unglaublich

That similar reasoning is there: using fission for power generation instead of banning fission altogther?

That's really a strange statement then...

Helicopter express transports only use 1% of the fuel of maritime shipping!

Mapping, legislation and licensing, marketing, existing taxi lobbies, consumer trust...

It's the same as encryption backdoors to stop the bad guys.

The bad guys work around it, and the rest is now in a vulnerable position.

Antrophic plays security theater by blocking their LLMs to work with security.

The bad guys work around it, and those that want to make their software robust against them are in a vulnerable position.

Luckily they made it safe to use so I can't hurt myself. Thank you Anthropic for holding my hand.

Cost will go up ~30%, so roughly 6.5 years instead of 5.0.

And for any non-trivial application, the space of possibilities grows so quick that you'll never even be able to _touch_ all the moving parts of the application and verify them.

And how are you going to determine which is the best? Going through all the possible combinations of users and usage? So mostly it shifts the work from generation to validation.

Why? What physical limitation will dictate that we can't have 1B tokens for cheap?

You turn the machines on when electricity is cheap, and turn em off when it's not?

Folks operating businesses that depend on oil prices would know these tricks?

Linear algebra can indeed not do it. You need non-linearity to get the expressivity that we see in LLMs.

Lol like EU chat control: politicians vote in favor, and politicians are excluded.

Most people are not, indeed. So this is the crux of religion: the "god" watches you, and will punish or reward you. Now social media and governments takes over that role.

Generally, electric networks benefit from economies of scale. So more customers will _lower_ prices per-customer in the long term.

Frontend developers must do this to their browsers too. Sure, your M5 Max or i7 on a 1gbit connection runs things well... but does it run well on a mobile cpu with flaky 4g?

What a ridiculous take. Just because solutions exist in the llm training data, doesn't make these problems 'toy' or 'easy'. The human 'engineering hardness' scale doesn't align with what an llm can and can't do.

Convenience. Want to run `docker run ...` without password, want IDEs and agents to be able to run containers...

This is why you need either a rootless container setup or user namespaces to remap the container user to irrelevant host users. https://docs.docker.com/engine/security/userns-remap/

Weak that this isn't the default.

The usual question: what's "aligned with generating alpha" that a human stock trader can do, but an ai can't?

It'll be shit for paying Linux customers too as a large user base of "free"-version Linux users could spot and report bugs easily.

Did we get days off when industrialization bumped output? Days off when industrial agriculture made food supplies abundant? Did we get days off when computers automated administrative office tasks?

That's how the design is supposed to work. But marketing realizes that no one voluntarily receives ads, so they mix em.

No one willingly says "yes" to advertisements, but people will say "yes" to important-updates(-and-advertisements).

So I guess we'll all be applying a random rotation to our matrices now to obscure their contents, like TurboQuant does. https://arkaung.github.io/interactive-turboquant/#rotation

Isn't any other significant economy willing to do business with anyone that can pay?

It's a bad short story. Whether it was generated at the hands of humans or a computer isn't really relevant. Speaking in riddles is cheap now, so authors better learn how to surprise the reader in simpler, more readable words. It seems to be something that LLMs are quite bad at.

Indeed, at 30tok/s make it pause for 20 seconds while "thinking" is streaming (and hidden); that's the real experience.

30tok/s looks fine when you're just streaming code, but the issue is that there's a lot of background noise like tool-calling conventions, metadata, "thinking", etc.

Light and moderate are mostly just "activities of daily life" (walking, commuting), and vigorous is whenever you exercise explicitly (running, swimming, speed cycling, soccer, etc). So it's more like 9 hours of active movement, or 4.5 hours of exercise (40 mins a day).

It's to mark the images for exclusion during training.