[B급 프로그래머] 8월 4주 소식(빅데이터/인공지능, 하드웨어, 읽을거리 부문)

(오늘의 짤방: Reality > Data > Model via @paulabartabajo)

1. 빅데이터/인공지능
   • 구글의 두 번째 의료 특화 LLM, Med-PaLM 2
   • Dive Into LoRA Adapters
   • Beating GPT-4 on HumanEval with a Fine-Tuned CodeLlama-34B
   • Dense Text-to-Image Generation with Attention Modulation
   • "LLM은 문서 초반의 정보는 잘 가져오지만 중간에 있는 건 제대로 못 가져온다" - 인간미가 느껴졌다
   • Perplexity is your AI research assistant. It has a conversational interface, contextual awareness and personalization to learn your interests and preferences over time.
   • Demo of reproduced VALL-E X
   • Very impressive. WizardCoder 34B outperforms GPT-4, Claude, and Bard on HumanEval.
   • Lots of confusion around bf16 vs fp16 for llama-2
   • Ahead of AI #11: New Foundation Models
   • We now have the most comprehensive cookbook on building LLMs with Knowledge Graphs
   • RAG graph example(colab)
   • A strong foundation in Mathematics can help you excel in the field of Data Science! Today, I'll share some top FREE resources on Maths for ML.
   • Fast Vector Similarity Library
   • [Week of 8/21] LangChain Release Notes
   • Embedchain is a framework to easily create LLM powered bots over any dataset.
   • How do we 𝗗𝗲𝗰𝗼𝗺𝗽𝗼𝘀𝗲 𝗥𝗲𝗮𝗹 𝗧𝗶𝗺𝗲 𝗠𝗮𝗰𝗵𝗶𝗻𝗲 𝗟𝗲𝗮𝗿𝗻𝗶𝗻𝗴 𝗦𝗲𝗿𝘃𝗶𝗰𝗲 𝗟𝗮𝘁𝗲𝗻𝗰𝘆 and why should you care to understand the pieces as a ML Engineer?
   • Automated Metadata Extraction for Better Retrieval + Synthesis
   • 🌲Multi Vector Retriever
   • AI2 Dolma: 언어모델을 위한 3T 토큰 오픈 코퍼스 (blog.allenai.org)
   • MS에 빼고 총집합?··· 구글, 아마존, 엔비디아 등 허깅페이스에 2억 달러 투자
   • Visualizations of Embeddings
   • "챗GPT 언급한 채용 공고, 21배 증가" 채용 시장의 AI 기술과 기업 동향
   • Llama 2 learns to code
   • "AI 기술 없으면 일자리가 위험하다" IBM-옥스포드 이코노믹스 보고서
   • “윈도우 그림판에 새로운 AI 기능 추가될 수 있다”
   • K-Means Clustering - An Explorable Explainer
   • Fine-Tuning Embeddings for RAG with Synthetic Data
   • Introducing Code Llama, a state-of-the-art large language model for coding
   • "잠재력은 인정, 열광은 조심”··· 과장 징후의 기술 4가지
   • Check out these new guides for 13 popular LLM use-cases.
   • 데이터 양이 ADAS 훈련에 정말 최선일까? - 열등한 데이터 세트, 사고 위험 증가시킬 수 있다
   • One major way to improve your RAG system is to fine-tune your embedding model ⚙️
   • 메타, 멀티모달 AI 모델 '심리스M4T' 발표 "최대 100개 언어 지원"
   • "쇼핑객 17%가 구매 위해 생성형 AI 사용" 세일즈포스 보고서
   • Understanding Transformers: A Step-by-Step Math Example — Part 1
   • Code Llama - 코딩을 위한 최첨단 거대 언어 모델 (ai.meta.com)
   • 오픈소스의 교훈을 생성형 AI에 적용하자
   • IBM, '코드 어시스턴트' 코볼에서 자바로 자동 변환
   • 김준구 네이버웹툰 CEO “저작권 논란 없는 AI툴 개발 중”
   • Google "We Have No Moat, And Neither Does OpenAI"
   • OpenCopilot solves this problem so building your own Copilot becomes intuitive, fast and reliable - all so you can build your copilot in a single day.
