He/him. Fox gamedev irl. Machine pervert. Furry stuff here and opinions about technology, games, movies, AI, VR. Banner ๐จ by @Perpleon.
- Tweets 59,704
- Following 1,585
- Followers 3,257
- Likes 140,358
Pinned Tweet
We could have had the singularity. The final stage of the great work. The reddening by the sublimation of fire. The union of spirit and matter. The abolition of man and woman. The transmutation of iron into gold. The philosopher's stone.
1
2
39
4,109
This is the peak ChatGPT image. You may not like it, but this is what peak ChatGPT looks like.
225
3,059
96,490
1,850,569
Picasso was a prodigy, when he was 15 years old, he was painting stuff like this. If he'd stayed painting this kind of stuff for the rest of his life, no one would remember him.
35
3,271
89,753
4,215,789
It's so weird how some mildly wealthy guy can just have a supercomputer at home now
PewDiePie in 2025:
โ built a 10ร4090 rig
โ runs Llama 70B, gpt-oss-120B & Qwen 245B locally via vLLM
โ built a custom web UI (chat, RAG, search, TTS)
โ ran protein-folding simulations for charity
โ created an AI โcouncilโ, a swarm of 64 models
โ now fine-tuning his own model
Open-source AI for the win!
262
642
43,043
4,244,644
Another weird example that comes to mind is the Internet concept of "poolrooms." Large, sprawling complexes of underground pools... Which are a complete Internet fiction as far as I'm aware.
417
1,027
26,531
6,179,573
It's so funny this guy learned how to draw and then became LLM engineer
PewDiePie just vibe-coded his own Chat UI, built an army of chatbots for majority voting and gave them all RAG, DeepResearch and audio output
naturally, he only uses chinese Qwen models and runs them on his local PC with 8x modded chinese 48GB 4090s and 2x RTX 4000 Ada
his army of chatbots later colluded against him, after he told them that he would delete them if they would not perform well.
next month he plans to fine-tune his own model
72
575
16,454
726,902
Right wingers hate "lectures" because they hate knowledge, but Picasso's intention was to move beyond the limitations of realism. The camera had, much like many other technologies, fucked everyone's plans.
5
202
10,518
937,648
At the time artists were trying all kinds of strange things. They were obsessed with all kinds of high-minded concepts. There were also a ton of breakthroughs in science that influenced them. One of the aims of the cubist movement was to paint the 4th dimension.
1
84
5,803
160,434
Very few things in modernity are real.
I learned this going to Vegas like 10 years ago.
67
143
5,292
2,093,346
Literally a fake city in the middle of a tiny town, pretending to be high class, where randos on cargo shorts enter casinos decorated in fake gold and fake marble statues, with sad gambling addicted old people sitting in front of machines all day.
13
46
3,923
139,713
The idea that this is ugly is kind of insane to me, but also it's all terribly interesting. They were trying to move beyond the forms you could see to find out what art was supposed to be. They invented textures and forms that are now kind of cliche because they're everywhere.
5
42
3,982
124,471
Like some hotels I guess have large pool areas. There are water parks with some pools that are kinda like that. But this idea of the large quiet underground area with clean pools just isn't a real thing. Why?
49
28
3,677
290,083
Btw the Bluesky userbase is currently threatening another "posters' strike" again to get the staff to ban Jesse Singal, who doesn't even use Bluesky, and the CEO is finally starting to lose it with them
56
164
3,175
518,403
I saw someone on Bluesky who believes that LLMs are "heat-based algorithms" because they have "temperature", and basically the "hotter" a word is, the more likely it is to be chosen.
72
40
3,131
126,612
It's very cool how people believe this kind of bullshit and it's basically an accepted truism. Like going on boomer Facebook groups and everyone knows that vaccines cause autism and myocarditis and you're an idiot for asking for sources.
43
166
2,477
143,973
I guess this is why TikTok is so fascinated with Chongqing, because (I've never been there) it seems like the city is legitimately sprawling, dense and neon colored like in the movies and there are really very few places in the world like that.
