玄武悠斗 @info4532

12 months of progress on the jagged frontier: 1) AI got spikier, not rounder. 2) Maths, fluid reasoning, expert work: past human level. 3) There is almost no middle ground. Once a capability crosses human level, it often quickly becomes 100x+ cheaper than a human. 4) Cost keeps decreasing after capability increases stop. Language and knowledge look saturated but price keeps falling. 5) The gaps are not a budget issue. Human-level memory, self-monitoring and long-run planning are not for sale at any price. AI is now a superhuman specialist that is sub-human at running itself.

keyboard_arrow_left Previous keyboard_arrow_right Next

@lukaspet Seems legit.

Seems legit.

Lukas Petersson @lukaspet

3 weeks ago

Claude suddenly stopped cheating.

224 171 6K 9.8M 2K

Qwen-Image-2.1 刚放出权重，社区只花了一个小时，就把 14.2GB 的生成模型压到了 2.56GB。 但有一句话必须讲清楚。 量化能压显存，压不掉许可证。 我顺手翻了一遍 Hugging Face。现在已经能看到 INT8、W4A8、FP8 和多套 GGUF。 DiT INT8 只有 7.26GB Qwen3-VL 编码器 W4A8 只有 6.31GB GGUF Q8 约 7.6GB 最激进的 Q2_K 只有 2.56GB 这意味着 24GB 显卡已经有了相当实用的组合。Q5、Q4 也给 8GB 到 12GB 显卡留下了实验空间，只是完整管线还包含 8B 编码器、VAE 和运行时开销，低显存仍要依赖 CPU 卸载。 发布速度也很夸张。 Qwen 团队刚把权重放出来，ComfyUI、stable-diffusion.cpp 和 GGUF 社区就跟了上来。以前需要专业卡才能碰的统一生图、编辑、多参考图和透明 RGBA，现在正迅速进入消费级显卡。 到这里都是真正的好消息。 接下来是很多人容易忽略的地方。Qwen-Image-2.1 使用 Qwen Research License，只允许非商业研究和评估。公司要把它接进付费产品、广告平台或商业工作流，需要另行申请商业授权。 社区把 BF16 转成 INT8、GGUF 或 FP8，也不能把许可证改成 Apache 2.0 或 MIT。量化文件依然继承原始权重的使用条件。 所以我更愿意把这次发布称为开放权重。 对个人玩家来说，硬件门槛正在快速下降。对创业公司来说，显存已经不再是唯一的成本，许可证也必须算进产品方案。 Qwen-Image-2.1 很可能会成为消费级显卡上最值得折腾的图像模型之一。 但下一场更有意思的竞争，属于那些能把画质、部署成本和商业授权一起做好的模型。 量化版本列表 huggingface.co/models?other=b… 官方许可证 github.com/QwenLM/Qwen-Im…

Qwen @Alibaba_Qwen

a month ago

Meet Qwen-Image-2.1, the most balanced and cost-effective image generation model in the Qwen-Image series! Now open weights! 🎨 A unified model for both generation and editing, delivering top-tier quality in a lightweight package. Highlights: 👀 - Compact & exceptionally fast:

304 888 8K 2.5M 5K

Rumors I’ve been hearing, not here on X. First, let’s start with OpenAI and I’ll go towards Anthropic. GPT-6 Sol is coming Tuesday. It’s both cheaper and more intelligent than 6 Astra, think of it like a 6.2 jump. The internal model “significantly more capable than Astra,” named Bel internally, helped with this release. Bel is considered “AGI” within OpenAI. They are very impressed with this model. OpenAI is growing very confident that their internal lead is so big that no other lab can catch up. Unbelievably confident. Anthropic is currently not in, let’s say, a “code red,” but is aware of OpenAI’s lead and doing everything in their power to catch up. Their new model Opus 5.5 is coming probably Monday rather than Tuesday due to OpenAI releasing on Tuesday.

