I’m calling it now, the adoption of AI agents into software development will be one of the most costly mistakes in the field’s history. Agents cannot program, and it’s taking longer and longer to realize that they can’t. They are a highly sophisticated statistical model designed to mimic the distribution of programming. The output is broken, but in a way that’s getting harder and harder to detect. Which is exactly what you’d expect from an increasingly accurate statistical model.
geohot
Now there’s a name I haven’t heard in a long time. George Hotz was the guy who first jailbroke iOS and the PlayStation 3 and made the towelroot exploit for early versions of Android, before legal threats drove him out of the scene.
Calling them a statistical model is misrepresenting their functional understanding of concepts.
This guy does not get it.
He’s bang on. LLMs have no understanding of anything: they are literally just statistical models.
remember, when you interview for a job and they ask you, “do you have any questions?”, you ask;
- has AI ever been used to develop your product?
- what percentage of your product has been written by agenetic AI?
- is the use of AI tracked as a performance indicator?
It’s going to be a wonderful time to be a Freelance Senior Developer and above in a few years.
They are not the automated from 0 to 100 coders that some people claim them to be. But they are quite capable, definitely much more capable than what anyone could have imagined ten years ago. Given well defined problems they can excel at even relatively complex tasks. I pointed Claude at a latex file of a somewhat complicated nonparametric statistical estimate calculation to look for any mistakes and it was actually able to find some. I then pointed it at a code that replicates the calculations and it was also able to correctly identify some issues with the code. I think this is the way one should use LLMs, not let it loose on coding tasks. In the former way you won’t even be able to burn through your first tier account quota where as in the latter the LLM will likely end up getting in weird loops burning tokens like there is no tomorrow. Also this method of sane usage of LLMs is much more suitable for open local LLMs. I don’t think there is any doubt anymore that LLMs can be very useful tools, not just for doing stuff but learning it too. People should move past the stage of invalid criticisms like “they are just stochastic parrots” and move to more serious matters like environmental impact, greedy fucking CEOs pretending LLMs are replacements for humans, degredation of skills, getting lazy at checking AI code, ethics of capitalizing on collective human knowledge and the unsustainable AI bubble that tech companies are pushing for.
invalid criticisms like “they are just stochastic parrots”
That’s not a criticism per se, it’s a description of how they work.
Sure, and at that level of accuracy it’s also a description of how humans work. I didn’t invent these words myself, I’m just stringing them together based on a stochastic process my brain was trained into.
Like LLMs, some of my speech is semi-random initialization (dada wawa googoo), some of that is mimicry (some of that is mimicry), some of that is reinforcement learning (downvotes incoming), and some of that is the output of a subprocess that uses the same systems prompted at the meta-level and without verbalization (maybe they won’t get the analogy between thinking and LLM scratchpads… how about I use this space to clarify).
Calling an LLM a stochastic parrot has the same social-emotional role as calling a human an animal. Yes, it is correct. But people can infer the connotation.
Humans are animals. LLMs randomly generate text based on the corpus they were trained on and the conversation so far, so stochastic parrot is an accurate description.
LLMs don’t learn. Humans do. LLMs generate text randomly using a massive matrix. Humans don’t; you lied. An LLM is incapable of lying because it has no understanding of truth. It just bullshits convincingly all the time. It’s very very good at it, but it’s all hallucinated for the LLM, true or false.
Expecting your random word generator to tell you truths is insane. The training measure is “sounds right” not “is right”. It passes if it sounds like the other discourse it read. Just like the confident drunk guy at the pub who thinks he knows everything passes of he convinces the other drunk guys at the pub.
Whereas humans learn at school and on the job and the training measure is “your teacher or supervisor approves”. LLMs were not trained on truth or accuracy. Trusting in them and treating them as equivalent to human intelligence, as you and a whole bunch of other folks do, is profoundly unsound, and soon the necessary price rises to pay for the processing costs (let alone the vast, vast, vast, vast, vast debts on the infrastructure) are going to make most slophouses which jettisoned their human talent go out of business. And very, very few people indeed will be sorry at that point.
