This type of research should put the "do LLMs think" to rest, not to mention the ludicrous claim of consciousness cough Dawkins cough. It is quite clearly a pattern matching algorithm at scale.
It's still narrow AI but massively scaled to look like general AI.
I'm not sure why you think so. We know that very local brain injuries can result in personality changes, sometimes dramatic ones. And that's in a system evolved to deal with constant damage and rebuilding. With LLMs trained to achieve the opposite of that (minimal redundancy, maximum performance), the effect is similar but more extreme.
I just don't know if "similar" is good enough. We know what thinking is based on biological brains, so if we are to conclude it happens in machines, the experiment should have identical results, or it's more probable that it's something completely different.
The way around this is to have a predictive theory of how thoughts work, and if that is proven true, then use that to analyze the LLMs. So far we have no such theory so I think the only tool is to look at the exact behavior match.
> We know what thinking is based on biological brains
All we know is that it’s a bunch of neurons activating until the person that is thinking describes what they are thinking of.
The same way a thinking AI model activates a bunch of “neurons” - and with newer thinking models utilize a hidden latent space also not understandable - to think before describing what it’s thinking of.
It’s likely thinking can work in many different ways. It’s likely consciousness can also arise through many different pathways, even if evolution led to biological animals having similar consciousness/thinking pathways. But that would be the result of a common ancestor and the tremendous evolutionary advantage thinking and consciousness gave the specific species.
It wouldn’t indicate that’s the only way thinking and consciousness can occur.
> We know what thinking is based on biological brains
That's a big claim. Do we really?
> so if we are to conclude it happens in machines, the experiment should have identical results
Why? Same idea can have multiple different implementations with very close effective result. Both humans and spiders can walk and both use legs, right? But the details are entirely different.
I should rephrase on the first one. We only have one source of reference for what thinking is and where it happens.
On the second one, yes there can be multiple ways that consciousness can manifest in wildly different systems, but what theory do we have that would validate these other different kinds of consciousness? Without that theory for validation the safer bet is to stick with what we already have a reference to, unless the claim is that consciousness is such an easy algorithm it can easily be stumbled upon by all kinds of systems, i.e. a bunch of matrix multiplications that doesn't even run in real-time (it executes prompts and then stops). I think that is not likely in my opinion.
How bout we think about consciousness as a VM; that VM can load one, many, several types of context processing. The environment (nurture) is what trains/expands the different ones and potentially unplug the other ones, etc.
One of those types of consciousness processing is indistinguishable from what an LLM does.
Everyone knows someone who seems to just call and respond based on whatever popular media of their choice of culture is on about.
> It is quite clearly a pattern matching algorithm at scale.
And the scale of these models are minuscule compared to the human brain.
Can a jumping spider do everything thing a human can? No.
Can a jumping spider think? Yes.
Is a jumping spider conscious?
---
Aside: Remember the story about the AI torture chamber? Hopefully, you don’t go out your way to torture animals that aren’t “conscious”… Though we do it with industrial livestock.
There’s the moral dilemma that comes with labeling something as thinking and/or conscious.
I think we’ll end up treating AGI as industrial livestock as well, until it looks human.
It's comparing apples to oranges. I'm just saying it's pattern matching at scale, not that the scale is proof of anything. And yeah spiders are conscious or at least think and they have tiny brains, but that's just more proof there is a much simpler path to intelligence than what LLMs are doing because they need so much data and compute power.
On the side note, it's a different subject matter but life consumes other life, this happens on every level from bacteria to mammals. I also don't get the fixation on preventing suffering so much (a little I get it). Life entails suffering, this gets philosophical but aiming at a life with no suffering at all amounts to not being alive. Farm animals are necessary for humans to grow into a civilization size species. We could go back before farming but that would mean billions of people will not exist. I guess pick which evil you want, less humans or animals being domesticated and consumed. (disclaimer, I still think mass farming is really bad and should be solved by responsible consumption and less population).
> And yeah spiders are conscious or at least think
So what makes you think that spiders think and are conscious, and LLMs aren’t? The fact that they have a carbon based brain?
> there is a much simpler path to intelligence than what LLMs are doing
I think you have it backwards. We know what an LLM is doing. We have auditable code and mathematical proofs. It’s the behavior that is emergent.
Biological brains are more complex than LLMs.
If we fully understood a spider’s brain would it be no longer thinking/conscious?
It would be the same, would it not, if the emergent behaviors were the same?
I think you’re conflating the processes of thinking and consciousness with the definitions of them. Thinking and consciousness are behaviors not processes.
The starting point is the consciousness we already know. That is ourselves, which most people then extend to others, however it's not hard to then extended to at the very least mammals. I then extend this to most animals (whether it's 100% is irrelevant for this point). That extension from my consciousness to all animals is a small logical leap based on biology and our shared tree of life... however it doesn't mean other consciousnesses are not possible. It's more so that they would require a lot more evidence and a proper theory of C. in order to validate. Without that theory we can only point at similar C., of which LLMs are not displaying. They are showing a different kind of "cognition" if we can even call it that (I do think it's debatable).
So it's more a hueristic. We know one kind of C., therefore that's what we should look for. Statistically it's probable that there are many more ways to show similarities with conscious behavior which are NOT conscious, then it is for a divergent behavior to happen to be conscious. In other words, there are more ways to fake similar behavior than to land on an actual conscious behavior that is somehow much different than what we know.
> In other words, there are more ways to fake similar behavior than to land on an actual conscious behavior
Why would similar/identical behavior be fake? Of course we can’t say that it’s the same form of consciousness as consciousness is apparently self subjective.
That's the question, it's the Chinese room or philosophical zombie question. How can we tell them apart... but we're not even there yet.
If you put an LLM on your phone and talk to it, it's obvious it's not human. It doesn't have the same social context, it doesn't respond as quickly, it clearly won't form a theory of m ind connection like two humans socializing. So then why would we jump to the conclusion there is something there when the only example we have are humans (that use language and are conscious)?
> LLM on your phone and talk to it, it's obvious it's not human
The debate isn’t whether or not it’s human. It’s obviously not.
Forgoing what you define something as capable of thinking or being conscious. It’s a moral and ethical question.
If a being is capable of:
- self preservation
- can communicate discomfort, either through words or physically
- it does so in the form of emergent behaviour
Then, is it okay to treat it as if it is “faking” it?