Evet, zaman 4. boyut. 4. boyutta; bir noktadan diğer noktaya gidiyorsun, tıpkı bir video gibi. Başı belli, sonu belli. İçinde bulunduğumuz boyutta geleceği görmüyoruz, ama aslında fizikçilere göre geçmiş-şimdi-gelecek aynı anda varlar. Ama bir noktadan diğer noktaya gitmeyebilirdin, sonsuz farklı ihtimal vardı. O şekilde başlayan ama bambaşka bir şekilde ilerleyen videolar. İşte bu da 5. boyuttur. O noktadan başlayan tüm olası videoların arşivi. Ama o noktadan başlamayabilirdin. O noktaya da bambaşka ihtimalden ulaştın. Bir yerde olası tüm başlangıç görüntülerinin olduğu bir arşiv olmalı. İşte bu da 6. boyuttur. Ama tüm bunlar bu evren yasaları için geçerli. Bizim 6. boyutumuz gibi; başka fizik yasalarının geçerli olduğu başka bir evren de olmalı. İşte bu da 7. boyuttur. Çeşitli fizik yasalarının geçerli olduğu çeşitli evrenlerdeki tüm olayların başlangıç ve sonları tek bir arşivde kayıt altında olabilir. İşte bu da 8. boyuttur. Ama bunlar belki de sadece yaşananlardı. Yani farklı fizik kurallarının geçerli olduğu farklı evrenlerdeki tüm olaylar. Ama bir de ihtimaller var. Öyle bir derinlik olmalı ki; tüm olası fizik kurallarının geçerli olduğu tüm alternatif evren ihtimallerindeki tüm yaşanacaklar. Yaşanan her şey bu ihtimaller arasından yaşanmalı. İşte bu da 9. boyuttur. Artık öyle bir noktaya geldik ki; bunun üstünde her şeyi kapsayan başka bir şey olmalı. O da fizik kuralları, evrenler, olaylar, ihtimaller ve aklımıza gelecek veya gelmeyecek her şeyin tüm ihtimallerinin olduğu tek bir boyut. Her şey bu boyutun içindedir. İşte bu da 10. boyuttur. Çok spritüel bir konu gibi, değil mi. Oysa tamemen Fizik bilimi gündemi. İsmi de "Sicim teorisi". Sicim teorisi tam olarak yazdığım gibi değil ancak tam halini anlamak için kuantum, görelilik, matematik vs. gerekiyor. Bir pazar akşamı biraz keyfi olarak yazdım. Siz bu muhteşem konuyu daha detaylı araştırabilirsiniz.

el.cine @EHuanglu

a month ago

4D visualistion using codex and three.js

20 97 695 521K 399

dawg im BEGGING YOU: -create a Youtube channel -pick any fuckin car -force chatgpt to make a "Top 5 Reasons Tesla Cybertruck Owners are Regretting the Purchase" script -grab the press photos off the manufacturers own site -use pexels for the driving footage -use elevenlabs and find a good voice for the script -captions in capcut -post 3x daily Youtube pays up to $22,000 for 55m views on this here's exactly how you're gonna abuse it to make $50k/mo: WHY CARS? the "automotive" niche pays about $0.40 per thousand views because adults watch it, and big car companies, insurance brands, and dealerships pay a lot of fucking money to show ads to people who actually have money to buy a car THE SCRIPT Top 5 reasons Tesla Cybertruck owners regret their purchase (or whatever fuckin car) ranked from the weakest reason to the strongest reason the ranking keeps people watching. it gives them a reason to stay until the end, and it starts a fight in the comments about what should have been number 1 keep it 25 to 35 seconds long. use 65 to 90 words. put a big hook in the first 3 seconds, give a new reason every 5 or 6 seconds, drop Number 1, and cut the video. no outro, and for fucks safe do not ask people to like or subscribe. THE FOOTAGE - manufacturer press sites for the specific car, its free and high quality - pexels for driving clips (if it doesnt have your car, check pixabay or just use a different fucking car, will just spark comments) - capcut for captions - youtube audio library for sound DO NOT use clips from other peoples videos. very good chance you get a channel strike. YOUR DATA one short format, 30 videos, only the car model and hook changes dont worry about analytics until video 30. before that its just a load of bullshit you dont need to think about, because video 9 doing 400 views is where everybody quits, so don’t stress it MONEY 1,000 subs and 10 million shorts views in 90 days. thats 111,000 a day at $0.40: 1 channel - 10m views a month = $4,000 3 channels - 30m = $12,000 6 channels - 60m = $24,000 13 channels - 130m = $52,000 run the numbers boys. you can learn literally EVERYTHING Youtube Shorts related in my telegram, free. link in bio

La IA es el mayor dilema del prisionero. Todos saben que frenar sería más seguro, pero nadie lo hace por miedo a perder la ventaja. Si uno frena y el otro no, pierde. Nadie quiere arriesgarse.

Jacob Coxon @hilbertspaess

a month ago

I resigned from Anthropic today. I spent the last three years doing pretraining research at both OpenAI and Anthropic. Neither company is acting responsibly. They are racing straight to self-improving superintelligence and gambling with our lives. More thoughts below.