Meanwhile LLM slop is shitting in github all day long, every day, and shitting on the internet, and it will eat it’s own shit and produce crappier shit.
Your analogies don’t change the truth, and that is that LLMs don’t know the difference between sounds correct and is correct any more than MAGA voters know the difference between sounds good to me and is good for me.
What do you mean LLMs don’t learn? How do you think they became capable of stringing a sentence together?
They don’t learn during a deployment, but neither do humans; humans only learn during sleep. The behaviors a human exhibits while “learning” in the moment are just stochastic parrot behaviors based on their immediate context window, if the human doesn’t sleep in time the event can slip out of their context window and they don’t learn despite having acted as if they do.
You seem to be very naive about human learning in general. What makes the “truth” of school lessons greater than the “truth” of an LLM’s curated dataset it is reinforcement learned on? Have you ever seen actual evidence that mitochondria exist, or are you just stochastically parroting your biology teacher?
I also oppose LLMs in almost all applications (live translation being an example of a good application). But please oppose it with arguments based in reality.
What do you mean LLMs don’t learn? How do you think they became capable of stringing a sentence together?
You’re confusing constructing the LLM, which is done with an actual AI (neural network) and a massive corpus of text (stolen from millions of humans in the greatest intellectual theft in history) and running the LLM, which is done with a random number generator and a massive matrix of probable next words.
They don’t learn during a deployment,
They don’t learn. They don’t change. They’re as random next time as this time.
but neither do humans; humans only learn during sleep.
False and false. Soooo much pseudoscience.
The behaviors a human exhibits while “learning” in the moment are just stochastic parrot behaviors
Wrong again.
if the human doesn’t sleep in time the event can slip out of their context window and they don’t learn despite having acted as if they do.
If that were true, most people would learn very badly first thing in the morning and get better and better later in the day. I think you’ll find that most school teachers would vehemently disagree with your nonsense conclusions.
Then again, perhaps by “doesn’t sleep in time” you mean stays up all night, then admittedly they might function less well cognitively but (a) we tend not to regularly torture humans that way and (b) you’re massively overstating the role of sleep in the learning process.
You seem to be very naive about human learning in general.
No, you seem to be very naive indeed, to extremes, about the intelligence and reliability of LLMs. When I ask them about general things that I know about, I tend to get the right answer about 60%-70% of the time. Why would I believe it when I didn’t know the answer. To trust an LLM to tell you the truth about stuff you aren’t checking when it clearly blags nonsense so frequently when you are is really really stupid.
What makes the “truth” of school lessons greater than the “truth” of an LLM’s curated dataset it is reinforcement learned on?
Most teachers tend to consistently teach the content of the syllabus rather than randomise what they say to classes based on the preceding conversation. They reinforce and update their prior knowledge by also learning from the mark schemes of the tests and exams their students sit.
Have you ever seen actual evidence that mitochondria exist, or are you just stochastically parroting your biology teacher?
No. I trust my teachers. I am rational to do so. I don’t trust LLMs. You are irrational to do so.
But please oppose it with arguments based in reality.
You are utterly deluded and have bought the hype. You seem unable to distinguish between distinct things and are dismissing a large amount of evidence that your “just as good as a human” is a crap-spewing shit machine, no more honest than donald J trump, and with no less sharting.
This is very obvious unless you are in tech leadership, in which case your job is now to push this at all costs and suppress dissenting voices.
In tech leadership. I don’t have to push it. My talented engineers took to it immediately.
They learned quickly that it is a tool. Instead of using a shovel and a wheelbarrow, they have a backhoe now. If you don’t know how to dig a hole, the backhoe is just a way to make a mess faster. It doesn’t replace intelligence.
They can use it to do the scutwork while they focus on the important stuff.
