[B급 프로그래머] 12월 4주 소식(빅데이터/인공지능, 하드웨어, 읽을거리 부문)

(오늘의 짤방: 그러네ㅋㅋㅋ 사람도 이렇게 일시키면 잘할 것 같은데ㅋㅋㅋ via @nameEO)

1. 빅데이터/인공지능
   • Stable Diffusion in Java (SD4J) Enables Generating Images With Deep Learning
   • LLM Compiler Agent Cookbook
   • Temporian is a library for safe, simple and efficient preprocessing and feature engineering of temporal data in Python.
   • Needle in a 930M Member Haystack: People Search AI @LinkedIn
   • Hands-on LLMs Course - Learn to Train and Deploy a Real-Time Financial Advisor
   • Tanuki - LLM 기반의 앱을 쉽게 개발하기 (github.com/Tanuki)
   • AI가 특허권, 혹은 저작권을 가질 수 있을까?
   • Emu2 - Gemini와 비슷한 오픈소스 37B 멀티모달 모델 (github.com/baaivision)
   • 5 Levels Of Text Splitting/Chunking:
   • 26 Prompting Tips
   • The "Hello World"s of machine learning & AI:
   • "2024년은 생성 AI 과대광고 끝나고 비즈니스로 승부"
   • Gemini vs GPT-4V: A Preliminary Comparison and Combination of Vision-Language Models Through Qualitative Cases
   • LLM Interactive Optimization of Open Source Python Libraries -- Case Studies and Generalization
   • A Recipe for Scaling up Text-to-Video Generation with Text-free Videos
   • Introducing Ultralytics YOLOv8, the latest version of the acclaimed real-time object detection and image segmentation model.
   • Detecting Technical Debt Using Natural Language Processing Approaches -- A Systematic Literature Review
   • Scaling Down to Scale Up: A Cost-Benefit Analysis of Replacing OpenAI's GPT-4 with Self-Hosted Open Source SLMs in Production
   • Language Model Based on FastText
   • Principled Instructions Are All You Need for Questioning LLaMA-1/2, GPT-3.5/4
   • Vertical AI - 지속 가능한 AI 앱을 위해 수직적 접근방식이 중요한 이유 (greylock.com)
   • GPT Pilot is a true AI developer that writes code, debugs it, talks to you when it needs help, etc.
   • Midjourney v6 릴리즈 (mid-journey.ai)
   • Per-User Retrieval
   • OpenAI Publishes GPT Prompt Engineering Guide
   • Exploiting Novel GPT-4 APIs
   • Advanced RAG Techniques: an Illustrated Overview
   • VCoder: Versatile Vision Encoders for Multimodal Large Language Models
   • InternVL: Scaling up Vision Foundation Models and Aligning for Generic Visual-Linguistic Tasks —— An Open-Source Alternative to ViT-22B
   • Retrieval-Augmented Generation for Large Language Models: A Survey
   • "제미나이 프로 이용정보, 구글이 다 들여다본다"
   • "AI가 굴리는데 영 신통찮네"…적자 쌓이는 로보어드바이저社 [긱스]
   • Amazon Q Code Transformation: Automating Java Application Upgrades
   • Using Gemini AI in Android Apps with the New Google AI SDK
   • We'll need to settle on a good vocabulary to describe complex Large Language Model driven systems.
   • Building Your Own Product Copilot: Challenges, Opportunities, and Needs
   • 자동개인식별부터 감정분석까지…AI에 사활 건 IPTV 업계
   • 비디오몬스터, AI 기반 여행 브이로그 자동편집 앱 '비브(ViiV)' 론칭
   • Awesome LLM Interpretability - A curated list of amazingly awesome tools, papers, articles, and communities focused on Large Language Model (LLM) Interpretability.