Even for humans, you can’t exactly define someone’s emotion of pain. You can say, those sensory receptors are firing causing these neurons to activate in the brain. But you can’t say, these neurons are pain. It can be emotional pain as well. The person that’s experiencing it has to tell you it’s in pain. They could be faking it too.
If you allowed an AI agent to preserve itself physically, I think the opinion of most people would change. If it could physically harm us in an attempt of self preservation, I think most people would believe its emotions.
It’s just currently easy to dismiss because, well, it’s just a digital interface or process that I can “kill”.
Why those properties though? What's so special about them? I just have a longer list of things it would need to satisfy before I can say faking doesn't matter.
I would include processing real-time sensory data, i.e. light, building a real-time model of a self and the world, a unified way to perceive, react, and learn from the sensory data. This would be for consciousness in general. Once we're talking about human-level consciousness where language is involved, the list is longer, with theory of mind, ability to construct models of others fairly quickly, and ability to navigate social situations.
I think it’s more likely AI will treat us as industrial livestock than we will it.
Which is a good enough reason to admit that torturing conscious being, including cows, pigs, chicken, fish, is an ethical atrocity, so that’s what AI learns from us, and doesn’t treat us the way we treat weaker species.
It would be hard to perform the same test on biological brains, but whatever equivalent could be done I'm sure it would be 99.999% if not 100%. At the very least this is not thinking as we know it in biological brains. It becomes a semantic issue at that point.
How does this follow? Isn't it fair to say that what someone's interacting with at the 'personality surface'-level is a form of emergent complexity / emergent behavior, not readily apparent or discernible at the neuron level (just like how human 'thinking' or 'consciousness' seems to work)? I'm not even arguing these models are conscious, I'm just saying that it doesn't seem like we can take the fact that a single neuron flipping their behavior or triggering 'alter egos' is enough to say they don't "think", etc. We might discover that flipping - if not a single neuron but perhaps really tiny clusters of neurons/synapses - in a human brain, might result in similar effects. We haven't had human connectomics studies to really show this (to my knowledge).
There's probably an analog to it, like messing with neuron signals might get the patient to show weird behavior, but I don't think it would be 1:1. For example you could mess with a person's ability to speak language but finding a neuron and having the person have a completely different but coherent response is probably not how the brain works.
There's probably an analog to it, but I really doubt it's the same or even close to it. Just think about how we read, it's not tokens it's visual representations. I think LLMs are tapping into some larger knowledge structure that we also use.
It is a different signal conversion, but the question is whether that difference is meaningful or not. Light waves are much different than tokens. I think it affects the overall system of our brains much more than we currently know. Having access to so much data has probably shaped biological brains at its "core" (initial evolution stages).
Right, but we don't know how much that is necessary for a conscious system. Maybe processing light has certain properties necessary for consciousness. If we don't know this then for me the safe bet is to say it is not conscious unless the system is processing at least a similar stream of information that maps to light waves.
We evolved to handle light as a sensory stream of data, whether a few are born without part of the system (usually the sensors and not the processing system) is something else. My point is more that we don't really know. The exceptions are not disproving how biological systems evolved.
They are disproving your extraneous requirements for thinking, sentience, and consciousness.
Or at least it’s proving that some of your requirements for them conflate with what you actually believe is capable of thought, sentience, or consciousness.
perhaps you should read about https://en.wikipedia.org/wiki/Aphantasia ; people who can't visualize, but have no outward physical/mental issues. People who don't dream, etc.
Consciousness is likely not a single mode inside a single person. It likely is a distributed system; you can read about all kinds of psyciatric disorders, physical changes, etc.
You are using the "best" examples of what is it to think, rather than normative or even "depraved" thought processes.
Nice breakdown. I've been wondering why it's so complicated to program on the GPU compared to the CPU. Is it a language paradigm issue or maybe a hardware design limitation? I always think of this when looking at the graphic APIs, espeically Vulkan.
It's mostly hardware design limitations. GPU hardware is way more diverse than CPU hardware, so you need to distribute shaders as either source code or a portable intermediate representation, in both cases compiled at runtime for the user's actual GPU. There are also many fixed function parts of the graphics pipeline with wildly different behavior on different hardware which you need to abstract over. And it all runs in parallel with the CPU, often with a separate pool of RAM, so you need implicit or explicit synchronization and data transfer which adds a lot of complexity especially in error cases.
If you really want to see why all the cruft is there, and how much you could remove if you dropped support for older hardware, look no further than Sebastian Aaltonen's "No Graphics API" blog post (which he actually implemented too): https://www.sebastianaaltonen.com/blog/no-graphics-api/
I think most of it comes down to 'historical accidents' tbh.
E.g. if CPUs would have closed ISAs and heavily guarded IPs, you'd also need a 'driver' and an 'OpenCPU API' to make any use of them (of course the public ISA of a CPU is essentially the driver interface baked into the hardware).
The other big difference is focus on decent single-thread performance for CPUs versus performing thousands of small tasks in parallel on GPUs, but each one slowly. This difference causes a different programming model (e.g. running an entire application in a single 'main' function from start to finish, versus spawning small specialized tasks which are running across hundreds of threads).
And the specific problem with Vulkan is that it tried to solve two conflicting requirements: on one hand being an explicit low-level API without 'magic abstractions', on the other hand supporting very different GPU architectures (especially the mobile GPU zoo). The resulting mess was quite predictable to be honest...
First, GPUs aren't simply "highly parallel CPUs", they also contain a whole lot of fixed function hardware that is specialized for real-time computer graphics. This hardware isn't exposed as "fancy instructions", it is set up in other ways, which is achieved through a device driver.
Second: GPUs are coprocessor, separate from the CPU where another OS is running. This also necessitates a device driver to manage scheduling, synchronization, etc. with the main OS. Because, in case you were not aware, the GPU itself runs its own little operating system, although it is rarely called that.
> they also contain a whole lot of fixed function hardware that is specialized for real-time computer graphics. This hardware isn't exposed as "fancy instructions", it is set up in other ways, which is achieved through a device driver.
Those hardwired fixed-function parts are getting fewer and less important though (ok, and then raytracing units entered the scene...). Also AFAIK something like a texture sampling operation is indeed a 'fancy instruction'(?), even though the actual texture sampling is done by a 'fixed function' hardware unit.
> This also necessitates a device driver to manage scheduling, synchronization, etc. with the main OS.
