would go out and buy a novel or collection of poetry written by ChatGPT. But that doesn’t mean the output is unpleasant to consume. It pretty undeniably produces clear and readable summaries and explanations.
But have we ever had a general purpose technology (steam engine, electricity) that failed to change society?
This is likely a selection bias: you only notice the obviously bad outputs. I have created plenty of outputs myself that are good/passable -- you are likely surrounded by these types of outputs without noticing.
Not a panacea, but can be useful.
Another several unfounded claims were made here, but I just wanted to say LLMs with MCP are definitely good enough for almost every use case you can come up with as long as you can provide them with high quality context. LLMs are absolutely the future and they will take over massive parts of our workflow in many industries. Try MCP for yourself and see. There's just no going back.
I think this is a great analogy, not just to the current state of AI, but maybe even computers and the internet in general.
Supersonic transports must've seemed amazing, inevitable, and maybe even obvious to anyone alive at the time of their debut. But hiding under that amazing tech was a whole host of problems that were just not solvable with the technology of the era, let alone a profitable business model. I wonder if computers and the internet are following a similar trajectory to aerospace. Maybe we've basically peaked, and all that's left are optimizations around cost, efficiency, distribution, or convenience.
If you time traveled back to the 1970s and talked to most adults, they would have witnessed aerospace go from loud, smelly, and dangerous prop planes to the 707, 747 and Concorde. They would've witnessed the moon landings and were seeing the development of the Space Shuttle. I bet they would call you crazy if you told this person that 50 years later, in 2025, there would be no more supersonic commercial airliners, commercial aviation would basically look the same except more annoying, and also that we haven't been back to the moon. In the previous 50 years we went from the Wright Brothers to the 707! So maybe in 2075 we'll all be watching documentaries about LLMs (maybe even on our phones or laptops that look basically the same), and reminiscing about the mid-2020s and wondering why what seemed to be such a promising technology disappeared almost entirely.
its actually kind of scary to think of a world where generative AI in the cloud goes away due to costs, in favor of some other lesser chimera version that can't currently be predicted
but good news is that locally run generative AI is still getting better and better with fewer and fewer resources consumed to use
I always think back to how Bezos and Amazon were railed against for losing money for years. People thought that would never work. And then when he started selling stuff other than books? People I know were like: please, he's desperate.
Someone, somewhere will figure out how to make money off it - just not most people.
You have top scientists like LeCun arguing this position. I'd imagine all of these companies are desperately searching for the next big paradigm shift, but no one knows when that will be, and until then they need to squeeze everything they can out of LLMs.
Phase 1 - mid to late 1990s:
- "The Internet is going to change EVERYTHING!!!"
Phase 2 - late 1990s to early 2000s:
- "It's amazing and we are all making SO much money!"
- "Oh no! The bubble burst"
- "Of course everyone could see this coming: who is going to buy 40 lb bags of dogfood or their groceries over the Internet?!?!?"
Phase 3 - mid 2000s to 2020:
- "It is astounding the amount of money being by tech companies"
- "Who could have predicted that social media would change the ENTIRE landscape??"
The majority of humans will almost always take the path of least resistance, whether it's cognition, work (physics definition), effort. LLMs are just another genie out of the bottle that will enable some certain subset of the population to use the least amount of energy to accomplish certain tasks, whether for good or bad.
Even if we put the original genie back in the bottle, someone else will copy/replicate/rediscover it. Take WhatsApp locked secret passphrase chats as an example - people (correctly) found that it would lead to enabling cheaters. Even if WhatsApp walked it back, someone else would create a new kind of app just for this particular functionality.
On the other end of the spectrum is that people - demonstrably - like access to the ability to have a computer spew out a (somewhat coherent) relevant suggestion.
The distance between those is enormous. Without a vocabulary to distinguish between those two extremes people are just talking past each other. As demonstrated (again) in this thread.
Consequently one side has to pull out their "you're ignoring reality" card.
All because we currently lack shared ideas and words to express an opinion beyond "AI yes or no?"
Yesterday I wanted to rewrite a program to use a large library that would have required me to dive deep down into the documentation or read its code to tackle my use case. As a first try, I just copy+pasted the whole library and my whole program into GPT 4.1 and told it to rewrite it using the library. It succeeded at the first attempt. The rewrite itself was small enough that I could read all code changes in 15 minutes and make a few stylistic changes. Done. Hours of time saved. This is the future. It is inevitable.
PS: Most replies seem to compare my experience to experiences that the responders have with agentic coding, where the developer is iteratively changing the code by chatting with an LLM. I am not doing that. I use a "One prompt one file. No code edits." approach, which I describe here:
https://www.gibney.org/prompt_coding
AI bros will probably feel I'm being sarcastic and facetious; but I'm genuinely not. LLMs are an awesome tool to have in the toolbelt. I use them every day. The question is simply on the scope of their capability.
Is this the future of how all code is written? Or is it just the future of how mostly-mechanical refactors happen? Can these systems take extremely abstract prompts and deliver adequate results? Or do they need to be communicated with in a way that so-closely-resembles computer code that one might as well just write the code themselves?