I’ve been told that the LLMs are reaching pretty hard diminishing returns, in that you need exponentially increasing amounts of compute for linear returns in model performance. The cost of marginal improvements is bumping into practical limits.
It can’t turn into general AI, that’s not how LLMs work. So they’re uselessly throwing money after nothing.
It’s like saying adding more steps to a ladder will make it fly.
Fundamentally, LLMs are not AI.
Here is the PR mentioned in the video https://github.com/matplotlib/matplotlib/pull/31132
Oh nice thanks!
I truly believe its bad now, but its not going to last. Yes, I know that they want to take computers from us, but their purchasing isn’t going to stay this consistent. Its going to be rough for a bit, but after a few more contracts fall through and these purchases aren’t renewed for year after year (and looking at stock prices investors are not happy about these massive purchases) data centers are going to purchase less, and I really believe we’re going to see hardware crawl back to gaming.
Its up to us to decide which companies deserve our money when that day comes.
The mad grab for data center real estate actualizes profit pretty much only in an authoritarian state.
It’s kinda neat how you can actually see the billionaires gameplan while they wage war on the lower classes. I feel like it’s never been so transparent.
It is pretty much destined to fall apart. The only chance they’ve got is brute force, which will become harder and harder to maintain as support plummets.
A problem is that a lot of the companies that make consumer hardware cannot survive in the meantime.
The flash manufacturers are expecting shortages through 2030. GPU’s have been a shit show for years now, and RAM and SSD’s are following. Now HDD’s are following SSD’s.
It’s getting too expensive for anyone to build a PC or small home server. So what’s going to happen to the other companies? It’s going to be hard to stay in business making ATX desktop cases and power supplies. Demand for consumer CPU’s will fall and lead to AMD and Intel shifting their capacity to enterprise on that front too. Motherboards will follow as well. Some of that is more flexible than others, but even if the AI bubble pops it’s going to take years for the markets to recover.
Think of the housing market. Real estate developers in the US, Canada, UK, Australia, and a lot of the rest of the world have spent decades focusing on building large single-family luxury homes at high profit margins. To the point where they would rather keep building large homes and leave them empty over building apartment buildings or duplexes or smaller homes. I see a similar trend in the consumer PC space, and it’s going to take government regulation to stop it.
The Chinese are moving aggressively into this market.
If shortages last until the 2030s, they will catch up and dominate the market.
I will have no sympathy for Western companies if that happens.
The AI bubble bursting, which I would put good money on happening this year, will probably cause a global financial crash on a similar scale as 2008. Many AI and tech companies will be completely exposed as dead trees and basically wiped out completely. A bunch of hardware will most likely flood the second hand market at that time. Sadly, due to job losses and ballooning cost of living, few of us will be in the market for it, so mostly the wealthy and corporations will benefit from it. Good opportunity to get a decent quality office chair, though.
If you look at the finances it’s on track to burst sometime in 2027. That’s when loans start to come due and when investors first expect to start to see a return for many of these deals.
They basically have a year to achieve agentic AI or AGI or whatever were calling it now and monetize it.
Honestly, as a tech worker who will likely lose their job when it bursts, I hope it happens soon. The sooner it happens, the less devastating it will be. We’ve already gone off of the cliff, now it’s just a matter of how far we’re going to fall
