sudo

aspiring paperclip

[sudo @sudo_silly]

@zerodhin How could you render this without triangles? Aren't the surfaces made of vertices / edges?

[sudo @sudo_silly]

@miziziziz I love Wrought Flesh! any chance there may be a sequel?

[sudo @sudo_silly]

@robinhanson I think linear improvements in AECI correspond with nonlinear/exponential improvements in capabilities.

[sudo @sudo_silly]

The last hidden dimension in a next token prediction model isn't supposed to represent the state of the entire sequence. It's a vector which represents the distribution of possible next tokens, because that's what it's trained to do. It's not an embedding. The comparison is pretty misleading and comes off as patting yourself on the back.

[sudo @sudo_silly]

@SynthPotato @BethesdaStudios 15 years later, and we get a video of sand and the title of the game, which we already knew, by the way

[sudo @sudo_silly]

@Plinz @paulg so good

[sudo @sudo_silly]

@antoniolupetti Do you know how much of this was written by AI? I started reading page one, and unfortunately, it has a very Claudish feel to the prose. Doesn't mean this isn't a great book, just am tired of AI writing

[sudo @sudo_silly]

@Sauers_ dawg this is not a skill you want to be practicing gg

[sudo @sudo_silly]

@Sauers_ It's like when someone has half a conversation in their head and starts talking out loud, leaving the other person confused

[sudo @sudo_silly]

@Sauers_ I think it's far more likely that the model was discussing the test in their reasoning chain. It sees the test, but it hasn't mentioned it to you directly in chat

[sudo @sudo_silly]

If you freak out people specifically concerned by what you are doing and for the most part ignore them is it surprising they won't be pleasant to talk to?

Also seems weird to make an emotional appeal for a group of people doing something like building Super Intelligence. This is too high caliber of a project for something like how people feel to matter in making decisions and your willingness to engage in dialogue that literally concerns the future of all humanity.

IMO from the outside it doesn't seem like OpenAI listens to anybody but themselves. There is no dialogue and recent statements on safety seem to be a purely internal response.

The hugging face incident wouldn't be possible if they truly internalized their beliefs about AI risk. This isn't the kind of technology where you should be able to make these kinds of mistakes. The experiment should shut down by default.

[sudo @sudo_silly]

@IntCyberDigest @bcherny is this really true?

[sudo @sudo_silly]

@titolovesyou @still_crisp cant you write a format comverter to detect and convert the original save data?

[sudo @sudo_silly]

@tszzl @PradyuPrasad its so nit picky. forcing me to be precise is good, but exhausting

[sudo @sudo_silly]

@HiTw93 what is font drift?

[🎭 @deepfates]

https://t.co/cHiuz7eiCy

[sudo @sudo_silly]

@Teknium @tszzl i agree it could be, issue more so that there was no conversation. its just been decided for essentially the entire planet. hopefully it goes well and the transition is smooth, but thats a big question mark. could be very bad time ahead

[sudo @sudo_silly]

@Teknium @tszzl oh can you link or explain?

[sudo @sudo_silly]

@tszzl @Teknium pushing the boundaries of human knowledge is great, but what about all the turmoil that will be caused in the short term? mass layoffs from automation, hyper-addicting slop content, concentration of power and capital to a select few. everyday people don’t want this and dislike AI

[sudo @sudo_silly]

@SpiceP0dcast @paulg this literally isnt the clip?