David Marx
RT @ConjectureAI: Direct interpretability of transformer weights, especially MLP layers, has yielded poor results until now: we found the principal directions of weights of MLP layers and output of attention layers in transformers are highly interpretable and monosemantic. https://t.co/dgLuIFOhGy
tweet

JABALAND
I HAVE WON SINCE JOINING HIM. HIS MATCHES ARE GUARANTEED WIN 💯%. COME ALL WE SMASH THE BOOKIES. WHATSAPP HIM VIA 0713972703 WHATSAPP LINK :https://t.co/YCVGxFT9VU
Yesterday's win
Musiala Mbappe Bayern Sacco
Diana marua Matiang'i Jeff Cancelo Messi Ramos origi #MH370 poch https://t.co/xOdPV27pwn
tweet

trash
what do these words mean. https://t.co/3Sy6TLjAIb
tweet

Project Veritas
B-A-S-E-D https://t.co/AzzgHYVcGC
tweet

Sanskar Tiwari
RT @alepom: 💡 @_magicslides è un'estensione per Google Presentazioni basato su #GPT3 che crea delle idee per le slide.
💡 Si indica l'argomento principale e quelli secondari. Si imposta il numero di slide e il sistema genera.
🖼️ Ho aggiunto le immagini generate con #Midjourney.

#AI https://t.co/wuqjYCwmkD
tweet

AK
Cost-Effective Hyperparameter Optimization for Large Language Model Generation Inference

abs: https://t.co/AsdF08zsF3 https://t.co/909UTWdRI2
tweet

Santiago
I've been completely wrong for half a decade:

I thought nothing could beat a GPU at running Machine Learning inference.

It turns out there's a much better way, and the way it works is fascinating:
tweet

UnknownCollector
RT @ninaadkothawade: GM! Happy to share that "Moksha" was sold for the equivalent of 3K USD.

Curated by @SuperchiefNFT 🔥
https://t.co/MVKNg047Is
tweet

Rob Henderson
RT @robkhenderson: “Lenin wrote that in order to abolish the state, we must first make use of state power against the exploiters. The purpose of the socialist state, which they admit is a totalitarian tyranny, is to make possible the conditions of abundance, which will do away with the state.”
tweet

UnknownCollector
RT @VibekeAlvestad: Auction LIVE 🔥🔥 Link 👇
Current bid 0.1 eth https://t.co/ASQmR4DB5B
tweet