   • Whisper API - Speech to Text Transcription
   • fai - Everything you need to launch your AI product
   • Using LangSmith to Support Fine-tuning
   • Run more pods per GPU with NVIDIA Multi-Instance GPU
   • GPT-3.5 파인 튜닝 후기
   • The Impact of chatGPT (2023) by MIT Department of Physics
   • Salesforce, 5.3조원 가치로 AI스타트업 HuggingFace에 투자를 진행중 (theinformation.com)
   • Semantic search with embeddings: index anything
   • Mind Maps plugin for ChatGPT
   • Will be interesting how this LoRA-on-demand service will compare to open-source LoRA on prem.
   • Generative AI for Blender - AI generate video, image, and audio from text prompts or video, image, or text strips.
   • A Math Lover’s Guide to Hidden Markov Models
   • functime is a powerful Python library for production-ready global forecasting and time-series feature engineering.
   • Fine Tuning GPT-3.5-Turbo
   • OPENAI IS FUNDING AN APP FOR PARENTS TO MANAGE KIDS' LIVES
   • IBM says GenAI can convert that old COBOL code to Java for you
   • A few quick thoughts on why exploring multilingual language models would be beneficial…
   • Awesome 한국어 음성인식 (github.com/rtzr)
   • 11M document subset of OBELICS: an open collection of interleaved image-text web documents, containing 141M English documents, 115B text tokens, and 353M images.
   • LoRA-style LLM finetuning is now available for all models in Lit-GPT
   • Less than 23 hours since the new Midjourney feature "In-Painting" dropped..
   • SeamlessM4T is designed to provide high-quality translation, allowing people from different linguistic communities to communicate effortlessly through speech and text.
   • Cloud GPU 가이드 - AI에 어떤 GPU를 어디서 써야할까? (gpus.llm-utils.org)
   • [Tip] 스노우가 2달 동안 150억 번 AI 프로필 이미지 만드는 방법 📷 (코드 있음)
   • Topic Modeling with Llama 2 - Create easily interpretable topics with Large Language Models
   • SeamlessM4T is designed to provide high quality translation, allowing people from different linguistic communities to communicate effortlessly through speech and text.
   • Here are a few ways you can use @OpenAI’s new finetuning endpoints to optimize your LLM apps over your data 💡:
   • IDEFICS is an 80 billion parameters multimodal model that accepts sequences of images and texts as input and generates coherent text as output.
   • Benchmarking Question/Answering Over CSV Data
   • Why You (Probably) Don’t Need to Fine-tune an LLM
   • GPT-3.5 Turbo 미세 조정 및 API 업데이트 (openai.com)
   • Outlines 〰 helps developers guide text generation to build robust interfaces with external systems.
   • There’s two ways of performing more structured tagging/retrieval for production-quality RAG systems:
   • Llama from scratch (or how to implement a paper without crying)
   • 챗GPT 연계 이후 돈 잘 버는 AI 서비스 3곳
   • Image embeddings
   • AI-generated art cannot receive copyrights, US court says
   • LLM 연구의 공개 과제들 (huyenchip.com)
   • LlamaIndex + Metaphor: Towards Automating Knowledge Work with LLMs
   • Python Polars: A Lightning-Fast DataFrame Library
   • What motivated self-attention mechanisms in transformer-based LLMs in the first place?
   • OpenAgent is a library of modular components and an orchestration framework.
   • Master Finetuning LLMs on a single GPU!
   • LlamaGPT - A self-hosted, offline, ChatGPT-like chatbot, powered by Llama 2. 100% private, with no data leaving your device. Code Llama support coming soon.
   • NVIDIA AI Workbench Speeds Adoption of Custom Generative AI for World’s Enterprises
   • Excited to share that my newest article "Building Custom Q&A Applications Using LangChain and Pinecone Vector Database" on Analytics Vidhya is out now📝🚀
   • Deep Dive into pandas Copy-on-Write Mode - Part I
   • Five facts: How customer analytics boosts corporate performance
   • Welcome to the NaLLM project repository, where we are exploring and demonstrating the synergies between Neo4j and Large Language Models (LLMs).
