You can absolutely connect it to git. It's janky and everybody else's edits will show as a single commit created, I think, whenever you pull, and with a title "update on overleaf" or something. Yet, once it's in git, it's workable.

I'm also in academia, and I have one important correction to the OP: MORE BUILDING FROM SOURCE!!! It's never been so easy.

My personal no-go property of a platform is the login wall: I don't want to demand a platform account from my potential readers.

Twitter had many last drops for different people, but the login wall a couple of years ago was mine. I've been avoiding LinkedIn for the same reason.

My personal no-go property of a platform is the login wall: I don't want to demand a platform account from my potential readers.

Twitter had many last drops for different people, but the login wall a couple of years ago was mine. I've been avoiding LinkedIn for the same reason.

Well wait until you run out of code to write because all code has been already written.

Please don't misrepresent all criticism as bad faith. Please also don't reply-by-quote.

I looked through the paper and saw both where it explains the theorem and how it discusses its implications. I have also explained why it appears insufficient. You did not react.

bsky.app/profile/iris...

Your thread claims a "fatal conceptual error" in AI-as-engineering without qualifiers. The abstract or box 2 do not mention the distinction and seem to apply to AI-as-engineering. Your screenshot states: "for instance they have access..."—that's not a restriction of scope.

Do you agree?

Ah, I see that you do already consider scientific method uncomputable for similar reasons. bsky.app/profile/iris...

I can't help but think that there's a concept swap between "impossible for a P-time algo" to "possible for a human because humans happen to do it".

AI-as-engineering would also support a definition in terms of complexity theory (Dr. Ingenia runs an experiment and tries to find a simple model for an AGI), and that one would also be NP-hard—fatal error again!

This formulation, however, would be a "fatal error" of the entire scientific method 🧐.

It seems that the fatal conceptual error only applies to AI-by-learning, rather than AI-as-engineering, is that right?

From the manuscript I understood that one is about specifically learning from samples, another is about any approach to a computable representation as a matter of principle.

Also pixi global? 😉

There's still a question of how long it takes until the leak happens. I'd say that www.vice.com/en/article/m... is a comparable privacy violation, and yet it did happen.

It's surprising that you didn't consider all the creativity you could have invested into the task. And all the software out there from which the parser script was plagiarized.

If reputation isn't high, then neither are the stakes.

That may mean that a bunch of students don't actually engage.

So far I plan to let students propose 1-2 sentence additions to the list, and for others to vote for contributions.

Then at the end, I assemble the sheet from the most voted contributions.