RomanS

Karma: 840

RomanS Oct 6, 2023, 4:51 AM
3 points
0
in reply to: pjeby’s comment on: This anime storyboard doesn’t exist: a graphic novel written and illustrated by GPT4
At my first attempt to upload the images I just drag-and-dropped the image files to the post, without editing file names. But GPT4 saves the images with the file names like this:
DALL·E 2023-10-05 12.12.06 - Illustration in anime style Inside a vast cave, Dr. Ada Worthington, a palaeontologist with brunette hair and a pink bow in it, has an intense gaze a.png
Perhaps LW internally is saving the original file names, causing the problems.
I reuploaded the images with the filenames like 1.png, perhaps it will help.

This anime storyboard doesn’t exist: a graphic novel written and illustrated by GPT4

RomanSOct 5, 2023, 2:01 PM

12 points

7 comments55 min readLW link

RomanS Sep 29, 2023, 1:57 PM
10 points
2
on: AI #31: It Can Do What Now?
Julian Hazell (distinct thread): “Why would you think AI will end up taking control?”
“We will give it to them”
A personal anecdote on the topic:
A few days ago GPT4 and me were debugging a tricky problem with docker. GPT4 suggested to run a certain docker command. As usual, I was going to copy the output and give it to GPT4. The output was a long nested json. I then noticed that the json contains the admin credentials. It was really easy to miss the part and just paste it for GPT4 to consume.
So, I almost gave GPT4 the admin credentials, which would potentially allow her to hack my app.
With many thousands of software developers doing similar things with GPT4, there will certainty be the cases where the developer wasn’t attentive enough.
This means, for the AI to break from her digital prison, she doesn’t need to do superhuman hacking, to exploit zero day vulnerabilities etc. All she has to do is to try the accidentally leaked credentials.
There is a very short path from “the AI wants to escape” and “the AI is all over the Internet”. She doesn’t even need to have a human-level intelligence for that.

RomanS Sep 28, 2023, 7:12 AM
3 points
0
on: Petrov Day Retrospective, 2023 (re: the most important virtue of Petrov Day & unilaterally promoting it)
I selected “Quickly orienting to novel situations” (QOTNoS) because it’s strictly superior to the alternatives:
- If you have the QOTNoS virtue, you can deal with the novel situation of AI destroying civilization
- The necessity of “accurately reporting your epistemic state” is a novel situation for most people. QOTNoS helps again.
- “Resisting social pressure” is a common situation. But if the survival of civilization depends on it (as the poll implies), this is a novel situation. Thus, QOTNoS will help with it.
In essence, QOTNoS (as in being able to make the right decisions in novel situations) is a synonym for general intelligence, and thus is the strongest power.

RomanS Sep 12, 2023, 12:52 PM
−35 points
−7
on: Sharing Information About Nonlinear
I think of myself as playing the role of a wise old mentor who has had lots of experience, telling stories to the young adventurers, trying to toughen them up, somewhat similar to how Prof Quirrell[8] toughens up the students in HPMOR
Speaking about taking inspiration from fiction...
Several novels by Robert A. Heinlein feature Jubal Harshaw, a fictional wealthy rationalist polymath who is living and working together with 3 sexy female secretaries: a blonde, a brunette, and a redhead (e.g. in “Stranger in a Strange Land”).
I wonder if, by a pure coincidence, the 3 women involved in the Nonlinear situation are a blonde, a brunette, and a redhead?
I’m not implying anything, and I see no problem with such a setup at all, as long as everything is done with consent. But if there is indeed such a coincidence, that would make me update about Nonlinear in several ways.