   • My Everyday LLM Uses
   • Fooocus - SD 와 Midjourney에서 배운, 더 쉽고 편한 이미지 생성 소프트웨어 (github.com/lllyasviel)
   • pykoi - LLM을 위한 데이터 & 피드백 수집용 UI 라이브러리 (cambioml.com)
   • Become a Google Certified Data Scientist 🚀
   • A core retrieval idea that will lead to better results for your LLM QA system is decoupling embedding representations from raw text chunks (s/o @md_rumpf for inspiration). ✂️
   • Best Practices for Data Cleaning and Preprocessing
   • Beginner’s guide to Llama models
   • 퀴즈쇼 우승 ‘왓슨’, 암치료 나섰지만…결과는 ‘굴욕적 퇴출’
   • GN⁺: Cruise, 충돌 사고 후 즉시 로보택시 편성을 50% 줄이라는 규제당국의 지시를 받음 (techcrunch.com)
   • CoDeF: Content Deformation Fields for Temporally Consistent Video Processing
   • Here are 8 key considerations for building *production-grade* LLM apps over your data (RAG) 💡 (see 🧵):
   • Parent Document Retriever, as well as the cool diagram by @clusteredbytes! ❤️
   • CS224u: Natural Language Understanding (Code for the Stanford course. Spring 2023
   • Which GPU(s) to Get for Deep Learning: My Experience and Advice for Using GPUs in Deep Learning
   • NumPy Examples — Practice Questions Make You an Expert
   • Do LLMs Possess a Personality? Making the MBTI Test an Amazing Evaluation for Large Language Models
   • Incognito Pilot combines a Large Language Model (LLM) with a Python interpreter, so it can run code and execute tasks for you. It is similar to ChatGPT Code Interpreter, but the interpreter runs locally and it can use open-source models like Llama 2.
   • Patterns for Building LLM-based Systems & Products
   • 생성 AI 경제 모델, 기존 AI와 어떻게 다른가?
   • Awesome-Korean-LLM : 한국어 오픈소스 LLM awesome list (github.com/NomaDamas)
   • What's the connection between World War II 🪖 and Classification Models 👨‍💻?
   • Simpson’s Paradox and Interpreting Data
   • 엔씨, 자체 개발 언어모델 VARCO LLM 공개
   • These women fell in love with an AI-voiced chatbot. Then it died
   • CTransformers - Python bindings for the Transformer models implemented in C/C++ using GGML library.
   • The Poisson Hidden Markov Model for Time Series Regression
   • Why You (Probably) Don’t Need to Fine-tune an LLM
   • Should we be more persuasive?
   • Sweep - AI 쥬니어 개발자 (github.com/sweepai)
   • DeepEval - LLM을 위한 유닛 테스팅 (github.com/mr-gpt)
   • Visual Blocks is a framework that allows any platform or application to easily integrate a visual and user-friendly interface for ML creation.
   • 어떤 LLM이 더 낫지?··· 아서, AI 모델 비교·분석 도구 ‘아서 벤치’ 오픈소스로 출시
   • 나날이 풍성해지는 선택지··· 기업용 생성형 AI 모델과 유형 진단
   • “AI 아닌 기술에도 관심 가져야”··· 가트너, 신흥 기술 하이프 사이클 2023’ 발표
   • 'AI로 구글 검색에 도전했지만'...마이크로소프트 빙 점유율 예전 그대로
   • How to Develop LSTM Models for Time Series Forecasting
   • Catching up on the weird world of LLMs
   • Amazon Kendra로 모든 유형에 대한 자료 검색 구축하기 [2부 – 음성 및 영상 검색]
   • DevOpsGPT: AI-Driven Software Development Automation Solution
   • DoctorGPT - 미국 의사 면허 시험을 통과하는 LLM (github.com/llSourcell)
   • 'New York Times' considers legal action against OpenAI as copyright tensions swirl
   • “AI 콘텐츠, 팩트체크 더 세심하게 하라”··· AP, 언론인을 위한 생성형 AI 가이드라인 공개
   • Here is a little prompt we developed that will generate quiz questions for whatever PDF reading you have opened in Edge with the Bing sidebar.