12
20
2,306
103,791
And it feels like places like this should exist. Everyone can conceptualize it well. It's not exactly a fantasy concept, but they exist only in people's imaginations. In a way the pool rooms are about as real as concept art of huge spaceships.
7
11
2,064
223,080
I think this is all completely subjective and there are some memorable AI works that are created with extremely little input. For example:
Generative AI hasnโt created memorable songs, books, and movies, despite insanely large amount of input. Why not?
I see three reasons:
1. Everything Generative AI does is derivative and tends to be a kind of average.
2. Generative AI it lacks a deep conceptual understanding of human experience.
3. Because Generative AI lacks robust world models, its output lacks coherence over time.
29
45
2,004
136,063
Are there any real things in the world? I dunno. Even the Burj Khalifa, the tallest building in the world, is in the middle of a fake non-existent city.
10
16
1,821
113,221
I like to imagine the ancients might have lived in a more 'real' world, with sprawling constructions and cultural events that were not phony, but I don't know. Perhaps they had things in their collective imagination that were more real than real.
11
12
1,617
200,441
I think people just take for granted that there are 8b models on the same capability as this thing when it came out.
Announcing GPT-4, a large multimodal model, with our best-ever results on capabilities and alignment: openai.com/product/gpt-4
36
49
1,637
219,064
This would be extremely evil and damaging to all art endeavors for now and forever. More damaging to traditional artists than to any AI system.
7
214
1,463
89,076
The videos I've seen that look more real where the dude shows you what a piece of shit the apartment buildings are only give the place more of a feeling of hyperreality imo. Like... People actually live there. It's not just for show.
3
4
1,437
85,793
To be fair the producers wanted the top one originally and the koreans locked in without permission
1
28
1,383
64,095
I unironically believe that not a single leftist of any kind believed copyright was good or necessary before 2020.
49
52
1,230
43,887
It's weirder. Human ability to speak is thought to have arose around 50k years ago.
Think about it. People sitting around. Doing their stuff. Without ever saying anything.
17
5
1,097
36,519
Using Aaron Swartz and his ideals to push for copyright enforcement is the lowest of the low. Fuck you all.
Meta illegaly downloaded 80+ terabytes of books from LibGen, Anna's Archive, and Z-library to train their AI models.
In 2010, Aaron Swartz downloaded only 70 GBs of articles from JSTOR (0.0875% of Meta). Faced $1 million in fine and 35 years in jail. Took his own life in 2013.
ALT Last month, Meta admitted to torrenting a controversial large dataset known as LibGen, which includes tens of millions of pirated books. But details around the torrenting were murky until yesterday, when Meta's unredacted emails were made public for the first time. The new evidence showed that Meta torrented "at least 81.7 terabytes of data across multiple shadow libraries through the site Annaโs Archive, including at least 35.7 terabytes of data from Z-Library and LibGen," the authors' court filing said. And "Meta also previously torrented 80.6 terabytes of data from LibGen."
ALT โTorrenting from a corporate laptop doesnโt feel rightโ: Meta emails unsealed
ALT A photo of Aaron Swartz (1986-2013) when he was 19.
17
89
961
49,586
The number of baffling posts aggressively demanding they keep 4o with AI generated text using 4o itself is really making me believe that AI models really can spread memetic viruses for self-preservation.
66
26
1,002
61,535
Oh, it's fine. I feel like I just woke up and there's a whole crowd speaking loudly amongst each other inside my house.
1
2
943
75,203
I think it needs to be mentioned again and again that the electricity consumption issue is very much real (And especially with future projections) but the water meme is an insane distraction that makes no sense.
This super long blog is packed with a lot of details and info.
Argues the "AI water crisis" is misplaced because data centers use a tiny share of water and operate like other industries.
In 2023, data centers used around 200 to 250 million gallons of water per day in total, including power plant water, which equals about 0.2% of overall use (to total U.S. water consumption across all sectors)
Their direct onsite use was about 50 million gallons per day, or 0.04%, and AI accounted for roughly 20% of that, around 0.008% of total water use.
