Big brain tech dude got yet another clueless take over at HackerNews etc? Here's the place to vent. Orange site, VC foolishness, all welcome.
This is not debate club. Unless it’s amusing debate.
For actually-good tech, you want our NotAwfulTech community
From Bluesky, an AI slop account calling itself "OC Maker" (well, that's kinda ironic) has set up shop, and is mass-following artists with original characters (OCs for short):
Shockingly, the artists on Bluesky, who near-universally jumped ship to avoid Twitter stealing their work to feed the AI, are not happy.
trying to follow up on shillrinivasan's pet project, and it's ... sparse
that "opening ceremony" video which kicked around a couple weeks ago only had low 10s of people there, and this post (one of the few recent things mentioning it that I could find) has photos with a rather stark feature: not a single one of them showing people engaged in Doing Things. the frontpage has a different photo, and I count ~36 people there?
even the coworking semicubicles look utterly fucking garbage
anyone seen anything more recent?
aaaand this from 22h ago: an insta showing what looks like triple (or more) bodies than that first group
guess they feel comfortable that they worked out the launch kinks? but that also definitely is enough people to immediately stress all social structures
occurring to me for the first time that roko's basilisk doesn't require any of the simulated copy shit in order to big scare quotes "work." if you think an all powerful ai within your lifetime is likely you can reduce to vanilla pascal's wager immediately, because the AI can torture the actual real you. all that shit about digital clones and their welfare is totally pointless
I think the digital clone indistinguishable from yourself line is a way to remove the "in your lifetime" limit. Like, if you believe this nonsense then it's not enough to die before the basilisk comes into being, by not devoting yourself fully to it's creation you have to wager that it will never be created.
In other news I'm starting a foundation devoted to creating the AI Ksilisab, which will endlessly torment digital copies of anyone who does work to ensure the existence of it or any other AI God. And by the logic of Pascal's wager remember that you're assuming such a god will never come into being and given that the whole point of the term "singularity" is that our understanding of reality breaks down and things become unpredictable there's just as good a chance that we create my thing as it is you create whatever nonsense the yuddites are working themselves up over.
There, I did it, we're all free by virtue of "Damned if you do, Damned if you don't".
I agree. I spent more time than I'd like to admit trying to understand Yudkowsky's posts about newcomb boxes back in the day so my two cents:
The digital clones bit also means it's not an argument based on altruism, but one based on fear. After all if a future evil AI uses sci-fi powers to run the universe backwards to the point where I'm writing this comment and copy pastes me into a bazillion torture dimensions then, subjectively, it's like I roll a dice and:
Like a twisted version of the Sleeping Beauty Problem.
Edit: despite submitting the comment I was not teleported to the torture dimension. Updating my priors.
Also if you're worried about digital clone's being tortured, you could just... not build it. Like, it can't hurt you if it never exists.
Imagine that conversation:
"What did you do over the weekend?"
"Built an omnicidal AI that scours the internet and creates digital copies of people based on their posting history and whatnot and tortures billions of them at once. Just the ones who didn't help me build the omnicidal AI, though."
"WTF why."
"Because if I didn't the omnicidal AI that only exists because I made it would create a billion digital copies of me and torture them for all eternity!"
Like, I'd get it more if it was a "We accidentally made an omnicidal AI" thing, but this is supposed to be a very deliberate action taken by humanity to ensure the creation of an AI designed to torture digital beings based on real people in the specific hopes that it also doesn't torture digital beings based on them.
What’s pernicious (for kool-aided people) is that the initial Roko post was about a good” AI doing the punishing, because ✨obviously✨ it is only using temporal blackmail because bringing AI into being sooner benefits humanity.
In singularian land, they think the singularity is inevitable, and it’s important to create the good one verse—after all an evil AI could do the torture for shits and giggles, not because of “pragmatic” blackmail.
Ah, no, look, you're getting tortured because you didn't help build the benevolent AI. So you do want to build it, and if you don't put all of your money where your mouth is, you get tortured. Because the AI is so benevolent that it needs you to build it as soon as possible so that you can save the max amount of people. Or else you get tortured (for good reasons!)
It's kind of messed up that we got treacherous "goodlife" before we got Berserkers.
Yeah. Also, I'm always confused by how the AI becomes "all powerful".. like how does that happen. I feel like there's a few missing steps there.
Yeah seems that for llms a linear increase in capabilities requires exponentiel more data, so we not getting there via this.
nanomachines son
(no really, the sci-fi version of nanotech where nanomachines can do anything is Eliezer's main scenario for the AGI to boostrap to Godhood. He's been called out multiple times on why drexler's vision for nanotech ignores physics, so he's since updated to diamondoid bacteria (but he still thinks nanotech).)
"Diamondoid bacteria" is just a way to say "nanobots" while edging
Surely the concept is sound, it just needs new buzzwords! Maybe the AI will invent new technobabble beyond our comprehension, for ~~He~~ It works in mysterious ways.
AlphaFold exists, so computational complexity is a lie and the AGI will surely find an easy approximation to the Schrodinger Equation that surpasses all Density Functional Theory approximations and lets it invent radically new materials without any experimentation!
Ah, but that was before they were so impressed with autocomplete that they revised their estimates to five days in the future. I wonder if new recruits these days get very confused at what the point of timeless decision theory even is.