The duds are still typing shit into spreadsheets and emailing them as attachments while their coworkers are getting stuff done.
It is a tool. You can learn to use it or you can just be mad that it exists. In either case it isn’t going away. Like the telephone, the car, the computer, the internet, it is here to stay.
What’s fascinating about this conversation is … how do people think software used to be made ? With talented and knowledgeable developers who would never “hallucinate” an API or a library function ? With cybersec experts who would never put their user’s data in jeopardy ? With performance investigators checking the computational complexity of each function ? Bitch please…
Software engineering is not the kind of mystical cathedral building these people have in mind, it’s more like a musty workshop in Pakistan where they make tractor tires with no safety equipment and a cigarette in their mouth. We’ve been throwing imperfect humans in various states of lucidity at every problem known to man for 30 years but suddenly people start believing that their bog standard CRUD software should be written by monks having attained cosmic godhood.
If you’re letting your engineers find uses for it instead of constantly demanding that they generate lengthy “user stories” and decision documents and deferring thinking to agents instead of quickly planning stuff out using their experience then you’re probably quite an outlier by now.
I am a very lazy man, micromanagement takes so much fucking energy. Hire talented people, give clear and unambiguous guidance and trust them to do the work. It is amazing how easy management can be when you don’t get in the way.
Being in leadership is way easier than being an engineer and the pay is better too. Some people really overcomplicate shit.
part of the issue as well is that when they get something completely broken, people just re roll the output until they get something that’s broken in ways they don’t notice. Or re roll parts of it, or tell the system to judge if the output is broken and re roll the parts that it judges are broken automatically. Or increase the size of the context window to get it closer to that upper limit of accuracy.
All this together can get a more functional output with less effort, and as people find these tricks it gives them the illusion of an upward trend in capability, like this is all solvable issues that will improve as time goes on. Big problem with that though, theses tricks and methods explode the compute cost rapidly. That’s all fine and dandy when everyone is getting their compute costs for these tools subsidized by these model providers, but eventually they will need to charge the real cost of running this. The compute providers that host the model providers are also running at a loss, trying to help grow the market segment and maximize their market share. And then places that have the datacenters in them are giving tax breaks and discount utilities to attract new construction.
Everyone except the people making the chips is selling at a loss, and as people pile on usage to make up for the fundamental limitations of these systems, the demand balloons, validating to the providers at all levels that this is a growing market they should invest more in to.
But eventually… they need to make money. The bill comes due on all the debt and investment. What happens to the people who have fully embraced these to run their businesses? Or to all the people who have built their skill set around using these systems? It’s a crisis, a series of crisis, each time a debt wall gets hit by someone in the supply chain. A half decade of technical debt that just got really expensive to deal with, and not enough experienced people to handle it, since all the grey beared retired and not enough new people got brought in to replace them because the entry level work was automated.
This alarm’s being rung for over a year now, so “calling it now” means finally reading the writing on the wall
Let it be known that the first person to call it was actually Sam Altman when OpenAI’s paper on AI Scaling Laws in 2020 subtly showed that the diminishing returns will stop showing improvement with infinite power, compute time, and data before 94% accuracy is reached.
yeah I was kinda like. calling it?
You know the feeling that you want to rewrite a project? But you know that most rewrites are a bad idea.
Be it your own, old code. Or code you inherited.
There is a small chance that the world realizes that they went in the wrong direction and nothing can get fixed. That will be the time of rewrites.
No, I don’t expect this to be very likely. The agent code will remain, and human programmers get yelled at for not fixing it fast enough.
if it’s broken in a way that can’t be detected, is it actually broken?
all software is broken in some way. if the rate of bugs generated by llm and the severity of those bugs drops below the rate you would expect from a human programming team, then llm is offering something competitive.
