๐Ÿ“ข ๊ฒ€์ƒ‰ ๊ธฐ๋Šฅ ์ถ”๊ฐ€ ์˜ˆ์ •

์ฃผ๋ฏผ๊ฑด

๊ธฐ๋ก = ์„ฑ์žฅ

zoomg

23.04.06 (Thu)

Segment Anything
Meta AI Computer Vision Research
๋ฐ์ดํ„ฐ ๋•Œ๋ ค๋ฐ•์•„์„œ ๋ญ๋“ ์ง€ segmentation ํ•˜๋Š” model / ํ˜„์žฌ LLM์ด ์„ฑ๊ณตํ•œ ๋ฐฉ์‹ (๋ฐ์ดํ„ฐ ๋•Œ๋ ค๋ฐ•์•„์„œ ๋ชจ๋“  NLP task ์”น์–ด๋จน๊ธฐ)์„ ์ด๋ฏธ์ง€ ๋ถ„์•ผ์— ์ ์šฉ / ๋‚ด ์กธ๋…ผ ๋ฐฉํ–ฅ๊ณผ ๋น„์Šทํ•˜๊ณ  ์กธ๋…ผํ•˜๋ฉด์„œ ์Šค์น˜๋“ฏ ๋А๊ปด์ง„ ๋ถ€๋ถ„์ด์—ˆ๋Š”๋ฐ ์‹ค์ œ๋กœ ๋‚˜์™€๋ฒ„๋ฆผ,, / ๊ทธ๋ž˜๋„ ํ•ต์‹ฌ์ ์ธ ์กธ๋…ผ๊ณผ์˜ ์ฐจ์ด์ ์€ ์Ÿค๋„ค๋Š” label์ด ์—„์ฒญ๋‚˜๊ฒŒ ํ•„์š”ํ•œ๋ฐ ์กธ๋…ผ ๋ฐฉ๋ฒ•์€ label ํ•„์š” ์—†์ด ๋‹จ์ˆœ ์ด๋ฏธ์ง€๋งŒ ์‚ฌ์šฉ
[๋…ผ๋ฌธ ์ •๋ฆฌ] Segment Anything
* ์ž‘์„ฑ์ž: ๋ฐ•์ง„์šฐ (GitHub) * GitHub: https://github.com/facebookresearch/segment-anything * Demo: https://segment-anything.com/demo * Blog: https://ai.facebook.com/blog/segment-anything-foundation-model-image-segmentation/ Motivation ์ตœ๊ทผ ๋Œ€๊ทœ๋ชจ์˜ ์–ธ์–ด ๋ชจ๋ธ๋“ค์€ ๋†€๋ผ์šด Zero-shot / Few-shot Generalization ์„ฑ๋Šฅ์„ ๋ณด์ด๊ณ  ์žˆ๋‹คโ€ฆ

Towards Healthy AI: Large Language Models Need Therapists Too ๐Ÿ‘€ TL;DR The paper presents the SafeguardGPT framework...

Posted by Daily Arxiv on Thursday, April 6, 2023

23.04.05 (Wed)

[Kor/Eng by ChatGPT] What can RL do?
editor, Seungeon Baek(๋ฐฑ์Šน์–ธ) Reinforcement learning Research Engineer [Kor] ์•ˆ๋…•ํ•˜์„ธ์š”, ์˜ค๋žœ๋งŒ์— ๋ธ”๋กœ๊ทธ๋ฅผ ์“ฐ๊ฒŒ ๋˜๋„ค์š”! ์ด๋ฒˆ ๊ธ€์€ ๋…ผ๋ฌธ ๋ฆฌ๋ทฐ๊ฐ€ ์•„๋‹Œ, ๊ฐ•ํ™”ํ•™์Šต ๊ด€๋ จํ•œ ์ €์˜ ์ฒซ ํฌ์ŠคํŒ… ๊ธ€์ž…๋‹ˆ๋‹ค. ์ด๋ฒˆ์— ์ž‘์„ฑํ•˜๋Š” ๊ธ€์—์„œ ๋‹ค๋ฃจ๊ณ ์ž ํ•˜๋Š” ๋‚ด์šฉ์€ โ€๊ฐ•ํ™”ํ•™์Šต์€ ์–ด๋–ค ๋ฌธ์ œ๋ฅผ ํ’€ ์ˆ˜ ์žˆ๋Š”๊ฐ€?โ€์ž…๋‹ˆ๋‹ค. ์ด์™€ ๊ด€๋ จํ•˜์—ฌ, ๋น„์Šทํ•œ ๊ธ€์ด ์žˆ๋Š”์ง€ ์—ฌ๋Ÿฌ ํ‚ค์›Œ๋“œ๋กœ ๊ฒ€์ƒ‰์„ ํ•ด ๋ณด์•˜๋Š”๋ฐ ๊ฐ•ํ™”ํ•™์Šต์ด ์–ด๋–ค ๋ฌธ์ œ๋ฅผ ํ’€ ์ˆ˜ ์žˆ๋Š”์ง€์— ๋Œ€ํ•œ ๋‚ด์šฉ๋ณด๋‹ค๋Š”, ๊ฐ•ํ™”ํ•™์Šต์ด ํ˜„์žฌ ์ ์šฉ๋˜๊ณ  ์žˆ๋Š” ๋ถ„์•ผ์™€ ๊ด€๋ จ๋œ ๋ธ”๋กœ๊ทธ๋“ค์ด ๋Œ€๋ถ€๋ถ„์ž„์„โ€ฆ
ReAct: Synergizing Reasoning and Acting in Language Models
While large language models (LLMs) have demonstrated impressive capabilitiesacross tasks in language understanding and interactive decision making, theirabilities for reasoning (e.g. chain-of-thought prompting) and acting (e.g.action plan generation) have primarily been studied as separate topicsโ€ฆ