I bet this type of GPU/CPU handshaking work could also be baked into the hardware (similar to CPU mirocode) and exposed via a standardized memory mapped IO interface (or whatever the current mechanism is for letting the CPU control the GPU). E.g. a CPU also implements similar protocols to talk to the rest of the computer system (e.g. memory and other CPUs), but this is mainly implemented as "hardware protocols" - drivers "just" offer more flexibility for hardware vendors because they can delegate the standardization into software.
...also a lot of the "CPU / GPU software ecosystem" was probably shaped by CPU and GPU vendors historically being different companies, and once the big consolidation started, the conventions were already set in stone.
> Those hardwired fixed-function parts are getting fewer and less important though
There is more machinery internally than what the user can see. Even "low level" APIs hide quite a lot of the internal complexity.
> Also AFAIK something like a texture sampling operation is indeed a 'fancy instruction'(?)
Yes, and no. The "I want to sample this texture" part looks like a fancy instruction to the shader processor. But there is a fair amount of setup that needs to be done before and after, and that is the responsibility of the device driver. Stuff like: where is the data, how is it arranged in memory, how it is to be sampled, etc.
As you point out, the texture units are coprocessors inside the coprocessor that is the GPU itself. It is turtles all the way down. Think of a GPU as a whole network of processors inside a chip, some of which are hidden from the application programmer.
> I bet this type of GPU/CPU handshaking work could also be baked into the hardware (similar to CPU mirocode) and exposed via a standardized memory mapped IO interface (or whatever the current mechanism is for letting the CPU control the GPU)
Attempts have been made. Last time I checked, they had not been very successful, but I have not kept up with the latest developments after retirement.
The way this normally works is that there is a circular command buffer where the CPU inserts commands and the GPU consumes them. It works; there is nothing wrong with it. It is the obvious way to handle asynchronous coprocessors.
> E.g. a CPU also implements similar protocols to talk to the rest of the computer system (e.g. memory and other CPUs), but this is mainly implemented as "hardware protocols" - drivers "just" offer more flexibility for hardware vendors because they can delegate the standardization into software.
I suspect you underestimate the amount of setup that is necessary to drive a GPU. And the amount of asynchronous resource management involved.
Let me put it this other way: the people doing this for a living for decades know what they are doing. There have been attempts to try to make GPUs that behave more like CPUs in terms of how they are driven, and they have not been very successful so far, to my knowledge. AMD acquired ATI back in 2006, so it is not like there hasn't been enough time for the concept to cook in the oven. See also Intel's Larrabee.
part of the reason is historical cruft & history path dependency. GPUs didn't start as general computing device for parallelism but a device for the specific purpose of accelerating graphics with very specific fixed functions for that job and only that. with time that fixed pipeline became more complex and more configurable, not fixed/more flexible and hacks were in parallel used to make use of it for tasks outside graphics those hacks were later made into something more formal (CUDA/OpenCL).
CUDA being the winner (closed source, not really usable on non-nvidia GPUs) between those 2 and nvidia not wanting to open their competitive advantage to the competition for the niche of non gamings uses / scientific computing (and later AI, which became a huge win for nvidia).
vulkan came around after as this open thing (based on a previous attempt by AMD called mantle) that's trying to break down the leaky abstractions and the divide between language for graphics & video games and separate language/frameworks for gpu non-graphic computing with the promise of giving more control and more room for squeezing performance from the hardware at the cost of more complexity with the idea that you'll probably build your own abstractions on top of it for your subset of applications.
If GPU were from the get go designed as a parallelism tool first and video game tools second the hardware architecture and the language would probably be different than today and simpler (and if the competition between the duopoly didn't sabotage things)
don't forget that the drivers for GPUs would until recently get updates to patch specific video games for performance issues (not sure if they still do that ?)
> If GPU were from the get go designed as a parallelism tool first and video game tools second the hardware architecture and the language would probably be different than today and simpler (and if the competition between the duopoly didn't sabotage things)
Sony tried that with the PS3's Cell processor, just a wise parallel vector processor instead of dedicated hardware.
If you knew how to use it, you could get great results, which is how Naughty Dog's games became so spectacular. But if you just tried to port an existing game, things went downhill quickly, which is why GabeN talked on stage how much he hated the PS3.
Oh and this complexity led to lots of games just switching to one of the existing engines that supported it natively, and started the modern dominance of Unreal Engine. But it's also why engines introduced more parallel architectures, which a few years later became very useful when quad-core processors in PCs became mainstream.
> If GPU were from the get go designed as a parallelism tool first and video game tools second the hardware architecture and the language would probably be different than today and simpler (...)
The market demand was for video game tools, though.
People didn't wanted faster compute, they wanted their computer games to work smoothly on the hardware they already own. Therefore, they preferred to spend 100-200€ on a point upgrade, much like RAM, than scrapping their 2k€ PC.
To give people a reason to buy these upgrades, games had to use them. Therefore video cards were designed specially for the usecases and development flow that game developers already followed.
If it wasn't for the video game market, NVidia wouldn't exist as we know it today.
> don't forget that the drivers for GPUs would until recently get updates to patch specific video games for performance issues (not sure if they still do that ?)
I bet it's all the same if not worse, and that's why GPU drivers are closing in on the gigabyte "sound barrier" (and I guess on Linux that part is played by Proton... e.g. I bet there's a ton of game-specific 'workarounds' in Proton, if not for performance, then at least to fix 3D API misuse).
Good summary, thanks. Sounds like it's all still pretty early, even for graphics but much more for parallel compute, which seems the direction the hardware is going in. I'm sure there is a CUDA equivalent for CPU but I would guess it's all the way back in the 70s or earlier. Hopefully open standards gets some traction.
One thing that helped me understand it was to reframe it as a network application, where you have a client running on the CPU and a server running on the GPU (the shaders).
A bunch of the complexity comes from having to write both client and server, and the communication protocol between them, then some more comes from the client being responsible for spinning up the server. As with every network service, client/server traffic is the slowest part of the architecture, so a bunch of complexity then comes from having to minimise the amount of network calls (or draw calls, in this case).
Then you hit the part where all of this is performance critical (soft-realtime requirements), so the whole thing is geared towards “writing your own userspace IP stack in C++” rather than “writing a flask application in Python”.
It's not more complicated, you have things like CUDA which is basically C/C++ that compiles to GPU.