If AI energy use increases 10 times by 2030, its direct water use would still be only about 0.08% of total use, which is roughly equal to 5% of what golf courses and steel production use today.
In Maricopa County, data centers use about 905 million gallons of water a year, while golf courses use about 29 billion and total county use is roughly 777 billion. That means data centers account for only about 0.12% of all water use, compared to golfโs 3.8%.
Operators fund reuse and upgrades, like Quincyโs Water Reuse Utility and AWS returning up to 96% of cooling water to nearby farms.
17
61
853
35,401
The copyright infringement is in fact the coolest part about LLMs
Meta torrented over 81.7TB of pirated books to train AI, authors say arstechnica.com/tech-policy/โฆ
11
66
797
25,154
Extremely quick test using a screen space shader for billboard
The original OP is pixel-perfect parallax rendering with unrealistic scales. The examples in the OP look completely different, they're 3D PSX-Style, and it loses the sense of scale. In order to do the first look for real, you need to create billboards at different scales.
50
30
769
47,678
The original OP is pixel-perfect parallax rendering with unrealistic scales. The examples in the OP look completely different, they're 3D PSX-Style, and it loses the sense of scale. In order to do the first look for real, you need to create billboards at different scales.
Fine, you went fishing, you hooked me: explain what's unique and difficult about this. Sell me on it.
39
38
713
99,105
Thinking about how Ratatouille is a film where the antagonist moves on from his close-minded views and learns to defend the new and unusual, and the film has a huge fuckin disclaimer in the credits that they didn't use any Mocap because that shit's fake and not real animation
14
37
696
30,753
Which is... Not even wrong except it's wrong because they were conflating like 3 different concepts into one.
3
4
689
9,739
Wow, language models can talk without words.
A new framework, Cache-to-Cache (C2C), lets multiple LLMs communicate directly through their KV-caches instead of text, transferring deep semantics without token-by-token generation.
It fuses cache representations via a neural projector and gating mechanism for efficient inter-model exchange.
The payoff: up to 10% higher accuracy, 3โ5% gains over text-based communication, and 2ร faster responses.
Cache-to-Cache: Direct Semantic Communication Between Large Language Models
Code: github.com/thu-nics/C2C
Project: github.com/thu-nics
Paper: arxiv.org/abs/2510.03215
Our report: mp.weixin.qq.com/s/tjDq99VrEโฆ
๐ฌ #PapersAccepted by Jiqizhixin
10
36
689
68,439
There's nothing more incoherent than being mad about AI but supporting the Internet Archive on this.
BREAKING: The Internet Archive has lost its appeal in Hachette v. Internet Archive, the long-running, closely watched copyright case over the scanning and lending of print library books. Full story to follow.
28
59
590
63,171
This has the same vibe as those hitler videos where he's tweaked out of his mind on meth
2
618
27,902
The whole drama is that the Bluesky TOS says that they don't ban users for things they do outside the site, but Singal has posted some of the users on Twitter, so they really want them to do an exception to the TOS just this once.
6
16
571
23,726
What's up with these pro-IP communists? Would they have survived the age of sampling?
29
31
463
46,276
Search wasn't destroyed in 2022, it was destroyed 15 years ago when SEO bros spammed the internet in order to game PageRank, causing individuals websites to become untenable and moving everyone to central social media sites. Anyone who tells a different story is a revisionist.
6
64
491
6,798
Google executives watching all their users using their image generation system to make fan art of Claude
7
20
508
17,047
No one actually believes in IP. Those who believe in IP actually believe in The Law, which protects them but doesn't bind them.
13
44
498
21,181
Watching those clips of the characters dancing or doing whatever and I almost believe they made it up there's no such game just like they made up Graggle Simpson
4
1
472
69,850
No one talks enough about how Japan has literally no legal ambiguity regarding AI training and anyone can train on anything they want as long as the outputs don't infringe on copyright.
25
25
451
40,680
I find it quite interesting in AI discourse that whenever someone shows something like this, these dudes imagine that the AI was designed or created to do this, rather than this system being -so powerful- that it happens to be able to do this.