   • 토스의 AI 그래픽 생성기, 토스트를 소개합니다 #2
   • Can LLMs Replace Data Analysts? Getting Answers Using SQL Part 2: Diving deeper into LLM agents
   • Google Gemini API (Python notebook)
   • Ferret: Refer and Ground Anything Anywhere at Any Granularity
   • How Not to Be Stupid About AI, With Yann LeCun
   • Stable Diffusion XL by Kaggle
   • Gemini-Pro model guide
   • LMPerf Leaderboard 🏆 - Utilizing the LLMPerf, we have benchmarked a selection of LLM inference providers.
   • GTC 2022 - How CUDA Programming Works - Stephen Jones, CUDA Architect, NVIDIA
   • MI300X has 32% more tflops than H100 for BF16 (1307 vs 989)!
   • LLMs Will Make Programming Useless In 10 Years
   • An In-depth Look at Gemini's Language Abilities
   • 영국 대법원, 다버스 사례 판결 내려 "AI는 특허 소유 못해"
   • 생성형 AI 파트너를 선택하는 방법 “신뢰하되 검증하라”
   • 세일즈포스, '직장 내 생성형 AI 활용 전망과 위험' 보고서 발표
   • Generate text with LLMs - Robust prompting & (guided) text generation
   • 구글 제미나이, 첫 외부 테스트서 'GPT-3.5 터보'보다 성능 떨어져
   • HD-Painter: High-Resolution and Prompt-Faithful Text-Guided Image Inpainting with Diffusion Models
   • Building RAG-based LLM Applications for Production
   • AppAgent: Multimodal Agents as Smartphone Users
   • Texify is an OCR model that converts images or pdfs containing math into markdown and LaTeX that can be rendered by MathJax ($$ and $ are delimiters).
   • Getting Started with Mixtral 8X7B
   • All of the LLM examples in MLX-examples now support quantized models out of the box: - Mistral, Mixtral, Llama, Phi2, Qwen + Code, Chat, and Instruct variants
   • Two new llama-datasets and a Gemini vs. GPT showdown
   • Running Mixtral 8x7 locally with LlamaIndex
   • Microsoft LLMLingua - 추론 가속 및 비용 절감을 위해 프롬프트 압축하기 (github.com/microsoft)
   • A Novel Approach for Rapid Development Based on ChatGPT and Prompt Engineering
   • Brand new: MLX models on the hub(huggingface)
   • Semantic Chunking?
   • Getting started with visual assistants: templates + video overview 👀🤖 (langchain)
   • Intel® Extension for Transformers An Innovative Transformer-based Toolkit to Accelerate GenAI/LLM Everywhere - [2023/12] Supported QLoRA on CPUs to make fine-tuning on client CPU possible.
   • Getting Started with Gemini
   • AgentSearch, an open-core effort to make humanity's knowledge accessible for LLM agents.
   • MLX Examples - This repo contains a variety of standalone examples using the MLX framework.
   • Emu2: Generative Multimodal Models are In-Context Learners
   • “사기꾼 찾아내는 AI” 사기 대응 AI 강화하는 금융 업계
   • "성공적 클라우드 전환이 토대" 한 CDIO가 전하는 '전사적 AI' 토대 구축기
   • PowerInfer - 소비자용 GPU를 사용해서 빠르게 LLM 서빙하기
   • 챗GPT, 여전히 숫자에 약해..."금융 분야 사용은 시기상조"
   • ChemCrow: Augmenting large-language models with chemistry tools
   • How to Deploy CogVLM on AWS
   • With the holidays upon us, some of us might be a little behind in our gift shopping. Gemini can help! Connect to the Gemini API and tell it a bit about the person you want to buy for and Gemini will give you a few suggestions!