Story does not exist

โ€ณ์šฐ๋ฆฌ๋Š” ์ง€๊ธˆ AI์— ๋Œ€ํ•œ ์—„์ฒญ๋‚œ ํฅ๋ถ„๊ณผ ๊ณผ๋Œ€๊ด‘๊ณ ๊ฐ€ ๋‚œ๋ฌดํ•˜๋Š” ์‹œ๋Œ€์— ์‚ด๊ณ ์žˆ๋‹คโ€... ์Šคํƒ ํผ๋“œๅคง, ์ธ๊ฐ„์ค‘์‹ฌ์ธ๊ณต์ง€๋Šฅ์—ฐ๊ตฌ์†Œ โ€˜AI ์ธ๋ฑ์Šค 2023โ€™ ๋ฐœํ‘œ
็พŽ ์Šคํƒ ํผ๋“œ๋Œ€ํ•™๊ต(Stanford University Stanford)์˜ ์Šคํƒ ํผ๋“œ ์ธ๊ฐ„ ์ค‘์‹ฌ ์ธ๊ณต์ง€๋Šฅ์—ฐ๊ตฌ์†Œ(Human-Centered Artificial Intelligence. HAI)๊ฐ€ ๋งค๋…„ ๊ณต๊ฐœํ•˜๋Š” ์—ฐ๋ก€๋ณด๊ณ ์„œ โ€˜์ธ๊ณต์ง€๋Šฅ ์ธ๋ฑ์Šค 2023(AI Index 2023)โ€™๊ฐ€ 3์ผ(ํ˜„์ง€์‹œ๊ฐ„) ๋ฐœํ‘œ๋˜์—ˆ๋‹ค.AI ์ธ๋ฑ์Šค๋Š” ํ•™๊ณ„์™€ ์—…๊ณ„์˜ ํ•™์ œ๊ฐ„ ์ „๋ฌธ๊ฐ€ ๊ทธ๋ฃน์ธ AI ์ธ๋ฑ์Šค ์šด์˜ ์œ„์›ํšŒ(AI Index Steering Committee)๊ฐ€ ์ฃผ๋„ํ•˜๋Š” ์Šคํƒ ํฌ๋“œ ์ธ๊ฐ„์ค‘์‹ฌ ์ธ๊ณต์ง€๋Šฅ์—ฐ๊ตฌ์†Œ์˜ ๋…๋ฆฝ์ ์ธ ์ด๋‹ˆ์…”ํ‹ฐ๋ธŒ๋‹ค.์ธ๊ณต์ง€๋Šฅ์€ ์ƒˆ๋กญ๊ณ  ์ธ์ƒ์ ์ธ ๊ธฐ์ˆ  ์—ญ๋Ÿ‰์— ๋„๋‹ฌํ–ˆ

BloombergGPT: A Large Language Model for Finance ๐Ÿ‘€ TL;DR This paper presents BloombergGPT, a 50 billion parameter...