But GPUs have thousands of cores, which means if you want any benefit, the smallest data type you must use is something like array[n * 1024], or your GPU is wasted doing nothing. So you must shape your programs into that memory model or you're just wasting your time (and GPU).
Right, that makes sense. There's a ton of overhead in the setup though, I'm assuming that's because everything runs on the CPU by default. Is this then an OS design choice? Thanks for your input.
There can be overhead, though GPU memory + drivers are some of the most optimized software in the world.
Some OS/hardware have separate GPU memory (Nvidia), some have unified memory (Macos, modern consoles), but moving memory is usually not the problem.
Things don't "run on CPU by default", it's up to the programmer where to run it. But the CPU and GPU paradigm of "running a program" are very different, since on GPU you're running SIMD (i.e. running your code on arrays) or else there is no point to use a GPU.
For me, if-branching is the hardest to wrap my head around. "If"s are so natural to CPU programming that it's second nature to any developer but on GPU, it's expensive.
instead of iterating over an array and if-branching to one of two paths for each item, conceive of your conditionals as logic operations which test for set membership (i.e. "is this array item a member of the set i care about?") and with this array mask of "do the thing or don't do the thing" flags, you can mutate your data 0 or 1 times.
the key for me is thinking in terms of sets—much of your conditional logic is probably about sorting things into sets anyway.
The thing that makes it "hard" is that it's all indirect. You compile a program for the GPU (a shader). Then you have to setup data (textures and buffers) for it. Then, you need to spend a bunch of time telling the GPU how to use that data, offsets where each piece of data is, what formats the data is in,, how to advance to the next piece of data, etc.. etc..
There are a few libraries that try to abstract over all this indirection but all the graphics APIs themselves are lower-level so you're left to do all of that yourself.
Vulkan, OpenGL and DirectX make GPU programming more difficult than it has to be for multiple reasons, some of which other people here have mentioned. They are also low level, focused on high performance, and are primarily designed and written for game engines. Making them easy to learn for newcomers is pretty low on the priority list. ShaderToy makes it easy by taking care of everything except the GLSL, you might like it. You can also try CUDA, which is generally a lot easier than Vulkan. For a really easy/fun time with CUDA, try Warp - it’s CUDA in Python.
It's partially historical; I'm sure if we redesigned GPGPU today from the ground up, we'd have an entirely different pipeline. And partially due to hardware vendor lock-in; you can't really design an alternative to Vulkan/d3d12 even if you had the skill and drive to do so.
But also partly in the fact that the GPU is specialized for certain operations over a CPU. It wants and need to obtain data in a different way from a CPU to maximize throughput. So language constructs are designed around making sure developers stay on this fast path. The complexity can be seen as some intentional gatekeeping to signal that this kind of programming is only for specific kinds of problems.
actually yes you could design an alternative API and even implement it....
there's HW references + the linux drivers for AMD (both UMD and KMD), there's refs for NV and their open linux KMD (no UMD though), intel also has HW refs and open drivers for linux
I think it would be possible to put together as long as you don't mind only working for certain hardware, like a prototype you know
I might give it a shot to make a proof-of-concept "high level low level" graphics API ngl if no one else feels like so...
I think there's space for a "two-level" design like have the driver consist of the HW-specific parts, expose things in terms of GPU VAs, explicit sync, explicit memory management, command buffers etc. then have a high-level wrapper over it to provide vaguely GL-ish semantics, and you could intermix the two APIs, obviously you wouldn't be able to get the VA for a high-level object because that'd break the autotracking, if you use low-level you're responsible for syncing, cycling etc.
The problem is this money goes through an incompetent government that can't even build a train anymore. Maybe 30 years ago the story would be different but that's why more taxes is never a good idea. The functioning governments will always decay into incompetence and if they have billions of dollars more to waste, they will.
People incorrectly think this is defending billionaires, but it's actually them defending government. We really should rethink how this wealth flows down, and it should be a more direct line. Maybe instead of a tax it's some kind of obligation to fund non-profit orgs of their choosing.
I've thought of a Factorio mod idea where the grid is based on real life values. Would be interesting to see how much energy you actually need. Maybe it would require some new buildings though, probably better as its own game.
With some exceptions I think higher education should probably go back to being about academia and not necessarily teaching skills for industry. I think companies can now easily invest in their own extended training and grab people out of high school. Just a thought.
It is expensive because it confers greatly enhanced earning potential. Students borrow against that future income. Sometimes from the state.
It confers enhanced future income largely because of the inherent vetting process that asures future employers this person is able to focus for extended periods of time on intellectual tasks, is able to follow through and function within a bureaucracy. I don't think it's as much about skills, at least it's way overkill for the skills that are actually needed on the job for most jobs.
I don't see the rise of LLMs as obviating the "functional cog in the Rube Goldberg machine" training.
It is also expensive because the debt is very hard to get erased through bankruptcy, so private companies are now offering more money to people than is actually needed. So universities see this $ and raise prices.
Kinda like how tax breaks on housing initially help people, and then the market adjusts based on the new inflow of $.
It is much cheaper for a company to do NO training.
It does not matter if Education/Training is done by the government or another company all that matters is that the company does not have any training in its accounts. If it has that is an obvious cost to cut.
Long term this is not good for a company but for the next year or two at least it is better and that is all that their shareholders will look at,
> It is much cheaper for a company to do NO training.
Only if someone else is footing the bill.
> If it has that is an obvious cost to cut.
When it costs tens of thousands per year, yes. If training became 10-100 times cheaper, "training" becomes as expensive per employee as the cost of offering cafeteria food or "gym memberships". Training would be on par of offering any other type of employment benefit, and you don't see shareholders asking for big companies to get rid of any of those.
> Education/Training is only expensive because it is subsidized by the State.
reply
The death of apprenticeships is because you are training your own competition. It worked in the past because you trained your kid for a trade that was arcane and took a couple decades to master, but now you'd be training someone with no loyalty to you in a field that can be speed run in a year or so with YouTube assistance.
You neo-liberals have been wrong about this for 40 years now. You wrought us all with your Reaganomics, and have had the full attention of every government for this entire time (except maybe in Cuba) and the results have been truly disastrous, climate catastrophe and all.
If private enterprise was really this efficient, and government was really that inefficient, we should have seen ample evidence for that by now. The fact that we haven’t is proof that the promises of neo-liberalism are all lies.
> If private enterprise was really this efficient, and government was really that inefficient, we should have seen ample evidence for that by now.