   • “5가지 주요 직업군이 보는 생성형 AI는?”··· 세일즈포스, 연구 결과 발표
   • ‘일관된 판단이 장점’ 오픈AI, 콘텐츠 조정 업무에의 GPT-4 활용 방안 제시
   • 토종 AI의 반격…코난, 자체모델로 "기업시장 공략"
   • ‘K-초거대AI’ 윤곽… 네이버는 유료, 카카오는 무료
   • Department of Health and Human Services: Artificial Intelligence Use Cases Inventory
   • Reflections on “Making the Atomic Bomb”
   • Meta의 OpenAI에 대한 다음 공격: 무료 코드 생성 소프트웨어 (theinformation.com)
   • 크롬 브라우저가 AI를 이용해 전체 기사를 요약 예정 (blog.google)
   • TextFX is an AI experiment that uses Google's PaLM 2 large language model. These 10 tools are designed to expand the writing process by generating creative possibilities with text and language.
   • Roboflow Inference is an opinionated tool for running inference on state-of-the-art computer vision models.
   • Another blockbuster AI startup in town! Congrats David @hardmaru & Llion @YesThisIsLion on launching http://sakana.ai!
   • What Smart Companies Know About Integrating AI
   • It is fascinating how quickly our mediocre-but-good-enough-until-ChatGPT tools for identifying humans from machines have failed.
   • Meet Decicoder, an LLM proficient at writing code
   • Introducing: LoRA the Explorer 🤠🚀 Explore, try out, and download curated SDXL LoRAs to generate images with your favorite styles! Crayon, pixel art, 3D render, cyborg, watercolor, and much more!
   • Teach LLMs to Personalize
   • Finetuning LLaMa + Text-to-SQL
   • 토종 생성형 AI 출격 앞둔 네이버, 상반기 R&D에 1조 썼다
   • 생성 AI의 엔비디아 GPU 탈출은 가능할까 | 학습 인프라는 상황 종료...추론에 틈새 보여
   • Stanford XCS224U: Natural Language Understanding I Spring 2023 (Stanford Online)
   • GPTs are GPTs: An Early Look at the Labor Market Impact Potential of Large Language Models
   • Stable Diffusion XL playground by NVidia
   • NVidia NGC AI models(pretrained)
   • 구글 헬스, 엑스레이 분석 솔루션 업체 아이캐드에 AI 지원
   • Subs AI - OpenAI Whisper로 동영상의 자막 자동 생성하기 (github.com/abdeladim-s)
   • 가트너 ”생성형 AI 기대감 최고조” | 2023년 이머징 테크놀로지 하이프 사이클 발표
   • Datasette Cloud, Datasette 1.0a3, llm-mlc and more
   • AI Has Already Created As Many Images As Photographers Have Taken in 150 Years. Statistics for 2023
   • GN⁺: Opendream: Stable Diffusion을 위한 레이어 기반 UI (github.com/varunshenoy)
   • 챗GPT 넘어라··· 다양한 LLM 활용이 필요한 때
   • 생성 AI 만난 포토샵, 빠진 이미지 다 그려 넣어주네 | 어도비, '생성형 확장' 기능 공개…편집 절차 간소화
   • 생성AI 모델 레이스 새국면...구글, '제미니'로 하반기 대공세
   • FoodSAM: Any Food Segmentation
   • 삼성표 '생성형 AI' 나온다…10월 테스트 시작
   • Trustworthy LLMs: a Survey and Guideline for Evaluating Large Language Models' Alignment
   • OpenAI Adapter (langchain)
   • 🦜🔗 Text-to-graph playground - This playground explores the use of OpenAI functions and LangChain to build knowledge graphs from user-input text.
   • Do we really need a specialized vector database?
   • Creating a chatbot that can accurately answer questions about a product or service's documentation is a complex task. You can fine-tune on the documentation itself (but results won't be conversational), use embeddings (risks losing relevant context), etc. gpt-oracle-trainer is an experimental tool that aims to simplify this process and potentially produce better results.
   • Introducing DeciCoder: A new open-source LLM, specialized for generating code in Python, Java, and Javascript.
   • Did you know you could finetune Llama2 on Kaggle notebooks with multi-gpu for FREE?! Check out this notebook:
   • Deploy LLM on Kubernetes using OpenLLM
   • New breakthrough from Meta AI: Instruction Backtranslation. Applied to LLama this method outperformed Claude, Guanaco, LIMA, and Falcon-Instruct.
