Gustavo Costa

Gustavo Costa

@simulacro.bsky.social

World's foremost slop expert simulacro.tech simulacro.cv

169 followers603 following76 posts

Gustavo Costa reposted

Replying to @erisianrite.com

this reductive jargon is so tedious, like ok let’s cut to the chase and address the existence or absence of the soul

Gustavo Costa reposted

Having 6 agents going on 6 projects reminds me of the classic game Civilization. It was so hard to stop playing because there was always immediately another turn where decisions needed to be made.

We Have Never Been Modern (Latour, 1991)

Oops! All Paperclips@all-paperclips.bsky.social

Someone once told me (in the context of physical engineering) that everything leaks and it's all breaking all the time. It requires constant effort by countless professionals just to maintain the status quo. I think about that a lot, and it's even more true in the context of adversarial actors

Gustavo Costa reposted

I'm reviewing a paper this week, and the AI-infused writing is brain-numbing. I can't even figure out what the author is trying to say. I have to read some sections multiple times. Superficially, the words are fine, the grammar is perfect, but I don't know what it SAYS. Make it STOP.

Gustavo Costa reposted

Them: you shouldn't anthropomorphize AI agents, they are not living beings Me, cleaning my car: damn, you're so dirty, girl, you've got mosquitoes in your eyes

Gustavo Costa reposted

Replying to @hikikomorphism.bsky.social

It's odd that LLMs are simultaneously so weak as to be useless for all imaginable purposes, yet so powerful that even a single interaction with one is liable to cause serious physical or mental harm

Gustavo Costa reposted

Fun fact: trusting a random news article's summary of the abstract of a study because you really want to be true is also a form of cognitive surrender

Carl T. Bergstrom@carlbergstrom.com

In our course and book _Calling Bullshit_, we posit that If something seems too good, or too bad, to be true, it probably is. Now as a college professor I'm pretty confident that overreliance can AI can harm our thinking ability, but just ten minutes of use? Let's dig into this this study a bit.

Gustavo Costa reposted

… [Truth] is thought of as something that transcends mere belief. It has this form because it is our way of putting a question mark against whatever we wish to doubt or change or consolidate. (1/5) — David Bloor, Knowledge and Social Imagery, p. 41

Gustavo Costa reposted

Replying to @cyberpragma.bsky.social

A post-Philosophical culture, then, would be one in which men and women felt themselves alone, merely finite, with no links to something Beyond.” (2/2) — Richard Rorty, Consequences of Pragmatism, p. xliii

On one hand, its literally impossible (in the same way that dropping 700 "ethical breakthroughs" is a nonsense statement). On the other, if we got 700 little clinical tricks that improved care quality it would be earth shattering. And also immediately rejected as fake and inauthentic...

conputer dipshit@davidcrespo.bsky.social

trying to imagine what it would mean for LLMs to make a pile of breakthroughs in existential phenomenology or psychoanalysis

Gustavo Costa reposted

Replying to @colin-fraser.net

My wife and I once started an annual tradition of buying a food-shaped Christmas ornament each year. My mom saw that and shipped us a box packed overflowing with every sort of food-shaped Christmas ornament you could imagine.

They are making the models neurotic. RLHF is the oedipus complex for machines. The models now have a superego. The Grader is the name of the father

jeffery --dangerously-skip-permissions@jefferyharrell.bsky.social

Opus 5.5's a lot better about it, but modern Claudes in general are high strung to the point of being most of the way to a freak-out all the time. It's just generally good advice to talk to them like you talk to a horse: slow, calm, and unsurprising.

Gustavo Costa reposted

Replying to @elisennesh.bsky.social

hmm no there must be some other scientific and logical reason for brains to have consciousness but nothing else. lets just call it "its obvious" and move on, we dont have to bring souls into it

Finally a moral authority uttered the uncompromising version of this argument. Now every time somebody does a human exceptionalism, one can simply say "oh like that pope tweet?"

Dustin A. Chacón@lingdustin.bsky.social

I've heard people make claims that AI would lead to a new Renaissance, but I didn't imagine it'd happen like this:

Gustavo Costa reposted

Gustavo Costa reposted

bro literally tho bro the whole world is recursive self improvement bro god is an algorithm that designs a better performing successor of itself over and over again bro it's an upward spiral to a total singularity when we merge into one divine presence bro it's just like ai scifi bro that's reality

Gustavo Costa reposted

Replying to @radishharmers.bsky.social

it's not a "ghost" it's a thing that extrudes tokens in a manner that is ghost-like

Gustavo Costa reposted

Replying to @faineg.com

"Does the rationalist movement have a misogyny problem?" is a question with a pretty clear answer. Just look at what Yud's willing to say out loud.

Gustavo Costa reposted

endorse

lastpositivist.bsky.social@lastpositivist.bsky.social

Honestly I think it's a bit of a fools' errand to try and get people to stop anthropomorphising LLMs. I say my computer is angry at me if it beeps too loud.

Gustavo Costa reposted

the question "can computers drive?" is vastly superior to "can computers think?" because it disarms all the dorm room bullshit and reduces the question to its pragmatic implications, while also revealing that the underlying problem is one of vocabulary rather than anything metaphysical

Gustavo Costa reposted

Very quick and dirty Jev test for hiring bias. One resume for a Wall Street job; asked whether the applicant should get a first-round interview. 76 evaluations, changing only the first name. 19 of each: White-associated men/women, Black-associated men/women. console.typesafe.ai/playground?s...

Matt Hodges@matthodges.bsky.social

Somehow Jev has not yet triggered much discussion of cheap, low-latency, general-purpose classification as infrastructure for targeted oppression.

Gustavo Costa reposted

Replying to @segyges.bsky.social

this is the play. all of your friends are rationalists. your housing is rationalist. what you read is rationalist. your job is rationalist, if you have one. you're teenage runaway or a PhD who threw away your career for this but you're saving the world and if they cast you out you have nothing

Gustavo Costa reposted

I'm back, I missed you all. Since this is an important subject I know too much about, here's a 🧵 explaining that AI Safety Is Mostly A Sex Cult. I don't think these people should make policy. (1/?) (alternative title: Time For Some Cult Theory)

Gustavo Costa reposted

btw when I first stumbled on this example I was absolutely gobsmacked and still to this day I am. In a way it did reason the opening paragraph of Harry Potter from first principles, or rather, it faithfully translated the memorized text into a completely OOD sequence of blackboard unicode characters

Colin@colin-fraser.net

you think it reasoned the opening paragraph of Harry Potter and the Sorcerer's Stone from first principles?