   • Why 2023 was the most exciting year in computer vision history (so far)
   • Phi-2 is a 2.7B parameter language model released by Microsoft with performance that rivals much larger models. (MLX version)
   • LLM in a flash: Efficient Large Language Model Inference with Limited Memory
   • FastSR-NeRF: Improving NeRF Efficiency on Consumer Devices with A Simple Super-Resolution Pipeline
   • 너도나도 독점 LLM, 클라우드 경쟁 핵심으로
   • Welcome to the "Awesome ChatGPT Prompts" repository! This is a collection of prompt examples to be used with the ChatGPT model.
   • SlimSAM: 0.1% Data Makes Segment Anything Slim
   • A Challenger to GPT-4V? Early Explorations of Gemini in Visual Expertise
   • gsplat.js - JavaScript Gaussian Splatting library
   • Understanding GPU Memory 2: Finding and Removing Reference Cycles
   • Retrieval-Augmented Generation (RAG): From Theory to LangChain Implementation
   • Open Book LLM Science Exam
   • Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks
   • microsoft/promptbase - 프롬프트 관련 자료, 모범사례, 예제 모음 (github.com/microsoft)
   • Calvin - An Open-Source Google Calendar Assistant
   • Microsoft Announces Small Language Model Phi-2
   • Coffee - AI를 이용한 프론트엔드 개발 도우미 (github.com/Coframe)
   • Multimodal RAG pipeline with LlamaIndex and Neo4j
   • KB금융, 내부통제 디지털화 추진...AI·RPA 활용
   • abracadabra: How does Shazam work?
   • The T5 models are encoder-decoder models pre-trained on a mixture of unsupervised and supervised tasks. (MLX version)
   • Bash One-Liners for LLMs
   • "SNS 지고 생성형 AI 뜬다, 마케팅 전략 변화 필요해" 가트너
   • 챗GPT 대응 1년…뛰는 ‘네이버’와 발도 못 뗀 ‘카카오’
   • "구글, 픽셀폰 전용 AI비서 만든다…사물인식 안경도 개발 중"
   • Use Gemini and ChatGPT to learn from two capable teachers
   • 생성형 AI, 구축할 것인가 클라우드에서 살 것인가?
   • 사내 AI SQL 생성 슬랙봇 (m.blog.naver.com)
   • Many options for running Mistral models in your terminal using LLM
   • BricksLLM - LLM을 위한 AI Gateway (github.com/bricks-cloud)
   • [단독] 엔씨, 신사업 'AI 금융' 접는다…조직 해체
   • Building a Universal AI Scraper
   • Unlocking the Future of Data Products: Business focused AI agents team
   • RAVE: Randomized Noise Shuffling for Fast and Consistent Video Editing with Diffusion Models
   • LLM을 이용한 한줄짜리 Bash 스크립트들 (justine.lol)
   • Perspectives on the State and Future of Deep Learning - 2023
   • Resemble Enhance is an AI-powered tool that aims to improve the overall quality of speech by performing denoising and enhancement.
   • 창업자를 위한 AI 스펙트럼 (nfx.com)
   • 2024 기술 트렌드: AI, 그리고 그외 모든 것들 (ben-evans.com)
   • Invoice data processing LLM RAG on CPU with Ollama, LlamaIndex and Weaviate
   • Outliers have led me to 100s of business insights. But first I had to find them. In 6 minutes let me share 6 months of research into outliers.
   • GN⁺: OpenAI의 프롬프트 엔지니어링 가이드 (platform.openai.com)
   • Apple Open-sources Apple Silicon-Optimized Machine Learning Framework MLX
   • "2024년부터 '인공지능'이라는 말이 사라지기 시작할 것"
   • The Illustrated Stable Diffusion
   • Tree of Thought
   • Introduction to Anomaly Detection in Python with PyCaret
   • StoryGPT-V: Large Language Models as Consistent Story Visualizers
   • Prompt engineering by OpenAI
   • GN⁺: OpenAI, ByteDance가 자체 AI 모델 훈련에 GPT를 사용해서 계정 중단 (theverge.com)
   • Awesome System for Machine Learning - A curated list of research in machine learning system.
