Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Introducing Copilot+ PCs

blogs.microsoft.com

Post reply on HN
Print view View original

141–150 of 614 posts

Re: Introducing Copilot+ PCs

#141

Post by grugagag »

rogerthis wrote: post #116
drevil-v2 wrote: post #87

Being the designated "tech" person for my extended family and friends circle, I don't think I could recommend this to any of them because of the privacy nightmare. At least with Apple you have a single vendor who is vertically integrated and makes a huge song and dance about data privacy. Even if you discount their PR and marketing spin, IMHO you are still miles ahead of the likes of Microsoft + (pick one) HP, Asus,…

As it has been said so many times, tech privacy aware people are the minority. It won't make any difference for my non-tech neighbor (in fact, he will probably be delighted if it helps him with something).

Not if you tell them beware, all this will be used against you in some way in the future.

Re: Introducing Copilot+ PCs

#142

Post by __loam »

croes wrote: post #85
milkshakes wrote: post #65

Earlier quoted context omitted.

getting a warrant requires probable cause, and a judge to approve it. maybe that's not hard, maybe it is, but i'd personally prefer more steps being required to access what data i have.

In Germany it's the same in theory, in reality even bogus causes are approved by the judges who neither the time and knowledge to really check the inquiry.

In the United States, protection from unlawful search and seizure has been in our constitution since the very beginning through the fourth amendment. This is a critical component of what we believe are human rights, so we will continue to insist that law enforcement should "get a fuckin warrant" if they want to dig through our personal data, regardless of your opinion on how easily that warrant will be granted.

Re: Introducing Copilot+ PCs

#143

Post by cjk2 »

If the following does not happen in 2 years I will eat my hat:

They will have performance problems. They will have compatibility problems. They will have poor repairability and zero repair network and support. The software will be abandoned and completely useless within 18 months. They will fail very early but just outside a standard 1 year warranty. This will be a lot of e-waste. Regulatory or national bodies will step in and force privacy regulations which make all of this unworkable.

Re: Introducing Copilot+ PCs

#144

Post by Dalewyn »

josefresco wrote: post #107

Are these suitable gaming laptops? I run local LLMs (Ollama) using my NVIDIA graphics card but I can't tell if these ARM chips are suitable for gaming.

I hope you realize that most of the gaming world runs on ARM CPUs and iGPUs today. Smartphones and tablets far outnumber gaming PCs.

Re: Introducing Copilot+ PCs

#145

Post by frde_me »

grugagag wrote: post #141
rogerthis wrote: post #116

Earlier quoted context omitted.

As it has been said so many times, tech privacy aware people are the minority. It won't make any difference for my non-tech neighbor (in fact, he will probably be delighted if it helps him with something).

Not if you tell them beware, all this will be used against you in some way in the future.

The problem is that there's no concrete thing to point to as to why this would be used against them in the future

(I agree with the point, but it's just not trivial to make people aware of this)

Re: Introducing Copilot+ PCs

#146

Post by Tankenstein »

coder543 wrote: post #81
phkahler wrote: post #77

To really be an AI PC, I'd want to be able to run something like llama 3 70B locally, and that's going to need a lot of RAM. Even running a 7B will take 14GB if it's fp16. So these really just need to be able to run a few MS AI apps. A quick googling tells me things like the Gen-AI in photoshop is actually done in the cloud, so that should still work but doesn't really require a special "AI PC".

No, 7B LLMs only need about 4GB of RAM. There is extremely little quality loss from dropping to 4-bit for LLMs, and that “extremely little” becomes “virtually unmeasurable” loss when going to 8-bit. No one should be running these models on local devices at fp16 outside of research, since fp16 makes them half as fast as q8_0 and requires twice as much RAM for no benefit. If a model is inadequate for a task at 4-bit, t…

The original commenter mentioned 70B, not 7B.

Re: Introducing Copilot+ PCs

#147

Post by joshstrange »

There are a number of aspect of this I find very interesting if only so that Apple has to come up with an answer to them (which I'm sure we will see some of at WWDC) but the "Copilot" brand is clear as mud... It's like that period they added ".NET" to bunch of things but way worse.

Re: Introducing Copilot+ PCs

#148

Post by nvtop »

drevil-v2 wrote: post #87

Being the designated "tech" person for my extended family and friends circle, I don't think I could recommend this to any of them because of the privacy nightmare. At least with Apple you have a single vendor who is vertically integrated and makes a huge song and dance about data privacy. Even if you discount their PR and marketing spin, IMHO you are still miles ahead of the likes of Microsoft + (pick one) HP, Asus,…

The whole point of NPU-enabled devices is to run models locally, so they your data never leaves your device. This is a huge privacy win.

Re: Introducing Copilot+ PCs

#149

Post by __loam »

beretguy wrote: post #136

I hope gaming on Linux keeps getting better. I don’t see myself using anything beyond windows 10.

Same. Our household has one pc from 2019 running windows, 3 on Linux, and one Mac, plus two macs from work. The only thing keeping me on windows is league of legends and stone art programs and I might get a mac to do that.

Re: Introducing Copilot+ PCs

#150

Post by jsheard »

nvtop wrote: post #148
drevil-v2 wrote: post #87

Being the designated "tech" person for my extended family and friends circle, I don't think I could recommend this to any of them because of the privacy nightmare. At least with Apple you have a single vendor who is vertically integrated and makes a huge song and dance about data privacy. Even if you discount their PR and marketing spin, IMHO you are still miles ahead of the likes of Microsoft + (pick one) HP, Asus,…

The whole point of NPU-enabled devices is to run models locally, so they your data never leaves your device. This is a huge privacy win.

Although the line can get fuzzy when they want to ship a feature that's too big to run locally. Android has run into that, some of the AI features run locally, some of them run on Googles servers, and some of them might run locally or on Googles servers depending on which device you happen to have.
Post reply on HN