TechTakes

fucking christ. it takes a lot to fuck up my day, but a quick scroll through that thread seeing how quick these vultures (including one notable person who’s the reason why I’m ashamed to talk about my lambda calculus projects) are trying to capitalize on Vernor’s legacy is absolutely doing it

HN wanted a black bar[1] but were denied.

why in the fuck? is the famous sci-fi author with a heavy CS background not notable enough for the standards of the site whose creator is a much less notable self-help author whose CS background is failing to make a working Lisp 3 times and writing programming textbooks nobody reads?

The replies to somebody aggressively (and downvoted) pointing out some of the flaws in Jarts post are bad. Damn.

an LLM cannot be used to create a better LLM

By that logic most humans are also not intelligent.

No you dweeb, they are talking about model collapse, that thing what happens to this 90's tech.

Oh, it doesn't work? That's because IT'S NOT INTELLIGENT.

Ok, let's run this test of "real intelligence" on you. We eagerly await to see your model. Should be a piece of cake.

This is both a weird adhom and a god is hiding in the gaps style argument. (While I have some sympathy for this Peter Watts style argument it is incredibly weak (their post history (8) is more of this very weak stuff)).

Edit: forgot to mention what I actually initially wanted to say:

Idiocracy

I still think that movie actually is quite hopeful, it shows us a world where the current consumerist society in the USA can remain to exist for 500 years (that is how long he is frozen), and there is nobody invading and taking over when the USA becomes this automated and dumb. We should all hope the future is this hopeful (guess they fixed the climate, and achieved fully automated luxury capitalism (which still sucks)) and non-violent.

the old thing was use of ai in iron dome, because generally mistaking a civilian target for a rocket is pretty hard given differences in size and speed, and autonomous air defense is a thing that already exists for decades. the ai part comes from predicting if a given rocket will fall on populated area, and pointing radar-guided interceptors to those that do

the new thing, well nobody can tell you that you're doing things wrong if nobody knows for sure what are you doing. to even tell whether it's working or not you'd need to sit in heads of israeli military planners and know what are their exact objectives and acceptable collateral damage

people at palantir are probably making very detailed notes