Blaster M

deleted by creator

Exactly… the difference in energy use is like driving a Ranger vs. an F-650. Sure, the big chungus models are a lot smarter, but it’s diminishing returns on energy use to capability. Cutting out the fluff and using a much, much smaller and lighter model that is tuned for the task you are using it for is a lot more efficient, and people with macbooks or gaming computers can run a model on their computer instead of subbing to a cloud service.

Surely this won’t go wrong at all

This… this is the reasonable course that’s shunned by our divisive culture nowadays. I personally don’t have the time or the space, but I do have a computer, so I use what tools I have. Local models ftw

If it has Steven Hawking voice, I can hear the chaos in traffic

Explains a lot about how/why Grok leans the way it does

Virt-manager for a gui, qemu-kvm and libvirtd for the hypervisor. You will need to pass thru a graphics card, which makes it unavailable for the host machine entirely. I use debian stable for the vm, no gui, just ssh to manage it.

Ollama, Open WebUI, and Qwen, or one of the edits of it. The latest Qwen 26B is in a very good delta of smart enough to know how to help with stuff (I use it as a second opinion/initial proofread code checker) and small enough to run on a macbook or a gaming pc, and Open WebUI is the self host ChatGPT-like interface that works with Ollama, the turn key ai model runner backend. Couple that with your own searxng instance self hosted to give it the ability to search the web for answers if you want to give it that. Run it all in a virtual machine, of course.

For media content, wan2gp (wan2 for the gpu poor) has an all in one stop shop for self hosted audio, image, and video. Just beware people don’t like generated stuff in prod, but its useful for quick concepts, or to do certain types of edits of existing stuff.

There’s also tts and speech recognition via the whisper models, which sounds a lot better than the speak and spell voices from the past (unless you enjoy the degeneracy of spamming JOHN MADDEN JOHN MADDEN in the Steven Hawking voice) and can transcribe words a lot faster (realtime) and more accurately than in the past. The best ones run via Vulkan and thus work on any card that supports Vulkan.

If you run Nextcloud, you can also connect Ollama to it to get the Nextcloud AI working.

All that R&D, and they need to squeeze efficiency out more at this stage. Faster and more specialized is better than big chingus do everything, and means we can hopefully stop spending all that money on making more datacenters.

Personally, I don’t use the big chingus models at all… I much prefer local gen if I use it for things. Much better privacy that way, and I’m not throwing money at these jokers.

Subaru definitely uses the same hardware, but I know their software is different. Even the PCM performs very differently on the dinosaurs, even though they use the same Denso computers as Toyotas - just look into how aggressively Subarus fast idle when cold for an example… how many cars do you know of are programmed to hit fast idle come hell or high water? Most cars will cancel fast idle if you engage a gear, or limit the fast idle throttle, but the Subaru will send it if you try to keep the RPM below fast idle when cold. Also the “clear flood start” feature that deliberately doesn’t fire the injectors when you put your foot to the floor when starting the engine.

Interesting to see Subaru at the very bottom - barely anything, but Toyota is the opposite, sucking up big data like they’re General Motors

Edit: can confirm my car isn’t doing anything - when I upgraded my head unit, I discovered unconnected wires where a telemetry module would be

Unfortunately also true, and more likely than the Star Trek optimism answer.

WHO USES THEIR PHONE IN THE SHOWER

Science doesn’t really happen if we continually ask “Why do this?” Science happens when we ask “Why not do this?”

If we want to be a galactic civilization, we must first learn to live on worlds otherwise uninhabitable. The engineering challenges we must overcome to make that happen will give us the knowledge and experience we need to make it possible to expand past our solar system. This is a very long term advancent, and what we do today is what our offspring will do with ease.

Other than political roadblocks, the final roadblock to galactic expansion is FTL travel. We have theories on how to make it possible, but we are limited to the engineering and knowledge of our time now. Far future generations will be the ones to discover if and how to make it possible, but they need the foundational knowledge and experience of how to live in space.

This is the real tragedy of the modern data center. In the rush to be number one, they’ve captured the government at the top and use their influence to bypass environmental regulation and pollute at a grand scale just to save a dollar and some construction time. We could have had datacenters running on solar+batteries that could easily make money powering the grid as well, but no, we need to feed the oil machine and we let them run burner turbines unregulated. We could have had datacenters running closed-loop cooling, but no, we let them draw water directly for their cooling systems and dump it unfiltered back into the environment, further polluting it with heavy metals.

As long as one man desires to have more riches than another, this will never be solveable, I’m afraid. The shame.

Let me introduce you to the French phrase for perfume

In an internet funny, the year everyone was obsessed with the song Africa was also the year Africa was rolling tropical storms off the assembly line all thru summer (in the northern hemisphere)

That’s what the Nacelles do - radiative heat.

Nah, they just run on a starship-sized AI powered by an Antimatter Reactor