   • We present “Graph RAG” in @llama_index : a new method of augmenting LLMs with context from a graph database 🔎✨
   • GN⁺: 아마존, 제품 리뷰를 요약하는 생성형 AI 기능 출시 (apnews.com)
   • stable-diffusion.cpp - Inference of Stable Diffusion in pure C/C++
   • Anti-hype LLM reading list
   • 📐 The 🤗 Open Object Detection Leaderboard aims to track, rank and evaluate vision models available in the hub designed to detect objects in images.
   • How is LLaMa.cpp possible?
   • Understanding the patterns of the universe - the omnipresent normal distribution
   • I was surprised by a talk Yejin Choi (an NLP expert) gave yesterday in Berkeley, on some surprising weaknesses of GPT4: As many humans know, 237*757=179,409 but GPT4 said 179,289.
   • 새로운 루다를 지탱하는 모델 서빙 아키텍처 — 3편: 안정적인 LLM 서비스를 위한 서빙 최적화 기법
   • ChatGPT answers more than half of software engineering questions incorrectly - You may want to stick to Stack Overflow for your software engineering assistance.
   • PyTorch Model Performance Analysis and Optimization — Part 3
   • In bioinformatics, a sequence logo is a graphical representation of the sequence conservation of nucleotides (in a strand of DNA/RNA) or amino acids (in protein sequences).
   • How to Build a Real-Time Feature Pipeline In Python
   • The ChatGPT hype cycle:
   • Stable Diffusion Crash Course for Beginners
   • STUDY: Socially aware temporally causal decoder recommender systems
   • FastViT: A Fast Hybrid Vision Transformer using Structural Reparameterization
   • Benchmarking Question/Answering Over CSV Data
   • Multi-modality is another 10x UX improvement in the same way that chat was.
   • LLMOps: Deploy Open LLMs using Infrastructure as Code with AWS CDK
   • I created a simple example data science project. It includes a makefile that you can copy and use to easily make your own.
   • Fooocus is an image generating software. Fooocus is a rethinking of Stable Diffusion and Midjourney’s designs:
   • 인프콘 2023 핸즈온, ChatGPT API: 다양한 인공지능 서비스 개발하기 자료입니다.
   • Embedding speed may matter as much as performance if you're building RAG with local models! ⚡️ We ran a simple benchmark on @huggingface embedding models 🤗 - throughput as a function of model / batch size / string length.
   • 바른팀에서 구축한 한국어 모호성 평가 데이터세트를 소개합니다.
   • 최고 성능의 한국어 형태소 분석기 「바른」을 무료로 사용해 보세요.
   • ✅Image Classification in ML/Deep Learning- Explained in simple terms with implementation details
   • ⚡ Lit-GPT - Hackable implementation of state-of-the-art open-source large language models released under the Apache 2.0 license.
   • DoctorGPT is a Large Language Model that can pass the US Medical Licensing Exam.
   • Wrapper's Delight: An Enhanced OpenAI Wrapper
   • Zep - LLM/Chatbot을 위한 장기 메모리 저장소 (github.com/getzep)
   • 🤗 PEFT - State-of-the-art Parameter-Efficient Fine-Tuning (PEFT) methods
   • 네이버 “AI 집중”…판교 빌딩 팔아 현금 3000억대 확보
   • Tree-based ML algorithms - why they are the most popular models in the real world
   • Neuralangelo - This is the official implementation of Neuralangelo: High-Fidelity Neural Surface Reconstruction.
   • 🥇Top ML Papers of the Week
   • 이상 탐지 3부-머신 러닝으로 이상 탐지하기
   • Google DeepMind’s CEO Says Its Next Algorithm Will Eclipse ChatGPT
   • Dify - 사용하기 쉬운 LLMOps 플랫폼 오픈소스 (github.com/langgenius)
   • 인공지능(AI) 활용서: 6대 산업별 활용사례 by 딜로이트
   • "AI가 클라우드 스타트업 지형도 바꾼다" 2023년 포브스 클라우드 100
   • SKT 글로벌 AI사업 강화 앤스로픽에 1억달러 투자
   • Awesome Key Infomation Extraction
   • Retake - Postgres용 하이브리드 검색 오픈소스 (github.com/getretake)
   • liteLLM - 50개 이상의 LLM을 지원하는 프록시 서버 (github.com/BerriAI)
   • 프폼프트 인스트럭션 개선을 위한 힌트
   • More people should be training their own embeddings Why? Because it costs <$1000 to train a SOTA embeddings model
   • FaceChain is a deep-learning toolchain for generating your Digital-Twin. With a minimum of 1 portrait-photo, you can create a Digital-Twin of your own and start generating personal portraits in different settings (multiple styles now supported!).
