

German soldiers can refuse orders under certain conditions (e.g. if they against human dignity)
German soldiers can refuse orders under certain conditions (e.g. if they against human dignity)
Thanks for the comment. I have had exposure to similar claims, but wasn’t seeing anyone using AMD GPUs for AI unless they were somehow incentivized by AMD, which made me suspicious.
In principle, more competition in the AI hardware market would be amazing, and Nvidia GPUs do feel overpriced, but I personally don’t want to deal with the struggles of early adoption.
For inference (running previously-trained models that need lots of RAM), the desktop could be useful, but I would be surprised if training anything bigger than toy examples on this hardware would make sense because I expect compute performance to be limited.
Does anyone here have practical recent experience with ROCm and how it compares with the far-more-dominant CUDA? I would imagine that compatibility is much better now that most models are using PyTorch and that is supported, but what is the performance compared to a dedicated Nvidia GPU?
Here is an exported result list from Kagi that should be accessible without an account.
deleted by creator
I have been using it for the last 3 months to expose services from my home internet (plex, wireguard, etc.) through a VPS and I’m pretty happy with it. It’s relatively simple to set up, I haven’t had any outages so far, and it’s nice that it supports UDP port forwarding as well as TCP (for wireguard).
You could go even further and use hard links. That way, you can have two paths pointing to the same data on the partition, with the space getting cleaned up only after all references to it are removed.
Hey URL, go and fetch your friend JSON!
Makes perfect sense.
I’m pretty happy with Digital Ocean if I need a temporary VPS because I can pay by the minute and the UI is great. Anything that I want to stay alive for more than a month or two, I do on a single 6-core VPS rented long-term from Netcup, a low-cost German provider, deploying with Docker and Traefik.
Right! The last I remember hearing the “closed source is more secure” argument was about fifteen years or so ago, so it’s surprising that it is being pulled up from the dead.
From the FAQ of the Sunbird website (the tech powering Nothing Chats):
Will the app be open source?
Some of the messaging community believes that software that is open source is more secure. It is our view that it is not. The more visibility there is into the infrastructure and code, the easier it is to penetrate it. By design, open source software is distributed in nature. There is no central authority to ensure quality and maintenance and by putting that responsibility on Sunbird, development would not be feasible. Open source vulnerabilities typically stem from poorly written code that leave gaps, which attackers can use to carryout malicious activities.
To help satisfy our own ambitious goals of providing total privacy and security, we are currently undergoing a third party audit that will validate our security, encryption and data policies and plan on receiving ISO 27001 certification after launch.
This was a huge warning sign when the first round of news about Nothing Chats came around, so I’m glad we’re now getting early confirmation that security by obscurity still is a horrible idea and doesn’t work
Thanks for sharing! I agree with your main point about overall emissions not changing too much since most of that reduction comes from feedlots already.
One small addition: the product that I originally linked is based on 3-nitrooxypropanol, a petrochemical-derived active ingredient, not from red algae (so there is probably a different calculation about production cost and CO2 impact than growing, processing and transporting red algae on a large scale).
Fun fact, there is already a food additive to reduce methane emissions from cows
You can sign git commits using SSH keys, including the one you use to connect to GitHub/GitLab/Codeberg. These sites also support verifying the signature.