It will eventually be detected, but it passes tests before hitting production, that is the problem.
broken in a way that can’t be detected
Is not what anyone said and you’re lying when you pretend they did.
deleted by creator
No, humans make less mistakes. Less. That’s the key here, statistical models are trained on human data so by pure logic can never, ever, under any circuimstance, reach 100% accuracy. With current understanding of LLMs with a focus on AI Scaling Laws, and more importantly of natural human language adaptation, they will never reach 94% accuracy with infinite power and infinite training. That’s what the curve shows us in OpenAI’s 2020 research paper on AI Scaling Laws and later Deepmind’s paper correcting their math, that the diminishing returns will hit a limit far before convergence.
In addition to that, the AI also cannot detect subtle changes to established problems or any new unaccounted for variables, because they’re a statistical model and not capable of actual thought. They also lack any sense of responsibility for their actions for the same reason.
You fucking sloppers always try to say “HuMAnS mAkE misTAKeS, TOO!” Yeah and the fucking slopbots are trained on those mistakes and make them again but worse.
But you’re forgetting the key difference that makes it so much worse - we can fix human mistakes especially if we can talk to the human to figure out how. With an llm we have no external reference, only poorly designed code where the comments are there to guide the writing, not describe what was written. So it’s much harder to debug an output, and the llm cannot be trusted to clean it up either.
A human can be held responsible. A machine cannot. If the machine writes bad code, and someone gets injured or killed because of it, who takes responsibility?
I state again: a machine cannot be held responsible.
It is never the coder that is responsible, it is the one who makes the code available to use. Often with humans, they are one and the same. With machines, they are not.
You can totally fix AI-written code with AI. You tell it something is wrong, it tries to fix it.
I did a recent experiment with AI writing a document format converter and that’s exactly what I did. It wrote some code, I checked the output, found a formatting issue or similar, asked it to fix it, repeat. It works unreasonably well and with Fable the final code isn’t even bad.
Now keep doing this for months, with non-trivial software that other people use.
Probably fine if you review the code carefully. And if you’re working in a domain that AI is decent at (e.g. web stuff). But even if it wasn’t it doesn’t mean AI cannot program.
You can fix problems, if you know they are there and there is a model of that problem being fixed.
You can’t fix problems you don’t know are there, or do not have modeling.
That’s true for humans too surely? How would anyone fix problems they don’t know exist?
Humans generally don’t hallucinate libraries or documentation. If there is a bug or error on a human maintaine REPO the human in charge will generally know what went wrong and how to fix it, the AI will just gaslight your ass because the AI has no idea.
This is true. But it doesn’t invalidate any of my points. Humans have unique failure modes too.
To add to the other response - it is much more difficult to work with Ai to debug inconsistent issues or similar unless you can understand the code and step through with a debugger to check for race conditions or similar.
Recently I was working with an Ai tool for some c code that depending on machine ran wildly differently. The Ai was unable to identify any issues, and kept recommending fixes for hardcoding values or similar that I had to revert. The fix ended up needing to use valgrind to create a different enough environment to see how a race condition was made to properly have one async call delay for the other.
AI can be powerful, and humans can be dumb. But if the code was human made, I would not have needed 3 hours to find a problem, and I wouldn’t have tried to turn to AI for a simple fix because I’d know what I was looking for to start with.
Also, unlike say, writing articles, like 75% of coding is copy.pasting blindly from elsewhere anyway.
Not in my experience.
It’s so nice to see some people speaking reason. If only any of those people ran multibillion dollar companies.
Let them fail and then scramble to rehire.
No one paying the bills cares
Agents cannot program
This is just factually incorrect. Difficult to get past a false assertion of this magnitude.
They are a highly sophisticated statistical model designed to mimic the distribution of programming.
I thought we had got over the stochastic parrot nonsense by now.
You can totally have objections about the ability of AI to program - how good it is, poor failure modes, high cost, technical debt, knowledge debt, broken social contracts, etc. All valid.
But if you’re still in the “It’s just a next word predictor! It can’t really think!” stage of denial, even now… Sorry you’re an idiot.
