Yemp

Existing like everything.

[Yemp @YempYemp]

Efficiency and accuracy as imperatives ; speed as a luxury to afford as far as means allow.

[Yemp @YempYemp]

@yetooshi Them go fuck themselves!

[Jackson Oswalt @JacksonOswalt]

If you think Jensen resigned because of the affairs you’ve failed the IQ test of the week

[Yemp @YempYemp]

@justinskycak 😂

[Boşuna Tıklama @bosunatiklama]

Telegram'ın kurucusu Pavel Durov:

"Jensen acemice bir hata yaptı: Yönetim kuruluna sahipti. 2014'te casuslara hayır dediğim için yönetim kurulum beni görevden aldı.

2026'da ise Jensen, bir casusa evet dediği için yönetim kurulu tarafından görevden alındı.

Yani yönetim kurulu her 2 durumda da başınıza bela oluyor.

Şirket kurucularına tavsiyem: Yönetim kurulu oluşturmayın. Kendi yönetim kurulunuz kendiniz olun."

[Jensen Huang @JensenHuang]

For my first post, I’m sharing a letter @NVIDIA signed on why open models matter.

AI will transform every industry, power every company, and be built by every country.

Open models strengthen safety and cybersecurity, accelerate innovation and diffusion, and enable sovereignty.

The world needs both frontier closed models and frontier open models.

https://t.co/AUKzoQ5Ikb

[Yemp @YempYemp]

@SwedPaul That one came from heart 😅

[Nick Clegg @nickclegg]

Getting too close to Trump was a mistake – how could Silicon Valley have gotten (US) / got (UK) it so wrong?

Read the full article: https://t.co/ot428RpTT9

[Yemp @YempYemp]

@nvidianewsroom Isn't that stupid ?

Really stupid...

[Yemp @YempYemp]

Ramayana, that movie is a masterpiece !

#Ramayana

[Akito Hirai （YASAKANI） @yasakani_dev]

世界最速※のAIエンジン「YASAKANI」を、3日後の11/13（金）22時に無料公開します。

RTX 5090 1枚で 1,260 tok/s、他の8エンジンに全7分野で勝利。圧縮の品質でも、近い容量のGSQ-RCO・Unslothを上回りました。
もっと多くのPCに対応するため、クラファンを始めました。詳細はこちら↓
https://t.co/vZuWH9Ja4G

・速くする仕組みはRTX 5090専用ではありません。ほかのGPUにも広げていきます
・Qwen4にも、重みが公開されたらすぐに対応する予定です
・現在の対応はWindowsとRTX 5090です
・ソースコードと2つのモデル（最速・品質）をApache-2.0で公開

ご支援で対応GPUを広げます。累計300万円でRTX 3090・4090などの24GB・16GB機へ、その先はAMD・Macへ。1,000円から応援できます。

【動画と画像について】
・動画の48秒からは、実際に生成している画面の録画です。カットも早送りもなく、編集していません。別ウィンドウのタイマーも一緒に映しています（コードの書き換えが5.56秒で完了）
・画像1・2枚目：思考なし／思考ありの速度表（大きいほど速い）
・画像3枚目：圧縮品質の表（元のモデルにどれだけ近いか。KLは小さいほど近い）

※RTX 5090 1枚・Qwen3.8-27B・単一生成。掲載した8エンジン・7分野の比較条件での記録です。1,260 tok/sは長いJSONLの生成（思考なし）。品質は、自前の思考文で元モデルとの近さを近い容量どうしで比べた結果です。
#YASAKANI

[Yemp @YempYemp]

@alifarhat79 😂

[Netflix France @NetflixFR]

Maintenant Ils Le Savent.

NINHO : LE PRIX À PAYER, c’est dispo.

[Yemp @YempYemp]

@NetflixFR @ninhosdt Tu es le GOAT que tu pense être. Tu es le JEFE. Tu es l'HOMME.

[Qwen @Alibaba_Qwen]

The becoming begins.

[NINHO SDT @ninhosdt]

🫡

[Yann LeCun @ylecun]

LLMs accumulate a lot of declarative knowledge.
That's how they can answer a large number of questions.
That's also why they require so many parameters and hence so much memory.

But one can replace the accumulation of knowledge by *understanding*
That what JEPA world models do.
That's one reason they can be useful with considerably less memory and training data than LLMs.

This paper merely shows that the underlying similarity structure of the data, whether text or image (e.g. co-occurence of words in text or objects in images) uniquely determines the shape of the representation point cloud, when the encoder is trained trained with a common anti-collapse criterion like SIGReg or VCReg (not by reconstruction).

[Yemp @YempYemp]

@xingyudang Was not ready for this index...

[Yemp @YempYemp]

Without ai robots, there is no way USA go overdo China again. No way...

[Yemp @YempYemp]

Insiste 1000 fois sur un mensonge, ça n'en fera point une vérité. Celui qui sait déjà que c'est faux le saura toujours.
Seul les ignorants sont dupés.

#truth

[Yemp @YempYemp]

Rule 1 : This world is corrupted and deeply fake.
Rule 2 : Protect your heart so your soul not get corrupted.
Rule 3 : Your spirit is the key, focus on it.

#life #world