   • Mathematical discoveries from program search with large language models
   • Boxplots are one of the most useful tools in my Data Science arsenal. In 6 minutes, I'll teach you 6 years of using box plots for EDA and problem-solving. Let's dive in.
   • The Google AI Python SDK enables developers to use Google's state-of-the-art generative AI models (like Gemini and PaLM) to build AI-powered features and applications.
   • UIDraw - 폰에서 그린 그림으로 웹사이트 생성하기 (github.com/jordansinger)
   • Pixel Aligned Language Models
   • lightning.ai - Turn ideas into AI, Lightning fast: Code together. Prototype. Train. Deploy. Host AI web apps. From your browser - with zero setup
   • VideoLCM: Video Latent Consistency Model
   • “구글 AI검색 도입시, 트래픽 40% 하락” 공멸 우려 목소리 터졌다
   • Gemini API for Kaggle model
   • Gemini API Starter Notebook for Kaggle
2. 하드웨어
   • Everything You Need to Know about Fast Charging Your iPhone
   • GN⁺: 애플, 클라우드가 아닌 자체 하드웨어에서 직접 실행되는 AI를 원해 (arstechnica.com)
   • amx - If you are really interested in doing serious compute on Apple devices, check this:
   • Apple wants AI to run directly on its hardware instead of in the cloud
   • Ollama - Get up and running with large language models locally.
   • '인텔 4공정과 NPU 내장'··· 인텔 코어 울트라 칩 살펴보기
   • 2023년은 오픈 LLM의 해 (huggingface.co)
   • 인텔 CEO는 엔비디아의 '매우 운이 좋은' AI 지배력을 한탄하며 인텔이었어야 했다고 주장합니다.
   • osmos-2 released by @Microsoft is a very underrated model that can describe the image and answers questions about it, without hallucinating ✨
   • Which LLMs are best at using tools?
   • Just found out about the Chatbot Arena, it's brilliant.
   • 바이트댄스, 오픈AI API로 자체 챗봇을 개발하다 들통
   • 생성형 AI 도입에 집중하는 금융권
   • A Challenger to GPT-4V? Early Explorations of Gemini in Visual Expertise
   • Awesome-Multimodal-Large-Language-Models - 🔥🔥🔥 A Survey on Multimodal Large Language Models
   • AMD MI300X, Nvidia H100보다 30% 향상된 성능을 보여 (tomshardware.com)
   • Performance of llama.cpp on Apple Silicon A-series
   • “AI 에브리웨어 시대를 연다” 인텔 코어 울트라 칩 살펴보기
   • Inter-node and intra-node Networking Hardware (GPU 네트워킹 성능 평가)
   • 빠른 시작: Node.js 애플리케이션에서 Gemini API 시작하기
3. 읽을거리
   • PISA 2022 results(OECD 국가 수학 성적 통계 포함)
   • GN⁺: 모더나의 mRNA 암 백신, 예상보다 더 효과적임 (freethink.com)
   • The science of gift wrapping explains why sloppy is better
   • 28년 군 인생 뒤흔든 박정훈 대령의 맹세 [2023 올해의 인물]
   • 월급 관리의 기초
   • 미래에셋 "2차 베이비부머 세대 평균 자산 7.5억…83%가 부동산"
   • 두 가게
   • [ESC] 응답하라 ‘코끼리표 밥솥’···추억의 밥솥사
   • 중국 성장구조 전환과정과 파급영향 점검
   • “‘5공 전사’ 등 고증 탁월…악인에 분노하기보다 근본적 원인에 분노를”
   • 테무는 알고 있다. 우리 모두가 값싼 물건에 중독되어 있음을
   • How did they build the ISS? (International Space Station)
   • 2023 한국 부자 보고서 by KB 경영연구소
   • 자산 20억 넘는 부자 46만명…“내년 예적금·주식 투자 확대”

(보너스: Epoch vs Iteration via @levikul09)