RomanS Aug 27, 2023, 9:54 AM
10 points
0
on: AI #26: Fine Tuning Time
Regarding custom instructions for GPT4, I find the one below highly interesting.
It converts GPT4 into a universal Fermi estimator, capable of answering pretty much any question like:
- What is the total number of overweight dogs owned by AI researchers?
- How many anime characters have 3 legs?
- How many species of animals don’t age?
- How long would it take for one unarmed human to dig the Suez canal?
My remaining doubts about the intelligence of GPT4 evaporated after asking it a dozen of novel/crazy questions like this. It’s clear that GPT4 is capable of reasoning, and sometimes it shows surprisingly creative reasoning.
The custom instruction:
if you don’t have some required numerical data, make a Fermi estimate for it (but always indicate if you did a Fermi estimate).
If you’re doing a series of math operations, split it into smallest possible steps, and carefully verify the result of each step: check if it’s of the right order of magnitude, if the result makes sense, if comparing it with real-world data indicates that the result is realistic (if no such data exist, be creative about finding analogies). Show the work to the user, and on each step describe the verification. Additionally, check the final result by trying a completely different alternative approach to solve the same problem. Be bold with inventing alternative approaches (the more different it is from the first approach, the better), try to approach the problem from a different angle, using a different starting point.

RomanS Aug 3, 2023, 2:18 PM
3 points
2
on: AI #23: Fundamental Problems with RLHF
What would accelerate the use of AI in movies even more would be not striking.
Not sure if the strikes in the US have any effect on the spread of AI in film making (aside from making more creators aware of the AI option). The US is important for the industry, but far from dominant. Even if the AI script writers are somehow completely banned in the US, they will still be used in the EU, China, India, etc.
Additionally, there is Youtube and other places where anyone can publish their own AI-written movie, and profit from it (and if it’s exceptionally good, the movie or some derivative could end up on the big screen, if one bothers to pursue that).
The AI can help with the writing process, but it is a hack and it will remain a hack until after we have bigger problems than protecting scriptwriting jobs.
A few months ago, GPT4 wrote the best science fiction novella I’ve read in years, and it was written without agents etc. Just the plain vanilla ChatGPT web interface.
I also watched the episode of the South Park fully created by AI, and I rate it as in the top 10% episodes of the show.
This indicates that the much more formulaic art of scriptwriting is already solvable at a superhuman level with GPT4, if someone spends a weekend or two on automating and polishing the process (e.g. a step-by-step iterative process like this).
So, let ’em strike. They’are already obsolete, even if they don’t know that yet.

RomanS Aug 3, 2023, 10:54 AM
1 point
0
in reply to: Alex Vermillion’s comment on: Barbieheimer: Across the Dead Reckoning
I agree. It’s strange how otherwise highly intelligent people fall into the trap of using Hollywood movies as a learning tool. Especially given the fact that fiction is often harmful for your mind, and given the fact that the Hollywood fiction in particular is harmful in several additional ways.
There is nothing useful one can learn from the listed movies, unless you’re specifically studying mass media (e.g. as a movie maker or a sociologist). For every mentioned topic, it’s better to grab a non-fiction book.

RomanS Jul 30, 2023, 8:42 AM
6 points
−1
on: Self-driving car bets
One indicator that could be useful for estimating the progress in self-driving is the progress with openpilot, the leading open-source software for that.
It has a github page, and boy it has some issues. Things like:

Russian parliamentarian: let’s ban personal computers and the Internet

RomanSJul 25, 2023, 5:30 PM

11 points

6 comments2 min readLW link

RomanS Jul 6, 2023, 7:03 AM
1 point
0
in reply to: jimrandomh’s comment on: Twitter Twitches
And some of these bots have been through many iterations of detection and counter-detection, and are routing their requests through residential-IP botnets, with fake user-agent strings trying to approximate real web browsers.
As someone who has done scraping a few times, I can confirm that it’s trivial to circumvent protections against it, even for a novice programmer. In most cases, it’s literally less than 10 minutes of googling and trial & error.
And for a major AI / web-search company, it could be a routine task, with teams of dedicated professionals working on it.

RomanS Jul 5, 2023, 8:50 AM
4 points
0
on: Twitter Twitches
I think the both explanations can be true at the same time:
- Twitter is refusing to pay a bill to Google
- Twitter is severely abused by data scrapers.
One likely scenario is where Google itself is a main culprit.
E.g. Elon learned that Google is scraping twitter data on industrial scale to train its AIs, without paying anything to Twitter. This results in massive infrastructure expenses for Twitter, to be paid to… Google. Outraged Elon stormed into the Alphabet headquarters, but was politely asked to get lost. Hilarity ensues.