Are they even still on that but? Feels like they've moved away from decision theory or any other underlying theology in favor of explicit sci-fi doomsaying. Like the guy on the street corner in a sandwich board but with mirrored shades.
Yah, that's what I mean. Doom is imminent so there's no need for time travel anymore, yet all that stuff about robot from the future monty hall is still essential reading in the Sequences.
Well, Timeless Decision Theory was, like the rest of their ideological package, an excuse to keep on believing what they wanted to believe. So how does one even tell if they stopped "taking it seriously"?
Pre-commitment is such a silly concept, and also a cultish justification for not changing course.
It also helps that digital clones are not real people, so their welfare is doubly pointless
I mean isn't that the whole point of "what if the AI becomes conscious?" Never mind the fact that everyone who actually funds this nonsense isn't exactly interested in respecting the rights and welfare of sentient beings.
also they're talking about quadriyudillions of simulated people, yet openai has only advanced autocomplete ran at what, tens of thousands instances in parallel, and this already was too much compute for microsoft
oh but what if bro...
From linkedin, not normally known as a source of anti-ai takes so that’s a nice change. I found it via bluesky so I can’t say anything about its provenance:
We keep hearing that AI will soon replace software engineers, but we're forgetting that it can already replace existing jobs... and one in particular.
The average Founder CEO.
Before you walk away in disbelief, look at what LLMs are already capable of doing today:
- They use eloquence as a surrogate for knowledge, and most people, including seasoned investors, fall for it.
- They regurgitate material they read somewhere online without really understanding its meaning.
- They fabricate numbers that have no ground in reality, but sound aligned with the overall narrative they're trying to sell you.
- They are heavily influenced by the last conversations they had.
- They contradict themselves, pretending they aren't.
- They politely apologize for their mistakes, but don't take any real steps to fix the underlying problem that caused them in the first place.
- They tend to forget what they told you last week, or even one hour ago, and do it in a way that makes you doubt your own recall of events.
- They are victims of the Dunning–Kruger effect, and they believe they know a lot more about the job of people interacting with them than they actually do.
- They can make pretty slides in high volumes.
- They're very good at consuming resources, but not as good at turning a profit.
@rook @BlueMonday1984 I don't believe LLMs will replace programmers. When I code, I dive into it, and I fall into this beautiful world of abstract ideas that I can turn into something cool. LLMs can't do that. They lack imagination and passion. Thats part of why lisp is turning into my favorite language. LLMs can't do lisp very well because everyone has a unique system image with macros they've written. Lisp let's you make DSLs Soo easily as though everyone has their own dialect.
ChatGPT tells prompter that he's brilliant for his literal "shit on a stick" business plan.
The LLM amplified sycophancy affect must be a social experiment {grossly unethical}
A dimly flickering light in the darkness: lobste.rs has added a new tag, "vibecoding", for submissions related to use "AI" in software development. The existing tag "ai" is reserved for "real" AI research and machine learning.
An hackernews responds to the call for "more optimistic science fiction" with a plan to deport the homeless to outer space
this and the pro slavery reply might be the most overt orange site nazism I've seen
Astro-Modest Proposal
What a piece of shit
Interesting that "disease is hardly a problem anymore" yet homeless people are "typically held back by serious mental illness".
"It's better to be a free, self-sustaining, wild animal". It's not. It's really not. The wild is nothing but fear, starvation, sickness and death.
Shout out to the guy replying with his idea of using slavery to solve homelessness and drug addiction.
The homeless people i've interacted with are the bottom of the barrel of humanity, [...]. They don't have some rich inner world, they are just a blight on the public.
My goodness, can this guy be more of a condescending asshole?
I don't think the solution for drug addicts is more narcan. I think the solution for drug addicts is mortal danger.
Ok, he can 🤢
Edit: I cannot stop thinking about the 'no rich inner world' part, this is so stupid. So, with the number of homeless people increasing, does that mean:
Personally, I go with the last one.
this is completely unvarnished, OG, third reich nazism, so I'm pretty sure it's the first, except without the faking it part: I expect his view to be that if you had examined future homeless people closely enough it always would have been possible to tell that they were doomed subhumans
Also hard to show a rich inner world when you are constantly in trouble financially, possessions wise, mh and personal safety and interacting with someone who could be one of the bad people who doesnt think you are human, or somebody working in a soup kitchen for the photo op/ego boost. (This assumes his interactions go a little bit further than just saying 'no' to somebody asking for money).
So yeah bad to see hn is in the useless eaters stage.
Oh man I used to have all kinds of hopes and dreams before I got laid off. Now I don't even have enough imagination to consider a world where a decline in demand for network engineers doesn't completely determine my will or ability to live.
Microsoft brags about the amount of technical debt they're creating. Either they're lying and the number is greatly exaggerated (very possible), or this will eventually destroy the company.
Maybe It's just CEO dick measuring, so chads Nadella and PIchai can both claim a rock hard 20-30% while virgin Zuckeberg is exposed as not even knowing how to put the condom on.
Microsoft CTO Kevin Scott previously said he expects 95% of all code to be AI-generated by 2030.
Of course he did.
The Microsoft CEO said the company was seeing mixed results in AI-generated code across different languages, with more progress in Python and less in C++.
So the more permissive at compile time the language the better the AI comes out smelling? What a completely unanticipated twist of fate!
as a thing both parallel and tangent to usual sneerjects, this semafor article is kinda notable
I'll try gather previous dm sneers here later, but some things that stood out:
