That wasn't *all* I said, but it is apparently as far as you read. But let's stay there for now. You apparently disagree with this, whnich means that you think that LLMs are the only kind of AI that there is, and that language models can be trained to do things like design rocket engines.
It's been, what, a week, and they got 6000 jets through maintenance during a very busy holiday travel season?
That speaks to fantastic logistics. There are anonymous folks out there who did a great job and deserve a ridiculously generous Christmas Bonus.
This isn't true. Transformer based language models can be trained for specialized tasks having nothing to do with chatbots.
That's what I just said.
Here's where the summary goes wrong:
Artificial intelligence is one type of technology that has begun to provide some of these necessary breakthroughs.
Artificial Intelligence is in fact many kinds of technologies. People conflate LLMs with the whole thing because its the first kind of AI that an average person with no technical knowledge could use after a fashion.
But nobody is going to design a new rocket engine in ChatGPT. They're going to use some other kind of AI that work on problems on processes that the average person can't even conceive of -- like design optimization where there are potentially hundreds of parameters to tweak. Some of the underlying technology may have similarities -- like "neural nets" , which are just collections of mathematical matrices that encoded likelihoods underneath, not realistic models of biological neural systems. It shouldn't be surprising that a collection of matrices containing parameters describing weighted relations between features should have a wide variety of applications. That's just math; it's just sexier to call it "AI".
I too bought memory in April to avoid tariffs. I had to run a stupid python program to generate a dataset that required 96GB of RAM for a delayed project so I figured I might as well bite the bullet. DDR4 was still a good value at that point (it's a problem that can run overnight, performance wasn't too important).
But how are the tariffs limiting the manufacturing supply capacity of RAM factories in East Asia?
Do you have a mechanism to propose?
Do you think they're making enough to meet demand but then blaming tariffs to justify jacking up prices? All of them? It would be an interesting conspiracy but is there any evidence to support that theory?
> How much is this problem is down to AI and how much to beautiful tariffs?
What mechanism are you thinking of where tariffs could limit supply of VRAM from East Asia?
Simple price increases, sure, definitely, but this is described by manufacturers as a supply & demand problem.
Do you have a different angle we should consider?
> It did happened before, but not on this scale and speed.
Check out Meltwater Pulses 1a and 1b.
> It used to be my go-to site for all things computer related.
Me too.
They were slightly cheaper than Amazon for the same product, then I did a big project which got slightly downsized and I wound up with $400 in "restocking fees" for a couple of pieces of factory-hologram-tape sealed network gear, after I paid $100 in return shipping.
Learned my lesson real fast.
Them: Why don't you act your age?
Me: Well, I've never been this age before.
Like buying booze, renting a car, purchasing a handgun, buying a lottery ticket, getting a tatoo?
(some of these vary by state)
I don't see how you're too immature to order a Chianti with your steak dinner but you're mature enough to go $200K in debt based on a sales pitch of returns after investment.
These aren't even reasonable equivalents from a neuroscience perspective.
A read is supposed to be fine. At read time the firmware *should* rewrite the cell if the read is weak.
The firmware also *should* go out and patrol the cells when idle and it has power.
you can dd if=/dev/sdX of=/dev/null bs=2M once a year if your firmware behaves.
If your drive is offline you could
dd if=/dev/sdX of=/dev/sdX bs=2M iflag=fullblock conv=sync,noerror status=progress
to be sure, though write endurance is finite.
If you're running zfs you can 'zpool scrub poolname' to force validation of all the written data. This is most helpful when you can't trust the firmware to not be buggy crap. Which only applies to 90% of drive firmware out there.
Well 0-clicks and OTA attacks but yeah, to your point, device compromise lets you use apps on the device.
News at 11!
> I have found that streaming directly to my Plex home server over TLS is generally smoother without going through Wireguard. Not quite sure why.
I recently had to solve this.
Wireguard should work with a regular 1500byte MTU connection at 1440 or 1420 bytes (the default) --- however --- if your ISP is routing your IPv4 using 4-in-6 internally (like my major cable company) everything goes to hell.
Try dropping your wg MTU to 1360, MSS at 1320, and set up a mangle table to clamp MSS to PMTU (e.g. iptables rule).
I got a 10x bump in TLS over wireguard throughput.
Total pain in the ass and lightly documented.
The illusion of intelligence evaporates if you use these systems for more than a few minutes.
Using AI effectively requires, ironically, advanced thinking skills and abilities. It's not going to make stupid people as smart as smart people, it's going to make smart people smarter and stupid people stupider. If you can't outthink the AI, there's no place for you.
I upgraded from a Pixel 5 to 6 a couple years ago due to the unfixable OTA hardware security bug in the 1-5 models.
Some days I'd like more storage but really I can't see what else a newer device would get me. LineageOS 23 is asking me to upgrade to it; I guess I should after the December patches land.
Build a system that even a fool can use and only a fool will want to use it.
