hendrik

That was my question... How much on-chip memory do they have? And what are applications for that amount of memory? I think an image generator needs like 4-5GB and a LLM that's smart enough as a general porpose chatbot needs like 8-10GB. More will be better. And at that point you'd better make it unified memory like with the M-series Macs or other APUs? Or this isn't targeted at generative AI but some other applications. Hence my question.

Depending on the chip, they have somewhere from 100 to 400 GB/s. I'm not sure on the numbers on Intel processors. I think the consumer processors have about 50 - 80 GB/s. (~Alder Lake, dual channel DDR5) Mine seems to have way less. And a recent GPU will be somewhere in the range of 400 to 1000 GB/s. But consumer graphics cards stop at 24GB of VRAM and these flagship models are super expensive. Even compared to Apple products.

The people from the llama.cpp project did some measurements and I believe the Apple "Metal" framework seems to outperform the x86 computers by an order of magnitude or so. I'm not sure, it's been some time since i skimmed the discussions on their Github page.

Get a different hobby. Find some activity you're interested in. Then focus on that and slowly let go. And keep in mind what matters to you.

What kind of AI workloads are these NPUs good at? I mean it can't be most of generative AI like LLMs, since that's mainly limited by the memory bandwith and at this point it doesn't really matter if you have a NPU, GPU or CPU... You first need lots of fast RAM and a wide interface to it.

The Apple chips also have a wide interface to the RAM. That means you can run chatbots (LLMs) and other AI workloads that are memory-bound at crazy speeds compared to an Intel (or AMD) computer.

Ja, Englisch und Deutsch sind auch beides germanische Sprachen, sind also durchaus enger verwandt. Wobei ich finde, Englisch ist auch sowieso eine der einfachsten Sprachen zu erlernen. Man muss kein der/die/das mitlernen, die unregelmäßigen Verben sind finde ich ein Witz gegen Deutsch und Französisch (was ich mal in der Schule hatte), wo es ja zu jeder Regel ohnehin zig Ausnahmen gibt … Ich hab mal etwas im Internet herumgeschaut, die Leute sagen man kann B2 Deutsch so in circa 1-2 Jahren nebenher erlernen.

Software: https://github.com/awesome-selfhosted/awesome-selfhosted

Guide: https://github.com/mikeroyal/Self-Hosting-Guide

As a beginner you might want to start out with one of the all-in-one turnkey operating systems like yunohost.org , dietPi.com or unRaid or a bunch of others (see the awesome-selfhosted list)

That's a good question. I'd say be aware what you're ignorant about. Most people are ignorant. They have strong opinions about one conflict in the Middle East while they simultaneously ignore ongoing genocides in Africa. And my question is, does it help anyone if we argue with relatives or on the internet? I'd say it's alright to say I don't take sides, it's a shitty situation but not my field of expertise, so I don't have an opinion.

It's rarely a bad thing to be informed about things. And you always need information/education to make good decisions. Especially as a citizen in a democracy, it's your duty to elect your leaders, so you better have some idea about who's going to ruin the country and who's going to make it better. But that doesn't mean you have to know everything. And it also doesn't mean you need to blast your opinion out there.

And it's okay to be tired of US politics. Due to current circumstances. However, it shouldn't be that way. We learn about history and politics (in school) for a good reason. We're a part of the world and a part of what's going on.

I think that's overestimating the complexity. In my example you can just delete all data from people who cross the border regularly. I heard like >80% of Americans don't travel that much. So you'd still catch the vast majority. And there are additional giveaways. Visiting relatives will follow a pattern or coincide with holidays like every other thanksgiving. Weekend trips will start at the end of a week while work will be during the week and often someone would visit a worksite multiple times.

And correlating data and having multiple datapoints helps immensely. For example if you want to correlate license plates with cell tower data: One measurement will only narrow it down to a few hundreds or thousands of people who passed the highway at that point. But, a single additional datapoint will immediately give an exact answer. Because it's very unlikely that multiple of the people also return at the same time. Same applies to other statistics.

And you don't even need to figure out the patterns. It's a classification problem. And that's a well understood problem in machine learning. You need a labeled dataset with examples and ML will figure out the rest. No matter if it's deciphering hand writing, figuring out shopping behaviour to advertise, or something like this. We figured out the maths a long time ago. Nowadays it's in the textbooks and online courses and you just need some pre-existing data to start with. Maybe you're right and compiling a dataset will take more than 3 weeks. But it's certainly doable and not that complicated. And menstrual cycles follow patterns. That makes machine learning a precise approach. It'll home in on the ~4weeks cycle, find outliers and data that never followed a realistic cycle.

I agree, there are complications. People need to be incentivised to pay attention. Government agencies regularly fail at complex tasks. Due to various reasons. But it's probably enough to make peoples' lives miserable if they have to live in constant fear. So there is an additional psychological factor, even if they don't succed with total surveillance.

And this approach is a bit unlikely anyways. It's far easier to pass a law to force clinics to rat out people or something like that.

But my guess is that [predictive policing](https://en.wikipedia.org/wiki/Predictive_policing might become an issue. Currently we seem to stick to intelligence agencies and advertising with that technology (and Black mirror episodes and China). But that's mainly a political choice.

For using or not using loaded words? Generally, It's easier to criticize some words than to write down a long and nuanced opinion. You literally don't even need to read the attached article. And it's emotionally more rewarding to pick on things than write a comment that you agree. Also politics is an easy target for arguments and strong opinions. Try the same with gardening or the life of Johan Sebastian Bach and you'll see the same dynamics don't apply to some other topics. Unless someone writes something obviously wrong facts, that's going to be pointed out immediately.

Has that been tried since 1790 when the french decided to behead all the rich people?