I've held this view and talked about it many times here before.
It seems like an obvious conclusion to me that the end result will be a few AI owners trading among themselves should AI develop in what seems to be likely: recursive self improvement, robotics allowing it to displace manual labor and combat.
Then the owners will be trading for land, AI tech, minerals, energy, which will likely be owned by the other AI conglomerates, and maybe the odd thing that can't be replaced by AI like human entertainers that would make up 1% of the economy.
For what purpose exactly? So I am a rich AI owner and my goal is to get more land to build another AI data center? And my robots will combat the other AI owner's robot of that land and resources? What sort of trade am I going to be doing with the other AI owners?
That feels a bit silly. I mean anything is possible. Anything is possible even if you take AI out of the picture. All countries are like North Korea and their rulers fight and trade. Or all of earth is government by one oppressive dictator. So far it seems the broader incentives/forces push us in a different direction.
AGI and robotics do potentially change some of the dynamics.
To be fair, that doesn't seem to be stopping any of the billionaire class from trying frantically to accumulate more wealth today, I don't know why AI ascendant would change any of those incentives.
Why does it feel silly? There are already billionaires, and now Elon Musk is a trillionaire, and they still want more even though they have enough money to spend for several lifetimes.
Some people always want more. And defending against others like that will result in infinite demand.
This is similar to what I experienced when I tested mimalloc many years ago. If it was faster, it wasn't faster by much, and had pretty bad worst cases.
Agreed mostly. Going from standard library to something like jemalloc or tcmalloc will give you around 5-10% wins which can be significant, but the difference between those generic allocators seem small. I just made a slab allocator recently for a custom data type and got speedups of 100% over malloc.
I've been using jemalloc for over 10 years and don't really see a need for it to be updated. It always holds up in benchmarks against any new flavor of the month malloc that comes out.
Last time I checked mimalloc which was admittedly a while ago, probably 5 years, it was noticebly worse and I saw a lot of people on their github issues agreeing with me so I just never looked at it again.
I've benchmarked them every few years, they never seem to differ by more than a few percent, and jemalloc seems to fragment and leak the least for processes running for months.
Mimalloc made the claim that they were the fastest/best when they released and that didn't hold up to real world testing, so I am not inclined to trust it now.
> Mimalloc made the claim that they were the fastest/best when they released and that didn't hold up to real world testing
That’s… ahistorical, at least so far as I remember. It wasn’t marketed as either of those; it was marketed as small/simple/consistent with an opt-in high-severity mode, and then its performance bore out as a result of the first set of target features/design goals. It was mainly pushed as easy to adopt, easy to use, easy to statically link, etc.
I tried all three, multiple times, and it depends.
Using the last workload tested as an example, mimalloc just consumed memory like crazy. It was probably leaking, as it was the stock version that comes in Debian, so probably quite old.
Tcmalloc and jemalloc were neck to neck when comparing app metrics (request duration etc... was quite similar), but jemalloc consistently used only about half of RAM as opposed to tcmalloc).
Both custom allocators used way less RAM than the stock allocator though. Something like 10x (!) less. In the end the workload with jemalloc hovers somewhere around 4% of the memory limit. Not bad for one single package and an additional compile option to enable it.
I'm not sure that's entirely true. For most things, checking if a solution is correct is much easier than implementing it (page looks wrong, can't login etc...)
I think some of it might be genuine. For people that don't code (like management), going from 0 to being able to create a landing page that looks like it came from a big corporation is a miracle.
They are not able to comprehend that for anything more complicated than that, the code might compile, but the logical errors and failure to implement the specs start piling up.
This paper creates a new benchmark comprised of real remote work tasks sourced from the remote working website Upwork. The best commercial LLMs like Opus, GPT, Gemini, and Grok were tested.
Models released a few days ago, Opus 4.6 and GPT 5.3, haven't been tested yet, but given the performance on other micro-benchmarks, they will probably not be much different on this benchmark.
One of the tasks was "Build an interactive dashboard for exploring data from the World Happiness Report." -- I can't imagine how Opus4.5 could've failed that.
Waiting until the moment they get good enough is not a smart thing to do either. If you are a farmer and know it is going to snow, at some point in the next 5 months, you make plans NOW, you don't wait until the temperatures drop and you see the snow falling. Right now, people are waiting for the snowfall before moving their proverbial chickens indoors
Top AI researchers like Yann LeCunn have said that LLMs are a dead end.
It seems to me that LLM performance is plateuing and not improving exponentially anymore. This recent hubbub about rewriting a worse GCC for $20,000 is another example of overhype and regurgitating training data.
You don't know for sure if it is going to "snow" (AI reaches general intelligence) Snow happens frequently, AI reaching general intelligence has never happened. If it ever happens, 99% of jobs are gone and there is really nothing you can do to prepare for this other than maybe buy guns and ammo, and even that might not do anything to robotic soldiers.
People were worried about AI taking their jobs 60 years ago when perceptrons came out, and anyone who avoided a tech career because of that back then would have lost out majorly.
There is no reason why an AI model capable of pushing a significant chunk of devs into lower paid and highly competitive dev jobs as a result of automation needs to be a general artificial intelligence. There is a lack of nuance that comes with thinking that either AI is dumb or it has human level general intelligence. As much as devs hate to admit it, you don't need that much of what we understand as general intelligence to write software. Only a portion of your intelligence is needed and arguably not all of it at the same time.
While general purpose models might be plateauing soon (arguably they have for a while). Highly specialised models (especially for programming) haven't necessarily plateaud yet. And anyway, existing functionality seem like a good foundation to build upon systems that remove the need of hiring as many devs. It's not the "being out of a job" that should worry you. Open up your binary thinking and consider that facing a 08 job market for the rest of your career is not the same permanent unemployment but it is not a market you would like to have.
You don't need to be a genius or rocket scientist to write code, but llm don't even reach the bar for anything but the most simple things. Take a look at the video I posted earlier for an example.
And specialised models for programming HAVE plateaued.
> Can you imagine not being fired when you can only do 2.5% of all tasks?
You are not competing against LLMs though. You are competing against people (who in a pre-LLM world wouldn't be in tech) using LLMs tools to beat you in terms of value. In the new world, you either are a top 1% dev or you beat everyone in race to the bottom pricewise. The middle will become vanishingly small. Think of manufacturing in developed countries.
This technique showed that there are ways during training to optimize weights to neatly quantize while remaining performant. This isn't a post training quantization like int4.
For Kimi quantization is part of the training also. Specifically they say they use QAT, quantization aware training.
That doesn't mean training with all integer math, but certain tricks are used to specifically plan for the end weight size. I.e. fake quantization nodes are inserted to simulate int4.
Iirc the paper was solid, but it still hasn’t been adopted/proven out at large scale. Harder to adapt hardware and code kernels to something like this compared to int4.
Has anyone noticed a lot of polymarket posts on their X (formerly known as twitter) feed claiming to be making a fortune? It makes me feel like its some kind of coordinated guerilla marketing scheme.
It seems like an obvious conclusion to me that the end result will be a few AI owners trading among themselves should AI develop in what seems to be likely: recursive self improvement, robotics allowing it to displace manual labor and combat.
Then the owners will be trading for land, AI tech, minerals, energy, which will likely be owned by the other AI conglomerates, and maybe the odd thing that can't be replaced by AI like human entertainers that would make up 1% of the economy.
reply