Posted by Daily Arxiv on Monday, April 3, 2023

๋”ฅ๋Ÿฌ๋‹์—์„œ๋Š” ๊ธฐ์กด์˜ ์ด๋ก (์ „ํ†ต/์ƒ์‹)์œผ๋กœ๋Š” ๋„์ €ํžˆ ๋ฐ›์•„๋“ค์ด๊ธฐ ์–ด๋ ค์šด ํ˜„์ƒ๋“ค์ด ๋งŽ์ด ๊ด€์ฐฐ๋œ๋‹ค. ๊ทธ ํ˜„์ƒ์— ์ฃผ๋ชฉํ•˜๊ณ  ๊ธ์ •์ ์œผ๋กœ ๋ฐ›์•„๋“ค์ธ ์‚ฌ๋žŒ๋“ค์ด ์„ฑ๊ณผ๋ฅผ ๋‚ด์–ด์™”๋‹ค. ์–€๋ฅด์ฟค๋„ ๊ทธ๋žฌ๊ณ  OpenAI๋„ ๊ทธ๋žฌ๋‹ค. ๋”ฅ๋Ÿฌ๋‹์˜ ๊ฐ€์žฅ...

Posted by Sedong Nam on Tuesday, April 4, 2023
๋ˆ๊ธฐ

23.04.02 (Tue)

๊ทธ๋ฆผ์œผ๋กœ ์ดํ•ดํ•˜๋Š” ์Šคํ…Œ์ด๋ธ” ๋””ํ“จ์ „
์ œ์ด ์•Œ๋ผ๋งˆ๋ฅดJay Alammar์˜ โ€˜The Illustrated Stable Diffusionโ€™๋ฅผ ๋ฒˆ์—ญํ–ˆ์Šต๋‹ˆ๋‹ค.
ํŒ€์€ ๋ถ€์กฑํ•œ ์‚ฌ๋žŒ๋“ค๋ผ๋ฆฌ ๋ชจ์ธ ์ง‘๋‹จ์ด๋‹ค, ํŒ€์žฅ์€ ๋ˆ„๊ตฌ๋ณด๋‹ค ๋ถ€์กฑํ•œ ์ƒํ™ฉ์„ ๋น ๋ฅด๊ฒŒ ์•Œ์•„์ฑ„๊ณ  ๋น ๋ฅด๊ฒŒ ํ–‰๋™ํ•˜์—ฌ ๋งค๊ฟ”์•ผํ•œ๋‹ค, ์‚ฌ์‹ค ์ด๋Š” ํŒ€์›๋„ ๋งˆ์ฐฌ๊ฐ€์ง€๋กœ ํ–‰๋™ํ•ด์ค˜์•ผํ•œ๋‹ค

23.03.31 (Fri)

ChatGPT ์•„๋ฌด๋‚˜ ๋งŒ๋“ค์ž
TestingLLM/ChatGPT_NER.ipynb at main ยท ritun16/TestingLLM
Contribute to ritun16/TestingLLM development by creating an account on GitHub.
NER ๋ฐ์ดํ„ฐ๋ฅผ ChatGPT๋กœ ๋งŒ๋“ค๊ธฐ
Trends in AIโ€Šโ€”โ€ŠMarch 2023
LLaMA from Meta, an embodied PALM-E model from Google, and Consistency Models, and new OpenAI API endpoints plus a juicy pricing forโ€ฆ
Introduction to Reinforcement Learning with Human Feedback
Learn about reinforcement learning with human feedback (RLHF) โ€” a new technique for training large language models that has been behind many of the major advances in OpenAIโ€™s ChatGPT and InstructGPT LLMs, DeepMindโ€™s Sparrow, Anthropicโ€™s Claude, and more.
GitHub - microsoft/JARVIS: JARVIS, a system to connect LLMs with ML community
JARVIS, a system to connect LLMs with ML community - GitHub - microsoft/JARVIS: JARVIS, a system to connect LLMs with ML community
Huggingface + ChatGPT, ์ž๊ธฐ ๋ฐ์ดํ„ฐ ์—†๋Š” AI ํšŒ์‚ฌ๋“ค์€ ๋‹ค ์ฃฝ์„๋“ฏ
GitHub - Kent0n-Li/ChatDoctor
Contribute to Kent0n-Li/ChatDoctor development by creating an account on GitHub.
์˜์‚ฌ ChatGPT
ColossalChat: An Open-Source Solution for Cloning ChatGPT With a Complete RLHF Pipeline
Large AI models and applications like ChatGPT and GPT-4 have become extremely popular worldwide, serving as a foundation for theโ€ฆ

Great! Youโ€™ve successfully signed up.

Welcome back! You've successfully signed in.

You've successfully subscribed to zoomg.

Success! Check your email for magic link to sign-in.

Success! Your billing info has been updated.

Your billing was not updated.