You are right, but this doesn't mean that the amount of work is conserved as your original message implies. The correct statement would be that "algorithmic complexity is just one aspect of actual practical complexity and an algorithm with better algorithmic complexity can end up performing worse in reality due to practical considerations of the data and the processor doing the computations".
It was nothing like covid. The dot com crash lasted years where tech was a dead sector. Equity valuations kept declining year after year. People couldn't find jobs in tech at all.
There are still plenty of tech jobs these days, just less than there were during covid, but tech itself is still in a massive expansionary cycle. We'll see how the AI bubble lasts, and what the fallout of it bursting will be.
The key point is that the going is still exceptionally good. The posts talking about experienced programmers having to flip burgers in the early 2000s is not an exaggeration.
After the first Internet bubble popped, service levels in Silicon Valley restaurants suddenly got a lot better. Restaurants that had struggled to hire competent, reliable employees suddenly had their pick of applicants.
History always repeats itself in the tech industry. The hype cycle for LLMs will probably peak within the next few years. (LLMs are legitimately useful for many things but some of the company valuations and employee compensation packages are totally irrational.)
They already do this. Including peripherals which appear as an actual mouse, but they are there only so that cheat software can take control of the input without modifying the game memory. There are cheats which run on a separate machine and access the game memory via a dedicated DMA card (which itself presents itself as an innocent piece of hardware). Note, this can still be detected either via detecting the DMA card itself, or eventually these shenanigans will be killed off by IOMMU.
Unfortunately, there are also plenty of offerings which do not touch the game memory or process at all, and work purely based on image recognition and these days they actually use AI that is trained on specific games. I have no idea how they plan to detect these. All the cheat needs is the video feed and the ability to provide input via mouse and keyboard, and as you say this is trivial to do in a way that is entirely undetectable.
I'm writing this as a heavy python user in my day job. Python is terrible for writing complex systems in. Both the language and the libraries are full of footguns for the novice and expert alike. It has 20 years of baggage, the packaging and environment handling is nothing short of an unmitigated disaster, although uv seems to be a minor light at the end of the tunnel. It is not a simple language at this point. It has had so many features tacked on, that it needs years of use to have a solid understanding of all the interactions.
Python is a language that became successful not because it was the best in it's class, but because it was the least bad. It became the lingua franca of quantitative analysis, because R was even worse and matlab was a closed ecosystem with strong whiffs of the 80s. It became successful because it was the least bad glue language for getting up and running with ML and later on LLMs.
In comparison, Rust is a very predictable and robust language. The tradeoff it makes is that it buys safety for the price of higher upfront complexity. I'd never use Rust to do research in. It'd be an exercise in frustration. However, for writing reliable and robust systems, it's the least bad currently.
What's wrong with R? I used it and liked it in undergrad. I certainly didn't use it as seriously as the users who made Python popular, but to this day I remember R fondly and would never choose Python for a personal project.
My R use was self-taught, as well. I refused to use proprietary software for school all through high school and university, so I used R where we were expected to use Excel or MatLab (though I usually used GNU Octave for the latter), including for at least one or two math classes. I don't remember anything being tricky or difficult to work with.
R is the most haphazard programming environment I've ever used. It feels like an agglomeration of hundreds of different people's shell aliases and scripting one-liners.
I'll grant my only exposure has been a two- or three-day "Intro to R" class but I ran screaming from that experience and have never touched it again.
It maybe worked against me that I am a programmer, not a statistician or researcher.
When I used it I was a computer science student. But I wasn't reading anyone else's code or trying to maintain anything complex, which is why I asked what I did. I'm sure there are quirks I never had to deal with.
So is it just that the stdlib is really big and messy?
How is this going to solve the supply chain attack problem at all though? It just obfuscates things even more, because once an LLM gets "infected" with malicious code, it'll become much more difficult to trace where it came from.
If anything, blind reliance on LLMs will make this problem much worse.
I don't understand how is this "forced privatisation"? There's nothing preventing a state owned rail company from operating, no? All this says is that private companies must be allowed competitive access. Am I missing something?
No nothing preventing it but you cannot just have a state operator bid for it. The contract in question is basically the entire Netherlands passenger operations, with some small exceptions.
The key problem the EU has is the size of the contract. The contract cannot be for the entire country (more or less), it needs to be split into smaller ones (eg: one contract for intercity, one contract for local services near Amsterdam, etc). Netherlands defense is that the passenger rail system cannot be split up like this as it would cause operational problems and thus should not be franchised. I'm simplifying this from memory but it's roughly correct I think.
There is no doubt in my mind were it to be split up like the EU is requiring, that de facto some of it would be privatized. It has/is happening in every European state that has complied with the EU rules.
To me that is very close to forced privatisation. The current (Labour) UK model (government owned franchises with no bidding) would 100% not be allowed.
I'd argue that subsidizing the railway system from the taxes is inherently a better way to pay for it. In the UK, the trains are ridiculously expensive out of pocket. They are vastly more expensive than in Switzerland (a really high CoL country), which as you say is a much better run system at the same time.
The total costs might be the same, but if paid out of the taxes, the cost is distributed progressively. People who earn more will pay more to keep the trains running. In the UK, the cost is regressive by definition, because the portfolio manager who earns 1 million pounds a year will pay the same expensive ticket fare that the nurse who earns 40k a year, but for the PM it's a trivial expense. And the portfolio manager can work from home, or just drive in to work if they fancy, the nurse cannot. The nurse is forced to use an aging, unreliable and at the same time really expensive train.
And the service itself is positively atrocious even in London when it comes to any non-mainline route. It's a complete tossup whether the trains will have severe outages due to strikes, engineering works or the frequent signal failures.
You touch on the core issue of the UK, but there's both a cultural and policy aspect.
1. As you say, an aging population has placed the entire country into a gridlock - the pensions system as it stands is ridiculous and unsustainable. Unfortunately, it is an almost immediate political suicide to even attempt to reform it. Young people are disillusioned and wield little monetary or political power. The old generation has all the cards, and are actively destroying their children's future for their own short term benefit. They have mortgaged their children's future, which they won't get to see, for the short period of time they still have remaining.
2. There's no investment culture in the UK. Literally everything and everyone is rent seeking. Want a retirement? Buy property and rent it out. Want a pension? Pension funds just buy gilts - government bonds. Very little money goes into productive assets. This creates a vicious cycle where money is simply siphoned away from the little productivity that remains to an ever growing number of rent seekers.
There's no easy way out of this. One generation will effectively have to give up their welfare, but nobody wants to do this (understandable). The UK cannot afford to be the welfare state it wants to be.
> There's no easy way out of this. One generation will effectively have to give up their welfare, but nobody wants to do this (understandable). The UK cannot afford to be the welfare state it wants to be.
It can be done over a period of time. Canada moved from a non-funded pension model like the UK currently has, to a partially funded model which currently has over $731 billion in assets.
They began this change in 1999. If they hadn't made that change in direction they'd be in a much worse situation today.
It's possible to change from one system to another, if a government is able to look much further ahead than its own term in government.
With the Canadian model, your payments out at the end are tied to what you put in. Which is not quite the case in the UK, which allows extremely low payments in for just 10 years to get a very high amount out indefinitely.
The British government must take the same path the Canadians took.
This is so far out of my expertise that I have nothing to add to the actual topic, but I must say that the veritable forest of beam mirrors, splitters and whatnots looks fascinating. It reminds me a bit of the early integrated circuit research where people had similar forests of transistors.
reply