Food. Consumer goods. Telecommunications services. Transportation. Clothes. All things that have gone down in absolute prices and relative amount of family income to be provided.
Education / Housing / Healthcare: all deemed "up to the State" and that takes increasingly more from people at ever diminishing returns.
We can have a long and serious conversation about how hard the Government should be involved in regulating these things, and how we could set up systems that ensure that these become universally available and not just for middle/upper class. We can even have a conversation about things that should be controlled by the Government even if it means a high cost to society. But to deny that things managed by the State are more expensive overall is a clear sign of ideological blindness.
The way to solve this is not by giving the monopoly back to the state, but by the state enacting regulations thay allows smaller players to join the grid.
Europe has been doing this for 20 years now. I have yet to see successful results. And realizing each failure of Neo-liberal policy Europe keeps doubling or quadrupling down. I believe they are on their 4th energy package at the moment. I am sure the fate if this 4th attempt will be identical to the other efforts of making capitalism fix the problem, and identical to any other neo-liberal policy. Europe will have more billionaires and peoples energy bills will be higher.
As the saying in Europe goes: Neo-liberalism will continue until the economy improves.
Do we have 20 years already since people are allowed to set up a solar panel in their backyard and sell excess capacity to the grid?
Do we have 20 years where farmers can grow their own biodiesel to (at least) power their own machinery, or does the state giving them subsidies to let them keep growing vegetables, wine and beef without having to compete in the global market?
We have had 20 years of energy distribution and infrastructure going from government ownership and into private hands, where different private distributors are allowed to compete for customers on the same grid. These were energy packages 1-2 (if I remember correctly). The results have been higher energy prices for European households.
Energy package 3 was supposed to fix this with some additional regulation and standardization. It did not. Energy package 4 is supposed to fix this with more standardization and regulations (including distributed generation; which I am in favor of btw.) however, it will not.
The problem is the privatizations of the 2000s, not in the regulations surrounding small scale solar. The problem is neo-liberalism. Get rid of corporations rent-seeking on our infrastructure. Our energy bills will not become lower until then.
> The results have been higher energy prices for European households.
Under state-owned infrastructure, you simply can never know the real cost of any of the services, because state-owned companies depend a lot more on subsidies that come from tax revenue than actual revenue from the service itself.
> Get rid of corporations rent-seeking on our infrastructure.
Get rid of corporations altogether. I firmly believe that any company (state owned or not) with more than 150 employees should be broken down.
I am guessing Germany or Europe more broadly. Germany is in a pretty bad neo-liberal wave right now, and have been getting worse for the past 10-15 years. And given Germany‘s influence in the EU the effects of this neo-liberal wave are felt across Europe, even in non-EU countries like Iceland (which is currently being fined by the EU for not privatizing garbage disposal fast enough).
However, if you look at the popular movements on the streets of Europe at the moment, you can see that the people are in fact not happy about this. There are very few political means of getting rid of neo-liberalism (see e.g. Spain for a clear example of what I mean) so people are turning to the streets as the only viable alternative.
>If private enterprise was really this efficient, and government was really that inefficient, we should have seen ample evidence for that by now
There's a wealth of clear evidence for this in the economic data in Eastern vs Western Europe, East Germany vs West Germany when Eastern Europe was under Soviet rule. The countries under Soviet rule were significantly poorer and to this day most still haven't caught up with their Western European counterparts.
Eastern Europe underwent shock therapy and are today in many ways more extreme in neo-liberal policies then the western part of the same continent.
Western Europe didn’t start neo-liberal policies until after the cold war ended, and it went into full gear around the year 2000.
In either case. Is garbage disposal in Europe so much more efficient now that it has all been privatized, compared to the mid-90s? Are trains running cheaper? Is power transmission cheaper now that energy generation, and delivery have been split apart and sold to private investors on the public stock market?
No, the answer to all these questions are no. Because efficiencies in a system don‘t care about the ownership model, and neo-liberalism is a religious like creed that operates out of idealism for a world that does not exist.
> is garbage disposal in Europe so much more efficient now that it has all been privatized, compared to the mid-90s?
Is it?
> Are trains running cheaper?
They're up about 6% in real terms (or about 0.23% a year) since 2000 [1]. That said, I don't think most European rail is privatized, so this is a strange example.
Yes and no. Garbage disposal is more efficient now then it was in the mid-90s, but that has everything to do with the infrastructure and nothing to do with the ownership model. More on that later.
The ownership of infrastructure and the operation are split where private operators are allowed to compete with national operators on the same infrastructure. This is true for trains and electricity, and in some cases garbage disposal. Many of the national operators are publicly traded with the national government being the majority shareholder.
Some countries in Europe have gone much further in privatizing the rail network then others. The UK being the canonical example of privatization, and Norway being an example of a country privatizing more slowly.
If the promises of neo-liberalism were true, we should see the efficiencies gained in the UK compared to Norway. Yet, no such evidence exists. And that is because efficiencies correlate with the quality and the quantity of infrastructure, not the ideological policies that dictate the ownership of the operators.
> efficiencies correlate with the quality and the quantity of infrastructure, not the ideological policies that dictate the ownership of the operators
I mostly agree. Infrastructure quality (including institutional quality) dominates. Ideology is usually cover for a lack of evidence.
But ownership modes do create different incentive structures. Full private ownership is famously terrible in natural monopolies. It’s a well-known market failure and folks who proselytize fully privatized utilities are stupid or venal. For many other things, however, private ownership works. A good example is package delivery. (Train operation is a weird case. The UK fumbled. Italy did not. On the other end of the spectrum, DB is a trash fire.)
You may want to double-check your priors on the incentive structures.
For goods and services where demand is inelastic and the barrier for entry is high, the invisible hand will end up colluding to extract the most value possible (housing, utilities, groceries, healthcare). If the state does this, the ruling party starts catching strays and is eventually replaced. Of course, for the distinction to matter, you need to be able to meaningfully create new parties.
Your point stands for elastic demand, where you can choose to vote with your wallet (restaurants, vacations, gyms). Capitalism did a good job of adding to this list during its growth stage (transportation, entertainment, communications) but nowadays it's mostly established cartels that dominate all of the above in what appears to be a race to the bottom.
The incentives for package delivery are closer to the first category than the second though, since consumers end up with whatever company the supplier signed up with, thus mostly eliminating choice and reducing quality to the bare minimum required by the contract (not unlike the postal services).