RomanS Jun 17, 2023, 11:38 PM
1 point
−1
in reply to: dr_s’s comment on: I still think it’s very unlikely we’re observing alien aircraft
Not alien life, mind you, but crafts require interstellar travel to be plausible, and we have reason to doubt that. Even unmanned Von Neumann probes would have a very hard time arriving to their destination still functioning (never mind braking...), and non-inertial engines presume a violation of known physics so deep, it’s unbelievable we’ve missed all signs of it being possible until now.
While I agree with your general argument, I would like to point out that the aliens don’t have to be from another star system.
It seems that our Solar System has at least a dozen of separate places that could harbor life, from the clouds of Venus to the possible subsurface oceans of Pluto and beyond. And the list mostly considers the life that is similar to our own, requiring warm water (and not, say, solitons of the solar plasma). Extending the list with truly alien forms of life could increase the number of possible cradles to perhaps two dozens (??) in our Solar System alone.
Additionally, perhaps humans are not the first species on Earth that has created a technological civilization. So, theoretically there could be aliens originated from Earth.

RomanS Jun 7, 2023, 7:27 AM
5 points
−1
in reply to: TurnTrout’s comment on: The Sharp Right Turn: sudden deceptive alignment as a convergent goal
I propose the term Jasmine’s alignment, as a reference to the sudden (and fake) alignment of Jasmine in this famous scene of Aladdin (1992), right after Jasmine has realized that there is a possibility of escape:

RomanS May 25, 2023, 7:14 PM
2 points
0
on: AI #13: Potential Algorithmic Improvements
the other path isn’t guaranteed to work, but if the default path is probably or almost certainly going to get everyone killed, then perhaps ‘guaranteed to work’ is not the appropriate bar for the alternative, and we should be prepared to consider that, even if the costs are high?
I think it’s an extremely important point, often ignored.
Trying to prevent the AGI doom is not enough. If the doom is indeed very likely to happen, we should also start thinking how to survive in it.
My LW post on the topic, with some survival strategies that might work: How to survive in an AGI cataclysm.
But if you order up that panda and unicorn in a rocket ship with Bill Murray on request, as a non-interactive movie, without humans sculpting the experience? I mean let’s face it, it’s going to suck, and suck hard, for anyone over the age of eight.
Strongly depends on the prompt.
I would pay some real money to watch a quality movie about panda and unicorn in a rocket ship with Bill Murray, but with the writing of H. P. Lovecraft, and with the visuals of HR Giger.
The ship’s innards pulsed with eldritch life, cold metallic tendrils stretching into the vastness of the ship, their biomechanical surface glistening under the muted luminescence. Tunnels of grotesque yet fascinating detail lay like a labyrinthine digestive system within the cruiser, throbbing in eerie synchrony with the void outside. Unfathomable technologies hummed in the underbelly, churning out incomprehensible runes that flickered ominously over the walls, each a sinister eulogy to the dark cosmos.
Bill Murray, the lonely jester of this cosmic pantomime, navigated this shadowy dreadnought with an uncanny ease, his eyes reflecting the horrid beauty around him. He strode down the nightmarish corridors, a silhouette against the cruel artistry of the ship, a figure oddly at home in this pandemonium of steel and shadow...

RomanS May 24, 2023, 5:54 PM
1 point
0
in reply to: Waldvogel’s comment on: GPT4 is capable of writing decent long-form science fiction (with the right prompts)
I think you probably used my prompt for the one I got right, which is probably why I got it right (the tone and structure are very familiar to me after so much experimentation).
Nope, this one. But their prompt does incorporate some ideas from your prompt.

