Bojan Tunguz
In the land of instant gratification, the secondary effects guy is the king.
tweet

Genmo
RT @AI_Einstein_: I got access to the NEW @genmoai Chat!!!

Check out my new Video Review & Tutorial
https://t.co/6lYGwSoPoG

#genmo #AIart #AIArtworks #AIイラスト #technology #Science #aivideo #GPT4 #GPTwitter #midjourneyv5 #AIグラビア #aiphotography #aigirls #blockchain #tech #AI #innovation https://t.co/vUk7EhFrvx
tweet

Stephen Young
RT @KeyTryer: Imagine telling your employees they can't use ChatGPT to work, crippling their productivity compared to the rest of the world, all because Twitter told you to hate it 🤭
tweet

2000s
21 years ago today, brandy released ‘full moon’
https://t.co/pye330wGK6
tweet

mrdoob
RT @kitasenjudesign: 新しくなったONLINE PARCOのCMをWebGLで作りました。FileSystemAccessAPIを使ってブラウザからPNGの連番を書き出し、映像化😤
#webgl #threejs https://t.co/OgdfDlRjo6
tweet

AK
∞ -Diff: Infinite Resolution Diffusion with Subsampled Mollified States

abs: https://t.co/TH253juK15 https://t.co/1I5ytVCRLq
tweet

Adam.GPT
RT @Francis_YAO_: For those claiming 10B scaled models matching GPT 3.5: on GSM8K, the best performing 11B scaled model is from my own work with 27% accuracy. GPT-3.5-Turbo, presumably also 10B, has acc 76. 27% vs 76%, that's the difference. We should be honest to ourselves https://t.co/GzE7WKO5n1
tweet

AK
Evaluating GPT-4 and ChatGPT on Japanese Medical Licensing Examinations

experiments show that GPT-4 outperforms ChatGPT and GPT-3 and passes all five years of the exams, highlighting LLMs’ potential in a language that is typologically distant from English

abs:… https://t.co/Tqx8RPmFVh https://t.co/b7MdPiknln
tweet

2000s
keira knightley in atonement (2007) https://t.co/2ZvyV0ap8e
tweet

elvis
A Survey of LLMs

A new 50 pages survey on large language models just dropped on arXiv.

https://t.co/Tq2hBKUtNt https://t.co/86wEDNRlIV
tweet

AK
Self-Refine: Iterative Refinement with Self-Feedback

In all tasks, outputs generated with SELF-REFINE are preferred by humans and by automated metrics over those generated directly with GPT-3.5 and GPT-4, improving on average by absolute 20% across tasks

abs:… https://t.co/2oJKDiLYjv https://t.co/5ZtoerWBN5
tweet