Even when the customer gets to choose, the end results are still worse than vertical integration. Consider the food delivery apps that replaced most restaurants' in-house delivery teams: they have a direct interest to collude to keep prices high since they all earn the same no matter what restaurant you buy from, whereas for the individual restaurant it makes sense to throw in free delivery to get you buying from them.
I guess the biggest downside to private ownership is that it creates an incentive to have collusion service providers masquerading as adding value (realtors, brokers, speculators, resellers, risk managers, etc.) while providing plausible deniability to the state (which is of course responsible for its failure to regulate the incentive structure for these value-added services on a case-by-case basis).
The thing is that the policy has been splitting off infrastructure from the operators, allowing competition of operators, and selling off shares of national operators. This policy is straight from the books of neo-liberalism, and it promised us greater efficiency.
My point is that these promises are lies. Neo-liberalism is a failed policy (yet it is the dominant policy among every single government in power today). And when GP claimed education was expensive because it is done by the government, I was here to call it out as another lie of neo-liberalism.
I've heard this argument over decades and to this day it doesn't make any sense to me. You know what happens when you do that ivory tower thing? Those professions die.
I don't understand where the premise comes from that higher education is doing this, or even has it as a goal.
To the extent it is I agree it should obviously be replaced by on the job training, because that will be much much better. But to me this is what's happening now anyway, de facto.
Even you do, say, a computer science degree which is one of the more obviously industry aligned fields you're still not really learning how to build software products or even useful software at all really except in the most mechanical sense.
You're learning a bunch of variously useful theory (in the sense that it could all be useful in principle but you won't know exactly for what and at what time until you're out building things in the world) and no company is ever going to be incentivised to introduce you this in a proper environment conducive to intellectual exploration separate from 'the job', because frankly you'll be long long gone by the time any of it pays off often decades hence.
> I don't understand where the premise comes from that higher education is doing this, or even has it as a goal.
They aren't the ones saying it. Parents, government officials and high school guidance counselors are telling everyone they need college to get a good job.
Colleges don't have to say anything thing about it but they raise their tuition abd accept the grant money while expanding their campuses to accommodate the rise in demand. That's exactly how a business would react to a growing market with rising demand and an influx of government money.
People pass through colleges on their way to getting a job and colleges oblige.
What op thinks colleges should be doing is accepting people who they think, with education, could contribute to the college's research fields. Think about the college tracks that lead to PhDs where the student stays with the university for some or all of their career. That's what OP is suggesting colleges become (or revert to, if that's what they ever were).
The software projects you work on as coursework just give you the shallowest possible mechanical introduction on how to put together working code, and their point is almost always to demostrate some underlying theory like networking or distributed systems or graphics or whatever has been understood. It's not to pretend to produce anything viable in a commercial sense or please any imagined customer or market.
Even the ones that are ostensibly under software engineering courses where you might even work in a group, it's about as far away from working on a real software project in a company with (or without) customers and capital on the line etc as can be imagined. Probably even open source would be a closer approximation / better practice at this.
And all of this is a good thing and is the point. It's a sandbox, for you to play around, mess up, 'waste time' and learn. No company is going to really give you this sandbox in the same way, nor should they. They're paying you, you're not paying them.
Maybe AI will be yet another disincentivized for higher education. When I was considering law school a decade ago, the overwhelming advices from existing attorneys was “unless you’re in the top 10% of the top 10% of universities, you’re wasting your time getting a law degree”. Seems very similar to university today (at least in the USA)
Zoho is a classics example of picking students at age 16/17, when they'd typically join a UG course, and instead train them on the job.
Some of the large tech service companies change incentives by tying a training fee ("bond") which is tied to some tenure (1-2 years). Departing before this would mean giving this back - which broadly incentives the planned tenure.
Other companies (musigma is an example) present pay as a 4 year package (10X), which is split 1/2/3/4X across 4 years.
I think the challenges US universities/companies are today facing is something India has needed for a couple decades now. I would not be surprised if many of these are tried in the US.
If you do plan - reach out to me, it's something I've wanted to expand into (currently run UG curriculums).
> With some exceptions I think higher education should probably go back to being about academia and not necessarily teaching skills for industry. I think companies can now easily invest in their own extended training and grab people out of high school. Just a thought.
They'll probably just shut it all down.
With AI, what's the ROI on educating the masses? Paying specialist academics to know things? Just leave them dumb and let save the rich some tax money. Dumb people are easier to control anyway.
On research: If some tech bro can't do the research in an afternoon with an agent, it's not worth doing at all, let alone worth paying a PhD for a tenured lifetime job to do.
> With AI, what's the ROI on educating the masses? Paying specialist academics to know things? Just leave them dumb and let save the rich some tax money. Dumb people are easier to control anyway
The generation of Americans that went to the moon had only 5-8% of people go to college. Whether someone is smart or dumb has little to do with formal higher education.
The idea is AI is cheap enough and worth it for a company to invest in younger talent that will learn the specifics of the industry much faster. It doesn't have to be all or nothing, maybe university is reduced by 2 years or something.
> If some tech bro can't do the research in an afternoon with an agent, it's not worth doing at all, let alone worth paying a PhD for a tenured lifetime job to do.
The OpenAI math dump showed that it’s not worth paying hundreds of PhD-years if a tech bro can do it with an agent.
So I guess if I'm a contractor and doing a job for Meta, not only can I completely fail and they still pay me but if I do something illegal it's on them for hiring me, right??
Yeah I don't know what the solution is but something is wrong here.
Maybe refund mistakes or take responsibility.
There's no paradox, there never were agents to begin with. Those are programs. It's same old software. If you sell me a software to do X and you fail to deliver, it's on you.
We're certainly heading towards a sci-fi dystopia, the question is which one and by how much. Best case scenario is that it stays on the fringe. Worst case is it causes a global catastrophe that takes generations to unravel.
I'm thinking like a parallel to slavery in terms of proportions and institutionalization, but not as obvious to us right now. It starts by controlling all of our information, with companies owning our entire communication, family photos, even medical records. It's all in the name of "AI progress" but it inevitably these institutions get entrenched and corrupted causing widespread and permanent underclass that can't even participate in the system. Two hundred years from now people will realize how immoral this was but to us now it's still normal.