   • Meta AI, Vision 모델을 위한 PUG(Photorealistic Unreal Graphics) 데이터셋 공개 (pug.metademolab.com)
   • LLM As DBA
   • 올여름의 불타는 LLM: 세상도 불타고 나도 불타고
   • GPT detectors are biased against non-native English writers
   • AI가 이제 인간보다 CAPTCHA 테스트에서 더 뛰어나
   • PUG: Photorealistic and Semantically Controllable Synthetic Data for Representation Learning
   • AI 관련 종사자 관심 우선순위 ‘생성AI·컴퓨터 비전·데이터 분석·자연어 처리’
   • Code Understanding (langchain)
   • "Do Anything Now": Characterizing and Evaluating In-The-Wild Jailbreak Prompts on Large Language Models
   • llm-attacks - Universal and Transferable Adversarial Attacks on Aligned Language Models
   • Real-Real-World Programming with ChatGPT - Taking AI Far Beyond Small Self-Contained Coding Tasks
   • Introducing English as the New Programming Language for Apache Spark
   • California man's business is frustrating telemarketing scammers with chatbots
   • The Lone Banana Problem. Or, the new programming: “speaking” AI
   • Adversarial Policies Beat Superhuman Go AIs
   • AutoChain takes inspiration from LangChain and AutoGPT and aims to solve both problems by providing a lightweight and extensible framework for developers to build their own agents using LLMs with custom tools and automatically evaluating different user scenarios with simulated conversations.
   • Awesome LLMOps - An awesome & curated list of the best LLMOps tools for developers.
   • An introduction to graph theory
   • Welcome to Apache OpenNLP - The Apache OpenNLP library is a machine learning based toolkit for the processing of natural language text.
   • Zep and LlamaIndex: A Vector Store Walkthrough
   • Llama 2 Powered By ONNX
   • Top Generative AI tools for Data Scientists and Coders
   • Multilevel Regression Models and Simpson’s paradox - Avoiding false conclusions with the proper tooling
   • AI21 Labs Accelerates Generative AI Model Adoption Using Amazon SageMaker
   • Anserini: OpenAI-ada2 Embeddings for MS MARCO Passage Ranking
   • The NeurIPS 2023 LLM Efficiency Challenge Starter Guide
   • QAmeleon introduces synthetic multilingual QA data contaning in 8 langauges using PaLM-540B, a large language model.
   • Metadata Replacement + Node Sentence Window
   • 🦜⚒️Evaluating Q&A Systems with Dynamic Data 🔄
   • 어떻게 효율성을 유지할까?...오픈AI GPT-4의 비밀
   • 엔비디아 '쿠다'에 도전한다...모듈러, 6억달러 가치에 투자 유치 협상 중
   • api2ai - OpenAPI 스펙을 읽어서 코드를 작성해주는 API 비서 (github.com/mquan)
2. 하드웨어
   • 실용 사례 풍성··· 금융 산업은 양자 컴퓨팅에 대비 중
   • “올해 전 세계 AI 칩 매출 534억 달러 이를 것” 가트너 전망
   • AI로 날개 단 엔비디아··· 데이터센터 사업 약진으로 2분기 매출 135억 달러 달성
   • NVIDIA reported earnings yesterday, and wow: The underlying reason for their unprecedented spike in revenue? Cloud providers & data centers buying GPU units in bulk.
   • NVIDIA, 2024 회계 연도 2분기 재무 결과 발표 (nvidianews.nvidia.com)
   • 세계 최초 M1 GPU 드라이버 (rosenzweig.io)
   • I predicted great things for Apple AirPods before it was released.
