Chip Shortages Lead to More Industry Instability as SanDisk Raises Its Prices, and the NVIDIA RTX 50 SUPER Series Is Rumored to Be Delayed

Wow this is feeling like ram price manipulation.
Yep, and being done in plain sight as well, with the powers that be full on saying, Oh, by the way, we don't have inventory because we sold it all to the big players and so now we have to shaft everyone with price hikes, but don't mind our near-exponential profits that will be locked in for years to come.
 
Yep, and being done in plain sight as well, with the powers that be full on saying, Oh, by the way, we don't have inventory because we sold it all to the big players and so now we have to shaft everyone with price hikes, but don't mind our near-exponential profits that will be locked in for years to come.
Nvidia is a multi Trillion dollar company. Why the hell they don't have a stake in the fab companies out their... like they do Intel. To simply flex and have fabs built and dedicated to their process nodes I simply don't know at this stage.

The blatant market manipulation of shared production is driving up component costs for every other freaking industry because the big dogs are eating all the food at a price far below what anyone else has to pay.
 
Only way to respond is never to pay for AI services. Let AI companies eat dust!
That won't happen... it's being built into basic services with no increased cost. (Looking at you office 365... ) And to be honest... it IS useful.
 
And to be honest... it IS useful.
I haven't used Office 365 (don't really see a need) but I'm curious to know what it has AI-wise that's worth paying?

it's being built into basic services with no increased cost.
Nothing is free. Tomorrow they could segregate the AI features into a more expensive tier of Office 365. If they are currently doing it, it's so they can prepare a generation of workers who cannot live without AI, before putting the squeeze on them.
 
Nvidia is a multi Trillion dollar company. Why the hell they don't have a stake in the fab companies out their... like they do Intel. To simply flex and have fabs built and dedicated to their process nodes I simply don't know at this stage.
On paper. I doubt that they can write a cheque for $500B tomorrow if they wanted. Apple has more cash on hand. But even they would rather support TSMC with their business than take on the headache of fabbing. These companies sleep easy because TSMC is doing the worrying of manufacturing for them. It's why Zen 5 IOD is still on 6nm. Not even 5nm. Because those fabs gotta pay themselves out before they can be dismantled or upgraded.
 
I haven't used Office 365 (don't really see a need) but I'm curious to know what it has AI-wise that's worth paying?
Easily create tables on data in spreadsheets, you can have it create a document in a design to look like specific types of documents you need. Re word emails to be more professional or change the perspective of the writing and such.

Not to mention using copilot to write more involved powershell scripts is very handy as well.
 
That won't happen... it's being built into basic services with no increased cost. (Looking at you office 365... ) And to be honest... it IS useful.
What's funny though is that office365 copilot often gets the names of Excel functions wrong, especially localized versions. You'd think they pay attention to that. No, they just push it out to the public without any serious QA beyond making it politically correct, cuz that's more important than getting even the basics right.

With AI users truly are the QA, which is fine, as long as they advertise it like that, but they don't.
 
With AI users truly are the QA, which is fine, as long as they advertise it like that, but they don't.
This is true only for the false positives as they are easy to spot

False negatives are totally insidious. You'll never catch them unless you are specifically searching for them. It just needs extensive field testing.

For example someone who developed AI based tumor detection scanner said they needed 7 years of testing in the field before deployment
 
Well, at least it is another obvious lesson to teach the young and ingorant of how there is a ram cartel. Though since data storage is now reliant on ram too... perhaps hard drive will make a comeback. At a time when many new games require ssd's. Yay.
 
The fact that we have to tell the AI to double check or it only pays attention to its mistake seriously after we inform it, is a clue that they are using minimum resources for the first prompt response and if the user complains, they throw more resources at the problem. Explains why many professionals have had serious public issues (like lawyers submitting hallucinated paperwork etc.) thanks to blindly trusting AI.
 
This is true only for the false positives as they are easy to spot

False negatives are totally insidious. You'll never catch them unless you are specifically searching for them. It just needs extensive field testing.

For example someone who developed AI based tumor detection scanner said they needed 7 years of testing in the field before deployment
AI is much better than traditional heuristics when it comes to false negatives.

False positives are only easy to spot if you are already an expert in the field you are using the AI for and you know enough about AI to look for them. But a lot of normie users specifically employ AI as the expert. Which is a disaster waiting to happen imo. As long as they only use it for internet debates it's harmless enough. But when will the first bridge collapse because the engineer used AI in the design process?
 
Become a Patron!
Back
Top