According to Taiwan tech publication DigiTimes, most AI firms are unwilling to wait two years for HDD supplies to stabilize and are shifting to SSDs instead. To contain costs, they are choosing QLC NAND-based drives over the faster, more durable, and more expensive TLC variants.
Man, fuck those guys
Next year’s CES is gonna be every AI product from this year’s CES but with no progress on release date because they can’t even buy the consumer grade hardware to produce it since it all went into data centers lol.
I have one in my basement
the consumer hardware market is so tiny. It has almost no buying power or pull. Its hard to expect them to cater to it when its single digit % of their overall sales.
Back to punch cards we go. Or are they hoarding paper too?
“This just in. Panic buying of paper has led to shortages in timber, resulting in another round of toilet paper scarcity.”
Tariffs already gutted the barley starting to recover from the pandemic wood shortage.
“Why does the population hate us? We’re only completely destroying the consumer electronics market, accelerating climate change, aiming to eliminate countless jobs, increasing power costs, and stealing the works of millions of people to feed our system all so we can get even more obscenely wealthy? Please clap.”
Didn’t forget draining all their drinking water. Violating everyone’s copyright. Driving teens to suicide. Giving terrible medical advice. And Generating csam.
You forgot to mention that the product they’re sacrificing everything for is widely failing to meet expectations and that they will likely expect taxpayers to bail them out when their investments fail and threaten to take out the entire economy with themselves
Also in a few years when the AI generated nonsense code reaches a critical point and a ton of important systems grind to a halt, we’ll expect all of you that we fired to come back and un-fuck it for us so we can keep on making money. On temporary contracts and at reduced wages of course because times are tough.
To be clear, it only wildly fails to meet expectations in sectors that you hear about.
It’s most definitely medium expectations in sectors you don’t hear about because news and social media have a huge negativity bias because that gets views and engagement.
If we want to fight this scourge, we need to be more informed about it.
Like what sectors?
Contact centers, software development, automation, images and video analysis, data analysis, semantic search, entity recognition, advertising, misinformation campaigns, social media, security scanning & automation…etc
Many of these are cross-cutting across many sectors, some of these are sectors you don’t think of as they are driven by government entities.
And many of these have boring quiet tools and integrations that you don’t hear about because they “just work”.
You only hear about the shit that doesn’t work. Not the shit that does work.
Edit: inb4 a reply of a narrow use case or shitty implementation that, obviously, doesn’t work, which I already called out as a bias.
Law enforcement and military
I think it’s more like expectations have been deliberately lowered in those fields to meet exactly what AI can deliver. Unpredictable, arbitrary, non-negotiable decisions are the point, and the goal. It’s not about enforcing any laws or achieving any actual outcome other than making innocent people fear for their lives. And it’s doing a fine job at that.
Oh yeah I just read how AI keeps suggesting nuking everyone. That sounds like a great success. (I get what you mean, I just wanted to be a wiseass)
Doesn’t seem to be meeting expectations in that sector either.
I certainly expect that the visual models are meeting expectations in the racial profiling department.
I am start to think they are going on the acceleroism theory to push the society to the limits and destroy all for change
Some think this could lead to a more equallytary society, some wants to make a more centralized and controlled by one power sort of society, there is 2 different school of thought on this.
But we are clearly going to have big changes
Are now? Haven’t they been using flash first?
Hard drives suck for that work load. Their seek times are huge, their throughput is awful, and they’re worse storage density than HDDs. You can fit 24 2.5" ssds into a 2u server. 16TB in a 2.5" > 24TB in 3.5" SSDs were first to skyrocket in price vs HDDs.
No. AI companies have huge storage requirements for training data. Flash storage is not cost efficient for mass storage quantities.
For massive companies space is as costly as the price of the storage. Especially these AI companies don’t care much about the price of things.
Something tells me you’ve never worked for a massive company.
These aren’t massive companies. They’re startups. (and some branches inside of massive companies acting like startups)
… it’s Microsoft, Amazon, Google, and Facebook building all these huge datacenters. The biggest companies on the planet
Bruh, you’re the one who called them massive companies
A"I" companies want to store petabytes of “good” training data
Exabytes. Petabytes are easily surpassed by a lot of companies these days.
On the upside, we consumers get to have the superior TLC SSDs (better write durability).
Assuming that the chip manufacturers don’t drop TLC production to fill AI orders.
Let’s hope not
Compared to microns?
TLC is the technology, not the manufacturer. TLC means three bits per cell, versus QLC which has 4 bits per cell. Every extra bit the cell has to store exponentially reduces its write cycle.
