Major, major rabbit hole warning. You think you're about to read something about a philosophy professor, and what you get is an Alice Munro/Larry McMurtry mashup. His son seems to be a pretty amazing writer in his own right.

If all the stars are aligned, Qwen 3.5 will not exhibit outright looping, although it will still burn more thinking tokens than some other models. There are ways to tone down the overthinking or disable it entirely, though, and the models are still quite capable when configured that way.

No personal knowledge of this stuff, though.

Then he went off the deep end, seemingly around the time when the guy in Thailand insulted his submarine idea. It became clear that he can control trillion-dollar companies but not himself. And, well, life's too short to spend it working for Nazis, nutcases, or both.

Not really, though? The idea that Earth-based data centers need to be built in populated, developed areas is indeed dumb, yet it seems to be inexplicably baked into everyone's assumptions. In particular, the small discrete data centers that Musk wants to launch could go anywhere on Earth.

They could be powered by local PV arrays and batteries, they can be cooled by smaller radiators than they would need to use in space, and they could be networked via Starlink or something very much like it, just as they would need to be networked in space. There's nothing special about space, it just costs more to get there.

If he wants them to be out of reach of governments, why not put them on container ships in international waters? There are thousands at sea at any given time, and I'm sure their operators would be happy to rent them out.

Hell, put them on dirigibles that just drift around in international airspace for months at a time. Anywhere but space.

And power density for solar is another.

Does power density matter in terrestrial solar applications? If so, why? These things can and should be deployed in oceans, deserts, and trackless wastelands. Who cares how big the solar panels are?

But, power density in terrestrial I think we can do some math and reasoning:

First, oceans are WAYYY more hostile than space. Oxidation + salt water + .. I don't think it's even close there. I don't think they are comparable.

Deserts and trackless wastelands - I have some experience with sub-Saharan logistics; a couple of points -- I would not be surprised if actual deployment to trackless wastelands is more expensive than lift. Analysts estimate $55k-85k per ton under starship. (Elon estimates much lower; let's stick with low end of analyst numbers).

Trackless wastelands are really hard to get to. For instance, I've seen a fuel truck tipped over on its side in a river next to a small tow truck tipped on its side in a river next to a larger crane trying to rescue the original truck and the "rescue" truck in Southern Kenya -- by no means a trackless waste -- probably a week long ordeal, JUST for diesel delivery. This was in an area under former British rule with roads and stuff.

Second, trackless wastelands are really hard to find. There are people everywhere, man. And they like free metal, free power, etc.

If we imagine instead just deploying to West Texas, I think the square footage does add up. 40 foot container -> call it 16 racks. Nvidia estimates 600kw per rack in 2027 with Vera Rubin(!!JFC!!). So, 10MW of power per container. Let's imagine we magically found water in West Texas and have a PUE of 1.2, so 12MW. Solar panels are like 20 W/sq ft.

I got lazy; Claude tells me with 2.5x land needed for spacing, infra, etc, 6.5 peak sun hours, a couple of acres for storage, roughly 130 acres (0.2 sq miles) + 53 Tesla megapacks for storage per container.

I'll revise my above thoughts - there is NO WAY it's cheaper to do that in trackless wastes than space. I don't know about west Texas, but I don't think it's crazy to think that you might want to spend five years on engineering and production scaling instead of town and county and state and federal permitting.

And the cost of building all the infra to support humans living in an area that humans are not already populating is enormous.

As far as I can tell from random articles online, it seems that as a rule of thumb, you need about 6 humans +1.5 humans per megawatt - and that's just for running the datacenter part, different people maintain the power generation infrastructure. Now, if you have to house those people in space or fly them up whenever they have to do anything, that's going to destroy your budget.

If you want to assume a level of automation that makes that unnecessary, that's fine, but then you need to also assume that same level of automation in earth based data centers too, and everything that goes with that.

Here's an idea, let's do this instead: we put them in the desert, or on boats or zeppelins or whatever, and we pretend they're in space. If anybody asks, those fuckers are in space, man. Computin' in the cosmos.

Copyright isn't "done", copyright has just been restricted to the rich and powerful. AI has essentially made it legal to steal from anyone who isn't rich enough to sue you - which in the case of the main AI companies means everyone except a handful of giants.

I use AI every day in my dev workflows, yet I am still easily able to empathize with those who did not intend for their code to be laundered through AI to remove their attribution (or whatever other caveats applied in their licensing.)

Disney saw which way the wind is blowing and invested over a billion into OpenAI

A compromise might have been possible, based on treaties engineered by the people who brought us the TPP, but nobody in the current US government is capable of negotiating anything like that or inclined to try. And it wouldn't exactly leave the rest of us better off if they did.

As a result, copyright is a zero-sum game from a US perspective, which matters because that's where the majority of leading research happens on the majority of available compute. Every inch of ground gained by Big IP comes at America's expense.

So they must lose, decisively and soon. Yes, the GPL will be lost as collateral damage. I'm OK with that. You will be, too.

Tech is becoming universally hated whereas before it was adored and treated optimistically/preferably.

there are binary files that some companies are allowing you to download, for now. it was called shareware in the old days.

one day the tap will close and we'll see then what open models really means

For my own purposes, open weights are 95% as good, to be honest. I understand that not everyone will agree with that. As long as training takes hundreds of millions of dollars' worth of somebody else's compute, we're always going to be at the big companies' mercy to some extent.

At some point they will start to restrict access, as you suggest, and that's the point where the righteous indignation displayed by the neo-Luddites will be necessary and helpful. What I advocate is simply to save up enough outrage for that battle. Don't waste your passion defending legacy copyright interests.

At that point it will be far, far, faaaaar too late.

> Don't waste your passion defending legacy copyright interests

The companies training big models are actively respecting copyright from anyone big enough to actually fight back, and soaking everyone else.

They are actively furthering the entrenchment of Big IP Law.

China: lol