How many people actually use fable over opus? How far are we up the diminishing returns curve, and will their customers even care?
Or would it affect people's purchasing power. I guess to some extent it does for people with significant investments in the stock market, but that's not a huge percentage of the people wanting to buy RAM or devices with RAM
- Suppliers that memory companies depend on JIT are entangled and go under, halting production or stalling capacity increases
- Downstream suppliers / distributors for not vertically integrated memory companies go under, and there is a shortage of finished product ready for consumer use even though there is an excess of memory chips
- Memory producers / fabs can’t handle the depressed demand and go under / have to halt production, alleviating the demand glut for the surviving suppliers, who have less competition. Coordinating a cartel becomes easier when there are fewer suppliers and becomes more desirable now that everyone realizes it’s a matter of survival.
There are also plenty of good reasons to think prices will go down if demand is depressed, but it does seem possible for prices to go up
This might become the driver for memory and chips demand in the near future.
But i don’t know, of course, we’ll i guess.
It will happily walk me through all the legal and illegal ways to systematically kill rodents.
At least it is on-brand for something used by militaries that want to target civilians, I guess.
However a problem exists with scalper bots. They're going to fight tooth and nail to keep prices artificially high even when supply increases. Same with the hyperscalers blowing "free money" at eating supply as a strategic position against smaller competitors.
[1] https://en.wikipedia.org/wiki/DRAM_industry_price_fixing
[2] https://moginlawllp.com/dram-antitrust-suit-ai-memory-supply...
If I sold just the two sticks of RAM in it right now, it’d pay for nearly half of the total cost.
I still have it, but never installed it because I only use ECC memory in my desktops. I'm saving it for a future Desktop in case memory prices never fall again.
Basically everyone that makes memory is building new fabs, meanwhile I'm not sure how much longer AI datacenter demand for ram will last. I think the decrease in AI ram demand and the new fabs will likely coincide leading to a collapse in pricing.
That is, of course, assuming the memory manufacturers don't pull their favorite trick and collude.
For 1, AI datacenter builders have said that they have more equipment than they have places to put them. Leaving a ton of hardware shelved while you wait for datacenters to build out is bad business to say the least.
For 2, I think we are nearing saturation for the usefulness of AI. I certainly could be wrong, but I don't really foresee there to be a bunch of new exciting usages of AI that will ultimately justify the continued buildout.
Unless some other breakthrough in AI happens, like applying AI to some other large market.
Overall i think that AI is here to stay, if anything it will become more common.
Maybe in ten years we’ll all have some 256gb ai machine at home, like we most (all?) of us have AC.
If that happens, hopefully the ram price comes back to earth.
I’m sure some people will. But most people I know don’t think about software like that. They don’t know what use cases are, or ask themselves how their life could be simpler with different software. LLMs only make software if you prompt them. Most people don’t frame the problems they have in life as “I’m missing software to solve problem X”.
I think most people will never prompt LLMs to write software. Just like I don’t have a 3d printer, because I don’t think about how problems in my life could be solved with little plastic doodads.
I also know someone that thought about making a SaaS for his specific niche he works in, he asked me if I ever used Claude then told me his project.
Yes, most people will never do it because they're not the type of people that are self-motivated and courageous enough to dive into something with confidence to see how it goes.
But probably, as more people build stuff and it makes headlines that X person built Y service and is making Z$$ per month and he can't even write a hello world to save his life, a lot more people will try it out.
Or they’re just not interested. Just look around to see how many things that people are doing that you have no interest into. That’s the same disinterest people have towards programming and sys admin.
If person X didn’t contribute anything but a prompt, he probably wont be able to charge much.
First guy I mentioned spent a month tinkering, testing what he could, getting a basic high level understanding of the problem space to understand why his prompts didn't work, etc
But plumbers are still in very high demand, and charge really high prices for what is in many cases really simple stuff.
For a lot of folks, the idea just never enters their mind that they can solve simple problems themselves. Most people don’t want to spend the time, undertake the learning, have the possibility of doing it wrong, risk their property, insurance/code issues, etc. And this is an absolutely reasonable position to take! We can’t all be specialists at all things.
All that to say: people will long think that software is something that is made by someone of a special role. Even long after that’s not necessary anymore. That’s not great job security, but it’s what I’m going with for now!
This is why ROI is so critical, and I’d argue that picture looks a lot less optimistic for it to pan out functionality wise today than it did in March, when everyone was being forced to use it. Faced with millions of dollars in token costs and gains not materializing, the question is more about the overall usefulness of LLMs as a matter of returns.
We do seem to be spending trillions of dollars to drown our public spaces in grey goo
Could you provide more details about the Epyc + DDR4 server?
The above is a standard project management problem. We do this for lots of industry all the time. There is every reason to think you can get a new factory running in 5 years.
Note that I said 1 factory above. Some of the special machines we don't have the ability to make them fast enough to do 2 (I don't know the real number!) new factories in 5 years. Existing factories are using most of the special machine capacity to replace machines that wore out on the way - this can be corrected as well, but it adds another year and the expenses are much larger. Realistically though 1 new factory is likely enough.
https://transportgeography.org/contents/chapter3/transportat...
...
> I don’t think anyone truly knows when, but it will happen.
Do you know what cyclical means ... ?
Fab capacity is being bought online; there’s just lead time.
Noticeably greater intelligence is being achieved at the same number of parameters (see: Qwen3.8).
I think the future will be bright, it might be a matter of time. And for tinkers, a used Epyc + DDR4 server can be great fun and epic value.