RomanS May 24, 2023, 1:25 PM
5 points
0
in reply to: Richard_Kennaway’s comment on: GPT4 is capable of writing decent long-form science fiction (with the right prompts)
The key^[1].
You got 4 of 8 right. In two cases you failed to recognize humans, and in another two—GPT4.
It was a weakly adversarial test:
- I took a few less-known but obviously talented writers from the top of my head, and copied the excerpts from the first pages.
- For GPT4, I’ve used several prompts from the competition, and then selected the parts for their stylistic diversity.
I suspect that a test with longer excerpts would be much easier for you, as the vanilla GPT4 is indeed often easy to detect due to its repetitiveness etc (I haven’t tried the APIs yet).
If GPT4 already can fool some of us science fiction junkies, I can’t wait to read the fiction by GPT5.
David Langford’s science fiction newsletter Ansible has a regular item called Thog’s Masterclass, exhibiting examples of “differently good” actually published writing. Dare the Thog-o-Matic to see some random examples. ETA: or look at any Perry Rhodan novel.
Thank you!
BTW, have you read “Appleseed” by John Clute? I have a feeling you may be one of the few people on Earth who can enjoy it. A representative sample:
They passed the iron-grey portcullis that sealed off the inferno of drive country. A dozen ceremonial masks, mourning the hardened goblin eidolons of KathKirtt that died hourly inside drive country, hung within their tile embrasure above the frowning portal. The masks were simplified versions of the flyte gorgon. Their single eyes shut in unison at the death of one of the goblin eidolons, who spent their brief spans liaising with the quasi-sentient engine brother that drove the ship through the demonic rapturous ftl maze of wormholes. Even for eidolons with hardened carapaces, to liaise was to burn and die. When Tile Dance plunged through the ashen caltraps of ftl at full thrust, the engine brother howling out something like anguish or joy all the while, its entirely imaginary ‘feet’ pounding the turns of the maze, goblins lived no longer than mayflies.
1. ^
  ~~Only one, two, and seven are human: from “Ra” by qntm, “Contact” by Sagan, “Noon: 22nd Century” by Strugatsky. The rest is GPT4~~