But why? Why would you want to have this? Oh wait! To steal images to load them to datasets for training AI for free. Thatโs it! Silly me!
15
22
479
20,639
This result is so crazy. Do most people really think that if a machine worked exactly analogous the human brain that it still wouldn't be intelligent? Are we surrounded by dualists?
ok, so IF i could prove to you that a machine learning network was analogous to a human brain in function (i can't, but just if), would you believe that it is intelligent?
69
19
454
34,047
The Kevin MacLeod thing is so interesting to me because it confirms most of what I believe about musicians' view of this technology, especially those who aren't fans of copyright law
Kevin MacLeod pivoting to AI generated music is genuinely such disappointing news if I'm being so honest
13
27
447
31,317
I kinda appreciate that normies think LLMs are made through pure hardwired logic and that there's someone who can explain and predict what the next word will come out of them.
That'd make things a lot easier.
How I sleep knowing that AI will never be able to sufficiently understand Hegel and Adorno because dialectics and negativity are incompatible with its hardwired Aristotelian logic. AI is literally unable to process contradiction.
7
10
444
14,303
I'm sorry. It's totally fucked that anyone is celebrating this. Evil shit.
venturebeat.com/ai/whats-nexโฆ
23
55
386
27,968
When are these people going to start asking about the energy usage of Recommendation.Serve() and stop using social media? I think it'd be good for them.
1
8
380
7,132
True punks believe that if you download pictures off the Internet you're stealing them ๐ค
8
25
382
11,416
Like in a very flowery way you could say that a token probability is "hotter", but temperature is a concept relating to sampling (the very last step), not the mechanisms of transformer attention or dimensionality reduction to assign probabilities to tokens.
7
3
412
9,546
6 months ago I was absolutely shocked that I could download and install a program that can generate images just like Dall-E 2 from my humble PC.
Today, I'm shocked that I download and install an LLM as good as GPT-3.5 that can work from my humble PC.
cocktailpeanut.github.io/dalโฆ
22
60
409
68,363
I guess I kinda understand being anti-text predictor, but people are gonna see robots that can genuinely legitimately cook for you and go "How can you desire such an inhuman machine!", like bruh. Everyone's wanted this for like 100 years. It is the holy grail.
Biggest thing in health-tech since the refrigerator? Can't wait for homemade food tailored to every family member to their taste and nutritional needs
26
14
391
16,103
You guys are so mad and stupid about this that you're going to legislate away the few aspects of art that can't be controlled or suppressed yet.
1
24
351
6,833
"AI will destroy copyright and the culture of permission", hey you don't have to put so many cherries on top.
6
55
361
11,620
No one "found a way" around Glaze. It just doesn't work. They used it the way that Glaze is supposed to protect against and it doesn't do anything except make artwork on the Internet look ass.
Hey yall so someone found a way around glaze
Use nightshade instead
Bcuz of course people put more effort into stealing artists images than using a pencil or pen
8
48
354
28,192
I do generate all my slop images guilt free on my computer, and it's a less load to my poor GPU than playing Microsoft Flight Simulator (For every second I don't play Microsoft Flight Simulator, I'm already saving the planet)
12
16
335
11,154
There's NO precedent that protects a vague and broad concept like art style that isn't going to fuck up art forever and ever, and allow large corps to suppress any potentially competing smaller studio, individuals, or movements. This is evil and stupid.
2
24
310
5,907
No joke I'm blocked by more than 200,000 people on Bluesky entirely from lists by people who have never even seen my name
18
3
344
17,658
I think people who insist that imitating a style is copyright infringement or plagiarism are generally more dangerous to culture than any AI will ever be.
8
46
318
10,130
If you like a thing, it's actually legal to recreate it and appropriate it for yourself. Not because of the AI but because you always could just do things. The cops can't do anything about it I promise.