   • 비츠 스튜디오 프로 vs. 에어팟 맥스 : 애플의 프리미엄 헤드폰 구매 가이드
   • My Overkill Home Network - Complete Details 2023
   • “요즘 노트북 60만원 ‘헐값’돼도 안 사요?” 찬밥된 PC 재고 넘쳐난다
   • Arduino With Python: How to Get Started
   • “장단점부터 선택 기준까지” 올인원 CPU 수랭 쿨러의 모든 것
   • “진짜 돈이 되는 건 따로 있다”...세계 최고 가치 자랑하는 이 기업의 전략
   • 쫓느냐 쫓기느냐.. SDV 선두다툼 벌써부터 '치열'
   • 구글 픽셀 태블릿이 ‘2% 부족한’ 실패작인 이유
   • M1에 대한 찬가 (fabiensanglard.net)
3. 읽을거리
   • Germany reforms citizenship law
   • 네이버페이, “기술로 금융 넓히는 종합 금융 플랫폼으로 본격화"
   • “사무실 임대료 14억 밀렸다” 공개 저격…부러움 샀던 이 회사 무슨 일이
   • 건보료 부과 2단계 개편 1년…"지역보험료 산정 때 전월세 빼야"
   • '국가데이터센터 중단' 파문 확산…"신규 도입 6호기 운영할 돈도 없다"
   • 고물가에 2분기 가계 실질소득 3.9%↓…17년만에 최대폭 감소
   • 월 13조 돌파한 간편결제 시장…‘지갑 없는 세상’ 오는데 입법은 아직[머니뭐니]
   • 한달새 갑자기 오른 달러 환율, 이유는?
   • 한국 IT 역사 5화: 88 서울 올림픽 전산시스템 자이온스 (1부)
   • 한국 IT 역사 6화: 88 서울 올림픽 전산시스템 자이온스 (2부)
   • 한국 IT 역사 7화: 88 서울 올림픽 전산시스템 자이온스 (3부, 완결)
   • ‘이건 경제학이 아니다’라는 말을 들었던 재무경제학의 개척자 [자본시장 이야기]
   • 중국 부동산 위기…모범생들의 추락
   • 부가세 인상…‘나라살림 적자’ 돌파구 되려나
   • Linear TV Viewing Drops Below 50% of U.S. Television Usage for First Time, Streaming Hits Record High: Nielsen
   • '순자산 9.4억, 소득 686만, 소비 427만원'... 중산층이 보는 중산층
   • 스타강사에서 30억 빚쟁이 돼보니...“인생과 게임을 헷갈리지 마세요”
   • "이게 그림이냐"…공무원 관두고 '올인'했다가 '비난 세례'

(보너스: Data center revenue: Nvidia $NVDA vs $AMD vs Intel $INTC via @StockMKTNewz)

EOB

[독서광] 사실은 이것도 디자인입니다

이번 주는 사실은 이것도 디자인입니다를 소개하겠다.

방송 스크립트는 전체 공개되어 있으며, 슬라이드셰어에서 보거나 다운로드 받을 수도 있다.

하이라이트를 요약 정리하면 다음과 같다:

• 00:00 도입
• 00:44 한 페이지 요약
• 01:19 이 책의 목차
• 03:34 무엇이 이 책을 흥미롭게 만드는가?
• 06:10 대상 독자
• 06:40 결론과 마무리

EOB

[B급 프로그래머] 8월 3주 소식(개발/설계/경력관리/보안/클라우드/데이터베이스 관련 소식 정리)

(오늘의 짤방: 이 짤은 정말 보면 볼수록 희대의 명언인 듯. via @youngNrich777)

개발 관련 소식

1. 팁과 유틸리티
   • “작지만 알차다” 마이크로소프트 워드에 숨겨진 깨알 기능 9가지
   • 아직도 모르는가? 이제라도 써봐야 할 아이폰 기능 6가지
   • 새로워진 iOS 17 전화 경험에 불만 가질 필요가 없는 이유
   • 에어팟이 ‘말썽’이라면, 간단한 해결 방법 6가지
   • GN⁺: Cloudflare 가 제공하는 인터넷 속도 테스트 (speed.cloudflare.com)
   • How-To : 애플 워치 배터리 수명 늘리는 꿀팁
   • How Well Can You Hear Audio Quality?
   • 웹툴 - 다양한 도구들
2. 프로그래밍
   • 나의 사이드 프로젝트를 위한 기술 스택 2023 (jbernier.com)
   • 하시코프의 라이선스 변경과 오픈소스의 본질
   • How to convert an enum to string in C++
   • Trafilatura is a Python package and command-line tool designed to gather text on the Web.