RomanS May 24, 2023, 12:03 PM
2 points
0
in reply to: Matt Goldenberg’s comment on: GPT4 is capable of writing decent long-form science fiction (with the right prompts)
The original one.
Judging by my limited experimentation with submitted prompts, several of them are are already superior to mine. But mine has the advantage of writing in a more academic tone, which I think is more suitable for this story.
Among the submitted ones, my current favorite is this one. The resulting prose is more human-like, but the tone is of a young-adult work, which is a disadvantage in many cases. An example:
Chapter One: An Outlier Among Outliers
Somebody had to keep an eye on the squiggly lines, and as it turned out, that person was me.
Hello there. I’m Dr. Kiera Laine. The more I understand about how the universe works, the less I understand about how the world works. I’m an astrophysicist, no, the astrophysicist, if you ask any of my colleagues in the snobbish circles of Oxford.
I’ve made a career of studying the Cosmos, charting the glittering highways of distant galaxies, the uncharted nebulae and black holes. Yet, in the world of academia, my job is equivalent to the janitor who strolls in when everybody else has gone home, only my broom is the supercomputer running complex algorithms.
Take the seismic data we were currently analyzing, for instance. To the ordinary human eye, it was just an endless sprawl of jagged lines across the monitor screen. To mine, it was poetry in motion.
On one seemingly ordinary Tuesday, sitting amidst towers of towering servers, and screens flickering with quantum code, I noticed a blip. An anomaly. A hiccup in the heartbeats of Earth. So subtle that anyone else might have missed it.
But hey, it’s the subtle ones that turn your world upside down, right?
This little blip was buried in layers of geological data gathered from sensors scattered around the globe. To be specific, it originated from the Arctic tundra, one of the coldest, remotest regions on Earth. Which was in itself, weird. These things didn’t usually come from the frozen wasteland.
To confirm my suspicions, I ran the data through another round of computations. It held up.
Well, that’s… different.
A smirk played on my lips as I spun in my chair, letting the dim light of the screens blur into streaks. I liked different. Different meant interesting. Interesting meant I wasn’t stuck in the eternal loop of the same old patterns.
If nothing else, this will give me something to wave in the faces of those who call me a ‘backroom boy’, huh?
The calculations I ran were robust, I knew that much. And the implications? They were hard to digest, even for someone like me, whose job description involved digesting the undigestible.
A part of me wanted to dismiss it as an error, an oversight, a faulty sensor, maybe. Yet, my intuition and the frisson of excitement curling through my veins told me otherwise.
So, what do you do when you stumble across a seismic anomaly hinting at the possibility of an ancient, technologically advanced civilization that existed around the time of the late Cretaceous period?
I chuckled to myself. It was indeed a question for the ages. The thought alone was ludicrous, absurd, impossible. And yet…
Oh, isn’t that the definition of science? The art of making the impossible possible.
I pulled up the geological timelines again, brushing a hand through my messy bun. The anomaly was at the precise layer of Earth’s crust that correlated with the era of the Troodon dinosaurs. An era where, according to the established norms of science, civilization as we define it today, was a mere gleam in the cosmic eye.
There is no such thing as a coincidence, Kiera, I reminded myself.
What I had in front of me was more than an anomaly. It was a door, creaking open to the unknown. It was a question mark against everything we thought we knew about the history of intelligence on Earth.
What I had in front of me was potentially the biggest discovery of the millennium.
Is it too early to start practicing my Nobel acceptance speech?
I stared at the blip again, my heart pounding like a drummer gone rogue. My mind wandered to the realms of the impossible, painting images of scaled, feathered creatures crafting tools, constructing dwellings, staring up at the stars with the same curiosity that I did.
Here be dragons, indeed.
And as the fluorescent lights of my tiny office flickered, casting an otherworldly glow on the static lines of data, I made my decision. This was too big to be swept under the rug of daily routines.
The world was about to be hit by a comet of knowledge, a comet of truth, a comet that might redefine the story of life on Earth. It was time to assemble a team, to embark on a year-long journey of discovery.
I drew in a breath, filled with anticipation and a tinge of apprehension. This was going to be interesting.
Hold on to your hats, ladies and gentlemen.
And with a few quick keystrokes, I sent an email to the head of my department.
Subject: A Matter of Seismic Importance...

