Профиль

Fair. My expectation for general population's understanding of most basic physics is always skewed. Growing up in academic bubble etc.

Are we tagging ourselves? I might be the most decrepit old bird, then. 😿

Net result are server hardware at several times the cost of equivalent on Earth, plus the cost of getting it into an orbit, plus basically not being cost-effective to do maintenance, so short life. Orbital data centers are a scam.

Radiation is also a problem. You don't need anything terribly complex, but some of the control systems need to be rad-hard and you need redundancy for error correction. So we're looking at about 2:1 on server capacity required for the same amount of compute.

Thermos analogy isn't quite right, as vacuum-facing walls are reflective, minimizing radiative heating/cooling. But the radiators an orbital datacenter would need are quite massive, and plumbing needed to carry heat from servers to radiators quite complex. It is a lot easier to cool on Earth.

Our brain mostly works with world representations that are 2d projections, simply because so much of it is based off of vision. So it kind of makes sense, really, but also seems like a bad design and I want to complain to developers.

My brain is stuck on "epic diarrhea."

The models used to solve certain classes of applied mathematics and biochem problems are more interesting. These are places where you're literally looking for a million monkeys with typewriters approach. These you can often use LLMs for directly. But something specialized would still be better!

The core tech behind transformers, which are part of LLM, has good applications, but because all the money is being thrown into LLMs, very few people are actually working on these core principles and applying them in the most constructive way. Which is still a net harm.

I'd even put in caveat that while predictive models can work to aid in translation, LLMs aren't it. Specifically, because the model's tuned to work with randomization and generate a *plausible* continuation, which can be more misleading than helpful in academic context.

Aww, poor kitty.

Did it rupture?

Of course, we already know that LLMs are not the right tool for automatic translation even when machine translation is appropriate, so that kind of makes the point academic. And if you're trying to research hate speech to stay ahead of it, it only underscores importance of professional translation.

While I don't think it's the point the twitter user is making, as they seem to be upset purely because they can't weaponize hate speech in other languages, there is a point to be made that a translation tool that has biases in what texts it will or not translate makes it a bad translation tool.

Great for mobility too! If you can time it right. 😸

The alt-fire for the grenade launcher in KOOK creates a gravity well. Great for pulling in clusters of enemies to more easily blow them up! Wishlist KOOK: store.steampowered.com/app/2329690/... #ScreenshotSaturday #BoomerShooter

And because it is the battle Moscow forces won, it is one that's widely taught in history classes. Now, if you look at where the Ugra river is located, south-west of Moscow, in modern Kaluga region, and the fact that this is XVc, it gets really hard to argue against the expansion of the Mongols.

I was never big on church and saints, but I know the Stand on Ugra in 1480 was a big deal, and had all sorts of consequences leading to formation of Russian kingdom around Moscow and consequently Russian Empire. I'm not sure how one talks about Russian history while pretending it didn't happen.

Russia not using phosphorus because they care about people claim is particularly funny. We're talking about Russia that shelled a Nuclear Power Plant in Zaporizhzhia? That Russia. It takes a very dense denialist to be able to take any *decade* of Russian history and argue for moral high ground.