Soyweiser

joined 1 year ago
[–] Soyweiser@awful.systems 3 points 4 months ago

Turns out the story is even weirder, Durov wasn't always a 'everything is allowed' free speech guy. Iran must be defended

[–] Soyweiser@awful.systems 8 points 4 months ago* (last edited 4 months ago)

Unrelated to the posts, but in Dutch beffen is a somewhat vulgar verb for going down on a woman. Based Beff Jezos indeed.

(I'm sorry im 12 years old).

[–] Soyweiser@awful.systems 17 points 4 months ago (3 children)

I want to make a HN account and reply to the 'depression and schizophrenia is or isn't a hardware/software issue' posters and tell them there is actually is a working IQ test, and they all just failed it.

[–] Soyweiser@awful.systems 7 points 4 months ago

Yeah that they named it Q* and that they are very online (so they know about the implications) is quite worrying.

[–] Soyweiser@awful.systems 7 points 4 months ago

Why are you thinking he isn't running the startup by doing this? Number needs to be high!

Musk has shown that you don't actually need to deliver for decades if you just keep saying 'soon a thing!' every time there is a negative quarter.

[–] Soyweiser@awful.systems 7 points 4 months ago

Found some tweets (randomly I wasn't looking for them) of one of the people in the article above driving themselves mad. See the second tweet here. Oh no, the chatbot said it is the master of all!

[–] Soyweiser@awful.systems 12 points 4 months ago (7 children)

Finally it is today on the AS. So I can post my link. The AI Guys Are Driving Themselves Mad (nymag link)

[–] Soyweiser@awful.systems 10 points 4 months ago

Even Moot knew you couldn't do this. Congrats on being slower on the uptake than 4chan HN.

[–] Soyweiser@awful.systems 8 points 4 months ago

I was wrong, Manmade horrors...

[–] Soyweiser@awful.systems 11 points 4 months ago (1 children)

Yes, and then the problem is that this doesn't really scale well. Esp as it is always hard to regexp all the variants correctly without false positives and negatives. Time to regexp html ;).

[–] Soyweiser@awful.systems 23 points 4 months ago (3 children)

lazily regex

I'm having a sneaking suspicion that this is what they do for all the viral 'here the LLM famously says something wrong' problems, as I don't think they can actually reliably train the model it made an error.

view more: ‹ prev next ›