RomanS May 24, 2023, 10:46 AM
8 points
0
in reply to: Richard_Kennaway’s comment on: GPT4 is capable of writing decent long-form science fiction (with the right prompts)
This has less of the fingernails-along-a-blackboard feeling given off by every sentence of the original story
Below I’ve collected excerpts from the works of several less-known but talented human writers. Or maybe they were written by GPT4. Can you guess which ones are human-made?
Sample 1:
“Nottingham has enough pubs and clubs”, say the local police. If you wanted to get around every last one of them it would be a year at a brisk trot before you were starting to visit establishments more than one mile from the centre of the city. Pick a Friday or a Saturday, any Friday or Saturday of the year: the establishments will be rammed and jumping and the streets bustling with people in their most tightly-wound and elaborately crafted drinking costumes. It’s almost Christmas but the cold season has not added much to the average number of layers.
Sample 2:
The book was better than the movie. For one thing, there was a lot more in it. And some of the pictures were awfully different from the movie. But in both, Pinocchio—a life-sized wooden boy who magically is roused to life—wore a kind of halter, and there seemed to be dowels in his joints. When Geppetto is just finishing the construction of Pinocchio, he turns his back on the puppet and is promptly sent flying by a well-placed kick. At that instant the carpenter’s friend arrives and asks him what he is doing sprawled on the floor.
Sample 3:
He glanced outside at the buildings casting long shadows in the fading sunlight, the city frozen in the grasp of time. He took a deep breath. And then, he jumped.
He closed his eyes, feeling the rush of displaced seconds, the vertigo of time stretching, condensing, then snapping back into place. When he opened them, he found the world stilled. The shadows were now statues, the sun paused in its descent, and a bird hung motionless in the sky. This was Finn’s minute—his extra minute.
Sample 4:
Once upon a Martian sunrise...
Yeah, I know, sounds like the beginning of a fairy tale or a bedtime story, right? But I promise you, on my physicist-turned-astronaut honor, this isn’t fiction. It’s the raw, unadulterated truth. My truth.
Red soil underfoot, as fine as confectioner’s sugar. Low-grav shuffle making every step a dance move. Peaks and valleys sprawled across the horizon like a mythological beast sleeping off a hard night.
I’m Corporal Thea Kolinski, an astrophysicist by trade, astronaut by accident, and currently the number one recipient of the “Most Unlikely to Succeed” superlative in our six-person crew. A crew assembled to survive on Mars. A first in human history.
Sample 5:
Then, as abruptly as it began, the light faded. Billy blinked, expecting to see a spaceman, green and with eyes as big as dinner plates, but instead, there was a rock. A simple, gleaming, alien rock, sat innocently in the middle of his vegetable patch.
The spaceship blinked out of existence just as quickly as it had appeared, leaving Billy alone in his garden, a lump of extraterrestrial mineral his only proof of what he’d witnessed.
Sample 6:
For in the silence of their failure, a sound, inaudible to human ears but felt in the marrow of their bones, resonated from the vessel, wrapping the men in a shroud of madness. One by one, they fell, their minds invaded by images of cosmic horror, their sanity shredded by the unintelligible secrets whispered by the alien ship.
Standish was the last to succumb, his face a rictus of terror as he stared at the vessel. As the sun dipped below the horizon, plunging the valley into an unfathomable darkness, Standish’s final cry echoed through the hills of Belseth, marking the tragic end of their misguided endeavour.
Sample 7:
“Jump!” he shouted to Mandel. The crawler shuddered, throwing up clouds of sand and dust, and started to turn stern up. Novago switched off the engine and scrambled out of the crawler. He landed on all fours, and, without standing up, scurried off to one side. The sand slid and sank underneath him, but Novago managed to reach firm ground. He sat down, tucking his legs under him.
He saw Mandel, who was kneeling at the opposite edge of the crater, and the stern of the crawler, shrouded in steam and sticking up out of the sand on the bottom of the newly formed crater. Theoretically it was impossible for something like this to happen to a Lizard model. Here on Mars, at least. A Lizard was a light, fast machine—a five-seat open platform mounted on four autonomous caterpillar-tracked chassis. But here it was, slowly slipping
Sample 8:
No, I ain’t pullin’ your leg, sonny, it’s the God’s honest truth. It happened in our little town of Lonesome Hollow, right there in the foothills of the Appalachians. It’s a tale that puts the ‘odd’ in odds and ends, I tell you.
The lass in question was Amelia, known to most as plain ol’ Millie. Millie was as normal as the day is long, a pretty thing with a head full of chestnut curls and the sweetest smile you’d ever see. She ran the general store, knew the names of every critter in town, and baked the best apple pie this side of the Mississippi.

RomanS May 24, 2023, 8:41 AM
−1 points
−4
in reply to: SimonM’s comment on: Do humans still provide value in correspondence chess?
the most important things are compute and dilligence
I agree. Judging by the fact that AI is strongly superhuman in chess, the only winning strategy is to completely remove the human from the loop, and instead invest in as much compute for the AI as one can afford.
a sequence that no computer would consider or find
If it’s a sequence that no superhuman AI would consider, this means that the sequence is inferior to the much better sequences that the AI would consider.
It seems that even after 2 decades of the complete AI superiority, some top chess players are still imagining that they are in some ways better at chess than the AI, even if they can’t win against it.

RomanS

This anime sto­ry­board doesn’t ex­ist: a graphic novel writ­ten and illus­trated by GPT4

Rus­sian par­li­a­men­tar­ian: let’s ban per­sonal com­put­ers and the Internet

This anime storyboard doesn’t exist: a graphic novel written and illustrated by GPT4

Russian parliamentarian: let’s ban personal computers and the Internet