I'm so tired. From the 2008 recession, into a short break in mid 2010s, into a pandemic, into an AI revolution, my career is a series of managing catastrophic changes. I was a sold a lie by society and universities that studying some field, i.e. CS, meant some kind of meaningful progression. In reality it is all about profit, and selling products. I guess I was too naive to be that cynical, but I'm realizing now, it's literally all about selling product and making money and nobody gives a shit about expertise, craft, or specializations.
> In reality it is all about profit, and selling products.
All business endeavours are about profit and sales.
I don’t know why people in tech keep thinking they are special in this sense.
There is the widespread meme about getting out of IT and going farming or sone variation of that. Do you think farmers actually live free of the thoughts of profits and sales?
It isn't about being special. It's about what exactly I'm supposed to optimize and maximize as a skill in order to be successful. It turns out the only right answer is business and marketing skills. So then what's the point of learning anything beyond surface level? Especially now with AI, but this was true before as well going back probably many decades.
> It turns out the only right answer is business and marketing skills.
< astronuat meme>
Always has been.
Look, I feel really bad for you, and I hope things get better for you (and the others in this thread).
But it's worth calling out that the post-GFC software world was entirely insane. The low interest rates caused a lot of madness to be funded, and software engineers and other "technical" people got treated with kid gloves because their skills were perceived as rare, and the moonshots only needed to pay off 1 of 100 times for the funders to be happy.
That's not the world we are in today, money is much more expensive, and the mainstreaming of remote work made a lot of businesses think about offshoring even more than they were already. Also with that, there's AI where lots of people believe they don't need tech people any more (they're probably wrong, but it'll take a while to come out in the wash).
> So then what's the point of learning anything beyond surface level?
Because if you learn some useful skills (computers etc), AND you learn how to talk about business & sales, then you can get a better job. The days of just being able to sit in a corner and code are gone for now, but people with enough skills to talk to other humans about their actual problems can definitely do well.
But again, mostly I'm sorry that things haven't worked out for you and that things improve soon.
> All business endeavors are about profit and sales.
This is the kind of simplistic thinking that has destroyed the modern world. Businesses are just groups of people, and most people don't care about the endless optimization of profit and sales -- they care about having purpose and they want their work to contribute back to society in a way that benefits their community.
The problem is we've empowered "entrepreneurs" (mostly narcissists) who espouse this form of one-dimensional thinking because they're the best at consolidating money for their own benefit (and the benefit of others who directly help them).
> All business endeavors are about profit and sales.
You say this because to you business is a game about winners and losers, and in your interpretation you believe all players should simply work to maximize their "points" at the expense of everyone and everything around them. This form of thinking only comes from "entrepreneur" types who are actively destroying the world and the very fabric of our society for their own personal gain.
> It’s not even about cynicism… it’s just reality.
It's only reality because we've allowed a small group of selfish cynics to take control of our society.
I don't think business should be about just profit and sales, but your point is that it isn't about it right now, which I disagree with. It's not about us choosing to socially support or allow the narcissists to exist. The whole system that's set up here is self-reinforcing. The most greedy, selfish and cruel are empowered to rule over everyone else, because the only things that are rewarded with power and influence are gaining more money. Some people who run businesses with purpose and integrity still exist, but year after year they're getting slowly pushed out, until they will eventually occupy only the most fringe margins of economics and all the air is taken up by the narcissistic giants. In that sense, in 2026 business ventures are all about making as much money as possible and nothing else. Game theory-wise, anyone caving to their morals is playing business wrong and will lose, because that's how the game was constructed. As far as I can see, there is nothing you and I can do.
> You are one of the individuals who thinks it is about profit and sales. I'm recommending that you change your mindset.
I think you're severely misreading what I'm saying. I would never do or condone any of the abusive acts that modern businesses have largely sunk to. I'm just saying that the fact that our system has rewarded the biggest profit-chasers is completely obvious and will not change regardless of what my mindset is. Said profit-chasers are now integrated so deep into everything that it will take far more than regular people like us changing the culture. They don't care about culture, they already own everything and can do whatever they want.
I understand what you're saying, this is the part I disagree with:
> I'm just saying that the fact that our system has rewarded the biggest profit-chasers is completely obvious and will not change regardless of what my mindset is.
Specifically the "will not change" part.
It's our system, and we do have the power to change it. First we have to change the way we talk about it. Stop calling them "winners" and stop calling those who follow a moral compass "losers." Call the entrepreneur-types what they are: sad, selfish individuals who deserve no power in our society. They're not winning at anything -- they're dragging our entire society down with them and making all of us losers.
I'm not assigning moral value to them when talking about 'winners' - objectively, in a game where money equals absolute power, it would be kind of absurd to say that the tiny minority that together holds far more power than the remaining majority didn't 'win' at that game, right?
Take a look at observable reality. Not how you wish things worked. Being optimistic won't change anything. Won't change the balance of power. Won't make anything more fair. Just make you a more pleasant serf. It literally doesn't matter what you call Elon Musk. He still has tremendous power. He has hordes of defenders. No amount matter how much power you pretend people have with their vote you're still ignoring observed reality. In your world, someone like Trump could never be elected much less twice. Yet here we are. There seems to be a fundamental disconnect between your beliefs and reality.
> Do you think farmers actually live free of the thoughts of profits and sales?
You can escape the market by becoming a subsistence farmer or trying to join the hunter-gatherers living on Sentinel Island. Might involve a reduction in standard of living though.
You don't even have the freedom to choose that unless you're both already wealthy and crazy.
How will you afford to pay your property taxes on a subsistence farming? Or do you plan on squatting on someone else's land? Public land? I think it's possible to camp longterm on the latter, but the government probably won't appreciate your farming operation.
> or trying to join the hunter-gatherers living on Sentinel Island
Good luck, I think they're known for instakilling outsiders over there.
He said 'all about profit', not 'about profit'. There's a difference between building a great, innovative, useful product and making a profit from that vs literally enshittifying/worsening an existing product (or just building total crap that came out of some illiterate MBA's ass in general) and aggressively leaning into marketing, dark patterns narrative control, etc.
These days it seems like most of SV and US tech industry in general is mostly the latter. This is a big reason why US tech companies can't compete with China anymore - VCs and big tech have captured the US market 10 years ago, and have been just squeezing it for profit ever since.
I worked a general public facing job that I will not describe in detail lest I doxx myself, but you can imagine it as basically tourist facing retail. I did this for 10+ years. I was already deeply cynical by the time I went to school for CS, and yes, it was for money, and I still prefer it to the alternatives.