Um… but it is just a sophisticated statistical model… that’s literally what the math underpinning machine learning models is… and all it can do is make associations based on correlations within the field of the training data. That’s what it does.
This is like saying “but it is just a sophisticated network of neurons. All it can do is transform input signals into output signals. That’s what it does.”
Not really.
A machine learning model is a computer program. It is fundamentally a math equation, which we understand completely.
A living brain is not fundamentally a math equation, and is not purely a statistical model, at least not in any empirically demonstrable way. We don’t understand completely how it works, but we do know that it’s more complex than what you’re trying to imply.
The comparison is not valid. Machine learning models are not an equivalent to a biological brain.
A machine learning model is a computer program. It is fundamentally a math equation, which we understand completely.
Lol you couldn’t be more wrong about this. One of the most widely commented things about DNNs is that we don’t really understand them completely. I don’t know how you would miss that if you knew anything about AI at all.
A living brain is not fundamentally a math equation
It is. A very complicated one, sure. Which part of the brain do you think is impossible to simulate with maths?
Tbf, LLMs may be comparable in complexity to their specific brain.
Um… but it is just a sophisticated statistical model…
The mistake has been thinking this implies LLMs can never do X task, and using it as a catch-all argument for any value of X, but it isn’t a good argument because it has been wrong for most of those.
The mistake has been thinking this implies LLMs can never do X task
As this article points out, an LLM can spit out chunks of regurgitated code that it scraped from the internet, but that does not make the LLM a programmer. The resulting output is an attempt to find an existing pattern in the database which fits with what the user has asked for, but it is not a product of actually understanding the use case for the code. It is just statistical correlation.
So, sure, an LLM can be set up to generate output related to X task. If you can collect and clean data that can be used to train the kind of output you want, it should be able to produce an approximate facsimile of the results you want. Is that valuable for your use case? Maybe.
We’re still just talking about what is essentially a complex search function. The statistical model returns results from its database that correlate most closely to your input. That does not mean it returns the right answer. If there is no good correlation, it will still return a result.
As long as you understand that the result you get is just a correlation based on your input and may or may not be relevant to your specific problem, and you are not fooled into believing that the LLM actually understands what you’re asking and produced a result by “thinking” about it, then you might be able to use an LLM as an effective tool - to search a large collection of information for something that is relevant(ish) to what you’re asking for.
The real mistake has been broad misunderstanding of what LLMs actually do, and trying to use them as general-purpose problem solving tools (or worse, as accurate and reliable sources of information).
Language models are not databases and they are not markov bots (similar function but work directly using statistical word association maps). The big difference is that those things are algorithms someone wrote and can fully comprehend what they do, but machine learning models are large algorithms built by another algorithm processing training data. There is much more uncertainty about what is going on under the hood.
There is also great uncertainty about what concepts like understanding or thinking might mean in computer science terms. The main thing we can really know is that ultimately a human mind is a computer, which means that understanding and thinking have some yet unknown mathematical representation, and therefore a comparison can be made. We should eventually be able to quantify whether or to what extent a given algorithm thinks. But you said in another comment that you don’t believe minds can be represented mathematically; this should mean that such comparisons would be apples to oranges, but you’re making them anyway for some reason, and implying they have predictive power for the limitations of LLMs.
Certainly they do have limitations, at least individually and possibly as a technology. There are things given models are bad at, there are things they initially seem to be able to do well as humans but fail in different ways that suggest over-reliance on pattern matching. But these have been determined empirically through testing. The idea that they are “just statistical models” and this knowledge can be used to say what is impossible for them from philosophical first principles keeps getting repeated but has never worked in practice. The reality is that no one knows enough to say for sure where the line is.
Thank you for letting us know where you stand. Those who tag users can act accordingly to this public declaration of robo-handjob-giving.
Our apologies. Let us issue a correction:
“Agents should never be allowed to program”
There.
