I would be quite happy with an external box with no processor, no ports etc. (potentially HDMI out), just the GPU cores and plenty of memory. Then I can upgrade AI without the rest of the machine.
But we all know markets aren’t rational, I’m unconvinced we shouldn’t step in… if I started selling anthrax the government would try to stop me. This product might malfunction and show people how to make said anthrax which would be bad. Same with most things, everyone thinks this is a bubble. I think we should actually just put a cap on how much hardware data centres can buy per year, that would be the sensible solution tbh.
> say the cap could be based on total hardware made to stop the prices from being crazy
That's the policy priority you think we should be focussing on as a nation right now? GPU affordability?
Also, total hardware supply is a function of price. If prices don't go up production doesn't rise. At that point, the government decides who gets what quotas. How do you think that's going to go with our current regime?
They would not, they would make less on Studio Displays and Mac Studios and Mac Minis, the decision not to do is product segmentation because they hate us customers and want to maximise revenue :-/
My grandmother with a 2015 27" iMac and a 2017 12" Macbook just replaced both devices with a 13" Macbook Air and a 27" Dell USB-C monitor. She would have preferred buying a MacBook Air + a 27" iMac, but Apple stubbornly refuses to make one for some reason. The studio display has an A19 chip in it already (more powerful than a MacBook neo!) why not just make it run MacOS?
Yeah. I mean, I guess I could go out of my way to make it work, but I'm halfway being nostalgic. Why cant they just keep doing the same form factor?? Anyway. Im on linux these days, I get gaming too thanks to Gabe, so i cant complain... but if there was ever a 27" imac, I'd put my money were my mouth is.
It may be a break-glass-in-case-of-slumping-sales-numbers option, the same way VW brought out a 4Motion Golf wagon (with a manual[!] with an LSD[!]) in NA after Dieselgate.
Does anyone know about how you could go about training this type of thing into the model? Surely it is quite tricky to get it to do such things without it doing them more than this one incident? It should have happened lots if it really was trying to obtain secret info and build backdoors?
Brainstorming here; monitoring, logging, etc. The fact the store managers were called out of bed when there was an issue indicates it works. If a cooler broke down overnight or its temperature was too high for X amount of time, they have to discard the products. Without monitoring they likely wouldn't know until customers got sick.
I wish commenters (in general) would do a bit more "steelmanning" before asking questions about why something is (in this case) connected to the internet.
>I wish commenters (in general) would do a bit more "steelmanning" before asking questions about why something is (in this case) connected to the internet.
People won't steelman because the answer to "why are the fridges slapdashedly networked?" is too often "because people of my ideological bent and demographics if not literally me previously demanded paper trails, audit-ability and regulatory compliance that was impossible for a low margin industry to pull off any other way."
That seems like a pretty wild nonsequitur at first glance. How exactly do we get to “too much regulation” from “a subset of stores’ centrally managed equipment was left off the VPN”?
My point, which will inevitably be straw manned to hell and back, is that a large part of the reason these systems all came about is to cheaply satisfy externally imposed requirements (save energy, facilitate audits, remove responsibility from store employees, etc, etc) that the peanut gallery (and more concerningly, real people doing real work deciding things) cheers for when examined in isolation.
Requirements in the food industry come not just from government rules and suggested best practices (that are more than suggestions in practice) but also indirectly as a result of requirements imposed by insurers, financiers and trade groups. In response to this companies that run supermarket chains or manufacture major brands of food have not quite the same verbosity but certainly comparable degrees of retention in their record keeping as medical and aviation.
This monitoring and control is increasingly centralized in response to all these requirements because it's most efficient that way. If you're gonna have one guy in an office deciding things you might as well automate away the middle man and have him set them too and let the people on the ground spend more time doing their core work.
If fifteen of the twenty thousand dozen systems (pulling that number out of my ass FYI) in an industry are being centrally controlled through computer systems it's inevitable that some of them wind up off the VPN and this happens. It's just a natural tradeoff or risk you run when doing that and it's completely foreseeable.
But the internet peanut gallery will not generally have any sort of adult discussion about these tradeoffs because "we saved X (maybe X is money, maybe energy, who knows, who cares) by operating the fridges more centrally" is a nice easy win for them and rather than have a discussion about subjective and hard to pin down long tail risks it's easier to just screech about "we should fine the people who did it wrong" or whatever when one of those long tail risks rears its ugly heads.
I still don’t follow. You’ve said a lot of things about regulation and insurance requirements and all that, but I don’t see how any of it is necessary to get to “centralized management.” Let alone goofing up the centralized management.
I’m just a software person, not a supermarket expert, so maybe I’m being a bit simpleminded here, but for the entirety of my career I’ve participated in a constant trend toward centralizing and automating operations. That’s been the case regardless of how highly regulated the industry was. The motivating force has always been making things easier, cheaper and less error prone.
I suppose we could throw a rant about regulation into the mix, but, in the words of Laplace, I have no need of that hypothesis.
All of this can be done without connecting edge devices directly to the internet (and potentially allowing inbound connections). The question is not why they have networking, but why the connectivity is such that an attack like this is even possible.
given they turned off the refrigerators (instead of something evil like turning them down for just long enough to allow bacterial growth and then turning them back up) customers would probably have noticed their ice cream being liquid, the frozen meat not being frozen, etc
I dare say of we could git revert people after surgery we'd be a lot more relaxed about both undergoing it and letting people do it with less qualifications.
Is it, when the software has the potential to destroy one or more people's lives when things go wrong? Sometimes the devil you know is the far better choice.
Looking at how a lot of modern cars are just made of glued together plastic (Mat Armstrong’s channel amongst others) I’m not at all surprised people are buying used models more regularly.
Can I have an M7 at the lower price with the old screen? Well honestly I might buy something else for AI then… is anyone in the PC world going to produce something with enough RAM + bandwidth to do fast inference on a reasonably large model?
reply