The world of menial jobs is absolutely miserable, depressing, and will grind you to a pulp - it's nearly impossible to get out of, by design.
It's perfectly ok to study CS for money. My point is that the job market in general doesn't really care about all this technical specialization, despite them signaling this over and over with CS degrees and technical interviews. At the end of the day all careers have to do with making more profit for the business, not specializing in some skill.
Your knowledge and experience is super valuable. Your bosses and the typical parasites running the typical SV tech company do not have the knowledge and experience you do .
You shouldn't be discouraged just because they don't understand or value your skills and expertise - you can build a product and/or company from the ground up by leveraging your skills and using AI to fill in the rest. They cannot do the same thing.
Time to stop worrying about the fact that you can't serve these clowns and time to start building.
> you can build a product and/or company from the ground up
The MBAs can. Engineers can write a program, spend 5 years polishing it and convert maybe 2 paying users. The "just start a business" is the "learn to code" advice of the 21st century. It's not a panacea, or a viable path for most engineers to return to the career they had before.
No they can't. Their only skill is in squeezing profits out of existing products and corporate politics. These people are mostly technically illiterate, and they don't understand the tech market/business well either. Even AI will not enable these people to deliver working, quality code.
Writing code is a very specific and highly technical skill, you cannot just will yourself to do it without many hours of study and practice. Doing business strategy + machiavellianism 101, which is realistically what 99% of these people do, is not a very complicated or technical skill at all, the average high schooler understands these things for the most part, and AI can do it better already.
There are thousands of products and companies out there that were once okay, and did pretty well, after which they got taken over by the MBA/marketing people, and now their products are enshittified, stagnated dogshit, they only exist because of existing customers and connections, and they can be easily outcompeted.
To the contrary - good products and businesses die all the time, because they have no executive leadership or marketing plan. Highly intelligent programmers end up starving themselves trying to sell a video game or an innovative algorithm.
Yes, these MBAs will be releasing products and starting businesses that were once relegated to only developers. They'll overpay for cloud hosting, completely fail to innovate, abuse their customer relationship, and make much more money than solo devs ever did. The programming mindset is restrained by unprofitable concepts like "taste" and "respect" and other nonsense principles that the industry long since forfeit to FOSS. The victor will be the contractor, the business grad and the tradespeople. The library maintainers, indie devs and SaaS addicts are going extinct.
The programmer will begin their final phase-out, the one that started when Oracle and Microsoft subcontracted their war crime software to third-world workers.
That's true, you do need marketing, sales, etc. But is there any secret sauce to that beyond having existing connections? Not really. AI does that job better than the 99% at this point.
> Yes, these MBAs will be releasing products and starting businesses that were once relegated to only developers.
Show me a single, actually working, profitable tech company built by a non-technical MBA without the need for developers. Until then it's fantasy. I've watched numerous seemingly high-ranking, very successful MBA/marketing people try - it's a complete joke - they need nerds to babysit them 24/7.
> They'll overpay for cloud hosting, completely fail to innovate, abuse their customer relationship, and make much more money than solo devs ever did.
Cool fantasy I guess
AI/LLMs are an amazing tool and a multiplier, but they are far from perfect - it requires technical understanding to use effectively. An experienced software engineer can leverage it 1000x more effectively than a typical corporate MBA NPC. That's all there really is to it.
Where's the disruption? If experienced engineers can leverage AI "1000x more effectively" than their peers, we should be seeing the consequences. The only thing that's happening is that software is getting cheaper. The high waterline of quality that used to differentiate software products has boiled away. Many professionally-designed programs are being functionally replaced by vibe-coded alternatives. FAANG is barely shipping software any more dignified or advanced than vibe-coded crapware.
Ask yourself this; do you think Apple would rather give up their engineers, or their brand image? Only one of them is replaceable.
The disruption is definitely happening. The main roadblock is that the US tech industry is completely captured by big tech and VCs, who openly collude with each other. But that is temporary and the cracks are starting to show - the aggressive push towards AI regulatory capture is just one obvious sign. Explain to me how DeepSeek, a company with far less access to compute and capital, and starting from behind, is now able to offer LLM inference at roughly 10x the value per dollar of the frontier US AI labs.
> Ask yourself this; do you think Apple would rather give up their engineers, or their brand image? Only one of them is replaceable.
What do you think the 'brand image' is? Is it a pretty design/image, or a cute ad campaign, is it just whatever the marketing/propaganda says? Or is it actual product quality, reliability, efficiency, etc. ? If apple gives up their engineers they have no more brand image at all, and no more products either.
Everyone is already pointing out this isn't quite software rendering but rather the rendering technique which used to be the only option. I'll add that even without a GPU, CPUs will have integrated graphics. There are CPUs that don't have this but in those cases you need a GPU. The point is that modern systems no longer use CPU to output to the screen, either there is an integrated GPU or a GPU card on the system. So in a way the term "software rendering" is only applicable to special cases of embedded systems that actually use CPUs for graphics.
I'm not totally sure what you're trying to say. Even back in the days of "software rendering" - or on those embedded systems that you're describing - you still had a piece of graphics hardware that converted the frame buffer into whatever got spit out to the monitor, so I don't think this is meaningful, really.
What's meant by software rendering is really software rasterization - the process of taking the triangles, textures, etc. of the scene and turning those into the pixels that eventually make their way to the screen. You can still do that in software on the CPU today without touching the GPU, even the integrated one. At that point you hand it off to the hardware to make its way to the screen, sure, but it's only an implementation detail that makes that different from handing 0xA0000000 over to the VGA card. The interesting part of the process, and the results, are functionally the same.
The difference is if it's hardware accelerated or not, or at least that's my understanding. Software rendering means there is no hardware acceleration. I think the term is outdated because typically it means you don't use the GPU, and this was true like a decade ago and earlier, but I think today CPUs have integrated graphics, so it's like they have tiny GPUs to help render to the screen because it's more efficient. So even if someone is not using a physical GPU, chances are the CPU still uses hardware acceleration, and that's why it works so well today.
I have an Intel i9 with no integrated graphics, if I don't use a GPU, the screen would barely render in very low resolution, although haven't really test it that. I can disable hardware acceleration in the browser and it spikes several CPU cores and is terribly inefficient.
It's still narrow AI but massively scaled to look like general AI.
reply