cryptophreak

This declaration seems to be epistemically confused.

On one level, they admit that finding solutions and generating understanding are non-overlapping magisteria. They use this admission to make the point that AI models are not generating understanding by finding solutions.

But on another level, they assert that AI finding solutions somehow breaks the process of generating understanding.

In other words, they simultaneously do and don't understand the non-relationship between finding solutions and generating understanding, and they choose which mode to operate in depending on which helps them score a point against AI.

The only reason we have historically perceived a link between finding solutions and generating understanding is that the problem space was larger than human cognition allowed us to search efficiently. So we pretended, in a "security through obscurity" way, that finding a solution implied understanding. It's inconvenient that we can no longer do that, and institutions will need to adjust in response, but it doesn't break the foundation of mathematics.

How does the answer to that question relate to the article?

Can you explain why a low-hanging optimization that would reduce costs by 90% without reducing perceived value hasn't been implemented?

That would be even better. It would take longer than an evening, though.

That’s one of the great things about the approach demonstrated in the post. The developers of Handbrake don’t need to invest any time or energy in a minimalist interface. They can continue to maintain their feature-rich software exactly as it is. Meanwhile, there is also a simple, easy front end available for people who need or want it.

Selecting any text in the body of the article adds that text to your clipboard. Weird.

Technology may be more predictive of conflict abroad than at home. If we faced a land invasion, for example, we would not be able to bomb our way out of it.

Probably because one man's important details is another's gibberish.