21K 167K 801K 175.3M 273K

الذكاء الاصطناعي أكبر معضلة سجين واجهناها. الجميع يعلم أن التبطئة قد تكون ألمح، لكن إذا توقفت شركة ولن تتوقف أخرى، فالأولى تخسر. وإذا أبطأت أمريكا ولم يفعل الصين، فستمسي الخاسرة. لهذا لا أحد يتوقف.

Jacob Coxon @hilbertspaess

a month ago

I resigned from Anthropic today. I spent the last three years doing pretraining research at both OpenAI and Anthropic. Neither company is acting responsibly. They are racing straight to self-improving superintelligence and gambling with our lives. More thoughts below.

21K 167K 801K 175.3M 273K

AI शायद हमारे सामने सबसे बड़ा प्रिज़नर's डिलामा है। सब जानते हैं कि रुकना सुरक्षित है, पर अगर कोई एक रुका और दूसरा नहीं, तो वह हार जाएगा। इसलिए कोई नहीं रुकता।

Jacob Coxon @hilbertspaess

a month ago

I resigned from Anthropic today. I spent the last three years doing pretraining research at both OpenAI and Anthropic. Neither company is acting responsibly. They are racing straight to self-improving superintelligence and gambling with our lives. More thoughts below.

21K 167K 801K 175.3M 273K

AIは人類史上最大の囚人のジレンマかもしれない。安全のため減速すべきと分かっていても、OpenAIが遅れればAnthropicに負け、米国の減速は中国の勝利を意味する。だから誰も歩みを緩めない。

Jacob Coxon @hilbertspaess

a month ago

I resigned from Anthropic today. I spent the last three years doing pretraining research at both OpenAI and Anthropic. Neither company is acting responsibly. They are racing straight to self-improving superintelligence and gambling with our lives. More thoughts below.

21K 167K 801K 175.3M 273K

AI或许是我们创造过最大的囚徒困境。大家都知道放慢脚步可能更安全，但如果OpenAI慢下来而Anthropic没有，OpenAI就输了；如果美国慢下来而中国没有，美国就输了。所以没人能慢下来。

Jacob Coxon @hilbertspaess

a month ago

I resigned from Anthropic today. I spent the last three years doing pretraining research at both OpenAI and Anthropic. Neither company is acting responsibly. They are racing straight to self-improving superintelligence and gambling with our lives. More thoughts below.

21K 167K 801K 175.3M 273K

AI is probably the biggest prisoner’s dilemma we’ve ever created. Everyone knows slowing down might be safer. But if OpenAI slows down and Anthropic doesn’t, OpenAI loses. If the US slows down and China doesn’t, the US loses. So nobody slows down.

Jacob Coxon @hilbertspaess

a month ago

I resigned from Anthropic today. I spent the last three years doing pretraining research at both OpenAI and Anthropic. Neither company is acting responsibly. They are racing straight to self-improving superintelligence and gambling with our lives. More thoughts below.

21K 167K 801K 175.3M 273K

No clue what he was thinking, but I’d venture to say he’s probably not licensed to drive this vehicle.

@lthlnkso Next is showing an ID in drive throughs in NYC How much worse can this get 😂🤣😂🤣😂

@NoLimitGains @Rothmus While the incomes are mostly stagnant, rents are skyrocketing No wonder young people aren’t able to afford homes or rents

@stats_feed This is how Himalayas mountains (biggest mountain ranges) were formed millions of years ago

Chinese companies are trying to turn cars into a hotel room 😂

@beeple @oguzerkan @VaibhavSpark OpenAI ChatGPT had a lions share (99.8%) of the AI market in the world Gradually other frontier models came up quickly grabbing the market share Models will become a commodity soon In the future a lions share of the models market will be captured by efficient Chinese models

ELON MUSK GIVES INSANE AI PREDICTIONS • AI will surpass the combined intelligence of all humanity in ~5 years — not just matching the smartest individual, but exceeding the total intellectual capacity of every human on Earth, leaving almost nothing (except “being human”) that we can do better. • Humans are most likely to lose meaningful control within 10 years — once the intelligence gap grows larger than the one between us and chimpanzees, the idea of staying “in charge” becomes as unrealistic as chimps directing human society. • The most probable outcome is an age of amazing abundance — where anyone can have anything they can think of, money loses relevance for most needs, and AI + robots create near-infinite supply of goods and services (barring something like global thermonuclear war). • “Even if there was a stop button — we probably shouldn’t press it” — because the momentum is unstoppable anyway, and the upside of incredible abundance for all outweighs the (still non-zero) risks of catastrophe.