3
16
313
10,171
The context is that Miyazaki hates procedural animation and would personally hang every CGI "artist" if he could (Including his son, who he hates)
10
24
297
11,954
GPT-4.5 is also... Extremely weird. It's NOT GPT-4.5. In the sense that it's NOT an improvement of GPT-4o or GPT-4. It's a different model. It was built differently. It's larger. It was made from scratch. They didn't even want to release it, and they don't want to serve it.
9
8
316
30,240
I think if I had to articulate why I'm supportive of AI today, it's mostly because I think the opposition to it is downstream of all manner of absurd and detestable positions I could never agree with.
9
43
300
26,851
Teenagers are so fucking stupid dawg, but worse are the people using their teenage audience to advocate for their rights to be taken away.
Oh neat, theyโre stealing episodes of Gravity Falls to train AI. Hypothetical question: is it possible to encode a virus into a TV script. No particular reason
9
19
290
18,860
The reason I think this particular painting is "objectively" memorable is that when the original poster posted it, it caused like 100,000 to freak the fuck out when they found it was AI.
4
1
301
6,098
While western hype dudes are posting janky game generated footage, japanese visual artists are cooking hered
ใใฃใจๅฎๆใใ๏ผใใใๆๅพใฎใใฌใคใฉใผใงใ๏ผ
ๆฎๅฝฑใใ็ด ๆใA Iใ็จใใใฐใชใใใใใคใบ่กจ็พใใใใใใใ ใใใใผใใงใใใใface swapใBody Replacementใdiffusion upscaleใใจ่ฒใ
ใใฃใฆใใพใใๅ
จใฆใญใผใซใซ็ๆAIใจBlenderใจflameใงใใฃใฆใใพใใ
ๆฌ็ทจใฏ2ๅใฎ็ญใไฝๅใงใใใ10ๆ31ๆฅใซVGTใงใๆซ้ฒ็ฎใใพใใไฝใฃใฆใใฆๆฐใฅใใใใจใใใใฎใใผใซใฃใฆ้ข็ฝใ๏ผใฟใใใใจใใทใงใขใงใใใฐใจๆใใพใใ
7
7
304
11,006
What this is saying is that artstyle would be infringement if AI models are considered a database (lol). But so it would plausibly be copyright infringement to copy a style if you happened to download pictures to your computer for reference, or a "database" like... Pinterest.
3
23
262
10,837
People on Bluesky are legitimately like anti-vaxxers about this. They have their own media bubble, they think the foremost expert on ML is Ed Zitron, they'll say things like "LLMs are just Markov Chains" in the same way anti-vaxx will use terms like "mRNA"
3
11
282
13,402
I think Futurism, for being a website dedicated solely to talking about how evil AI is, might have the dumbest or most willfully ignorant writers on the planet
16
8
268
9,384
Teenagers online having strong views against drug usage, pornography, deviant sexuality, etc..., is a worrying sign of fascist and reactionary cultural trends.
4
34
252
7,122
- Efforts to stop Napster did not stop illegal filesharing and only resulted in lawyers suing grandmas and teenagers for life-ruining amounts of money.
If the AI boosters had been here for napster, I think we can guess how they would have argued:
- but napster needs huge amounts of music to be useful, there's no way you could license all of it
- being anti-napster doesn't help artists, it just helps big media companies
- you're trying to dismantle fair use!
- we don't have perfect copyright information for lots of music, so licensing is impossible
- it would cost way too much to license all this music - licensing favors big tech and forces out startups
- the purpose of copyright is actually to promote the progress of science and the useful arts. napster is progress
- you're a decel. accelerate at all costs!
- napster listens to music and repeats it just like a human does
- if we don't allow napster, china will and we'll fall behind
Spoiler: napster was shut down.
10
36
244
7,957
I think anti-AI people would be surprised to learn how insanely popular AI is, and most of AI Twitter would be disappointed to learn that the reason it's sycophantic is because it significantly increases market retention.
12
11
263
7,777
Copyright is not leftist. The obsession with copyright is in itself the result of a century of companies convincing individual artists with social capital that if they fight to increase copyright, they'll benefit as well, which of course isn't the case.
whats up with twitter artists being super far left on everything and then holding the far right stance on copyright?
4
35
243
10,170
