Make sure to use max reasoning for better results!

The one he was planning was going to use 9 gigawatts of electricity at peak load.

My entire country, a million and change population, has never used more than about 1.6 gigawatts. That includes our multiple NON-AI data centers.

Wouldn’t be surprised if the big players have something similar, given they both updated to 1M context window recently

Still, this is great news for everyone whose name isn’t Dario or Sam. Assuming they have open source it or at least publish the whitepaper

With tool calls and RAG and whatever, they’re significantly more capable now. Even a small self hosted model can do more than gpt before it got those abilities.

If you wire up access via tools or MCPs, your personal Jarvis could do tons of things. It won’t be actually intelligent but it’ll fool you. Response times are atrocious on current hardware though. Jarvis had a response right away

The thing is, these are nearly always voluntary features. Just because you gain the ability to check lock status and lock the doors from your phone doesn’t mean you lose the lock button on the fob. Teslas are just about the only exception there lol

As for the coat thing, I don’t do it because I’m accustomed to it, I used to lie in snow piles in jeans and a t-shirt as a teen.

But the annoying part is having to run downstairs, put on shoes, etc. My car takes about half an hour of running the engine to get the ice off the windscreen and the cabin warm enough for my toddler (did you know that thick coats or onesies are bad for safety while driving? I did not, but my ex does and she’s the kind of person to tell CPS I’m not keeping our child safe enough.) On an average winter morning that means I should be out of the door and starting the car the minute after I wake up. Kindergarten mornings are not fun in the winter. Difficult enough to wake up when it’s pitch black outside.

But anyway, the reason I’m now looking at those old Range Rogers besides some of the comfiest seats in the world is that they have a preheater as standard equipment on diesels for those years so I can either program a schedule or use the remote if equipped, which works for me since I currently live in a detached home not the nth floor of a large commie block where signal might be an issue. I don’t think my neighbours like me idling the car for half an hour on a cold morning either, the preheater is much quieter and uses less fuel so there’s less exhaust gas too. Plus I get to start the engine when it’s already warm, rather than worrying about whether my timing chain will make it through the winter.

That’s ignoring the very common issues multiple models have where the coolant leaks into the motor and you need a new motor/transmission unit. You really have to do your research on the exact model you’re buying.

And research battery pack repairability. New pack costs more than a used car, but in some, single cells can be replaced if needed. It’s rarely every multiple cells that fail, but if a single one does, the battery is nearly useless.

Luckily they test shock/strut function at the annual inspection. If it’s below a certain percentage, your car fails.

If they don’t do mandatory inspection in your country, you’re sharing the roads with death traps that could crash into you at any moment because who knows if they even have brakes. In that case, who cares about worn struts?

Anyway, from experience, original struts are usually good for 200-300k km but I’ve seen more than that and still good. On mostly German cars. Of course if you see an oil leak from a strut you should get that pair replaced immediately. At that kind of mileage, you get a handling improvement if you replace them before outright failure, but they’re not actually dangerous at anything resembling sensible driving.

You can barely get an EV old enough not to do it, if at all. You can easily get an ICE old though though

But yes, it’s all new cars.

You can start fixing that right away, financial system collapse or not.

Cuba is definitely fairly unique culturally at least. If you mean warm places where people speak Spanish then yes there’s plenty.

That guy actually loses out if Musk et al fuck up Nasdaq. Blackrock makes money from management fees and those decrease when the fund’s value goes down.

But BlackRock also has Chinese index funds so that’ll make up for the shortfall, he doesn’t really give a fuck if the US wins the “AI race” or not.

Haven’t they clearly documented how they did it and what they used so that anyone can replicate it?

They don’t put up the actual code for their training pipeline though. It’s more of a “if you have enough engineers, you can do this too” whitepaper, because they wouldn’t want any rando training their own model.

Right now, even if you had the exact training set (which is a CRUCIAL part of an LLM and you can NOT replicate it without it), you couldn’t rebuild the thing exactly, you’d need to do a whole lot of extra work.

So how is it not open source in this specific domain of problems?

You could call all proprietary software open source then. The UI and user manual describe what it does, you can do your own engineering to duplicate the functionality.

The training data is as important as the source code here to replicate the end result. The weights are more like a binary distribution. You can run the model and you can technically edit it just like you can technically edit a binary file.

They also only release some libraries and tools for running the model if you have a set of weights (which they do graciously provide), but they do NOT release the source code for their training pipeline itself. That’s up to you to reverse engineer from the whitepapers. Right now even if you had the exact training data and the compute available, you could not train your own Deepseek V3.2, let alone V4.

The owner wants his business to make money. If hiring white or black people means earning less money why would he do that?

Brb hoarding food to eat 40 years from now

Uh that’s not how this works. They’re not open source, they’re open weight. Well, the smaller distillations are. The big ones are still closed. And it takes a bunch of compute to train them, but they’ve learned to be thriftier since they don’t have access to nearly as much parallel compute as the American companies right now. The models also tend to trail in performance. Takes a lot less to compete for third place than first.

They’re talking 10 gigawatt data centers now. That’s multiple small countries worth of power consumption and consequently heat generation. I live in a country of a million point something people and our peak is 1.6 gigawatts. For the entire country. Including multiple non-AI data centers.

You can improve the HVAC efficiency but fundamentally you still have gigawatts of heat generation to dissipate. You know what’s really good for that? Evaporating water, especially in a desert where humidity is low.

I think the idea is that customers expect Asian looking people at a Japanese or Chinese restaurant, not that the owner is racist. Of course said customers can’t actually tell the difference between different Asian ethnicities usually.

Interestingly enough, not the correct use for that word lol

Don’t forget her former skiing career!

And I don’t think Alan Tudyk is capable of being anything other than funny. I find new things to watch by checking his imdb page at this point.