   • The Cross-Domain Thesis Part 1: Setting The Stage
   • Python Dictionary Iteration Quiz
   • What it means when you convert between different shared_ptrs
   • 5 GitHub Repos to Master JavaScript:
   • Show HN: I wrote a RDBMS (SQLite clone) from scratch in pure Python
   • ZGC의 기본 개념 이해하기
   • You're the OS game
   • hyperfine = A command-line benchmarking tool.
   • Quirks of Python package versioning
   • 5 ways to improve your system maintainability that will make your life easy.
   • 웹 콘텐츠 속 숫자의 접근성 높이기
   • 나 4년 차 서버개발자, 배달의민족의 지리 체계를 뒤흔들다
   • Concurrency and async / await (FASTAPI)
   • 제목 스타일 단계는 문서 구조의 기둥과 보
   • [I/O] Python의 Selectors
   • Infisical – 오픈소스 HashiCorp Vault 대안 (github.com/Infisical)
   • How to profile a FastAPI asynchronous request
   • PokerKit is an open-source Python library for simulating poker games and evaluating poker hands, developed by the University of Toronto Computer Poker Research Group.
   • This repository is the central place for Rust development of the libp2p spec.
   • Inside STL: Smart pointers
   • Talent Plan is an open source training program initiated by PingCAP. It aims to create or combine some open source learning materials for people interested in open source, distributed systems, Rust, Golang, and other infrastructure knowledge.
   • Secret of Binary ELF
   • The downsides of C++ Coroutines
   • VirGL - QEMU VM내에서 사용 가능한 Virtual 3D GPU (docs.mesa3d.org)
   • {fmt} is an open-source formatting library providing a fast and safe alternative to C stdio and C++ iostreams.
   • Narrowlink is a self-hosted platform that allows you to establish secure remote connections between devices within a network that may be hindered by network address translation (NAT) or firewalls.
   • Transcoding UTF-8 strings to Latin 1 strings at 18 GB/s using AVX-512
   • Unicode & Character Encodings in Python: A Painless Guide
   • Jujutsu is a Git-compatible DVCS. It combines features from Git (data model, speed), Mercurial (anonymous branching, simple CLI free from "the index", revsets, powerful history-rewriting), and Pijul/Darcs (first-class conflicts)
   • permission.site - 브라우저에서 permission이 필요한 기능(API)에 대한 테스트를 해 볼 수 있는 사이트.
   • Orb - 윈도우즈와 비슷환 환경을 웹브라우저에 구현해주는 프리/오픈소스 웹 데스크탑 (gitlab.com)
   • Exploring the internals of Linux v0.01
   • JIMP - 의존성 없는 Node.js용 이미지 프로세싱 라이브러리 (github.com/jimp-dev)
   • Enhance is an HTML-first full-stack web framework that gives you everything you need to build standards-based multi-page web apps that perform and scale.
   • Meta developer tools: Working at scale
   • Polyrhythmix (Poly) is a command-line assistant designed to generate MIDI files from the description of drum parts.
   • Don’t Confuse Complex with Complicated, Part 1: What is Information?
   • Rich는 터미널에서 풍부한(rich) 텍스트와 아름다운 서식을 지원하기 위한 파이썬 라이브러리입니다.
   • dpv (dee-pee-vee) is a dead simple alternative to pyenv-virtualenv and virtualenvwrapper
   • ECE4960 Lectures FA16/SP17 - Cornell ECE 4960 Computational and Software Engineering spring 2017, by Edwin Kan
   • Python Wheels
   • tonic - A rust implementation of gRPC, a high performance, open source, general RPC framework that puts mobile and HTTP/2 first.
   • Briefcase is a tool for converting a Python project into a standalone native application.
   • 닌자 코드 - 닌자라 불리던 전설 속 개발자들은 유지보수 담당 개발자를 혹독하게 훈련하고자 (아래에서 소개해 드릴) 다양한 편법을 사용하곤 했습니다.
   • Meta, 파이썬용 차세대 자동 수정 린터 Fixit 2 공개 (engineering.fb.com)
   • Java class data sharing upgrade would boost startup times
   • Java 21 - Java 17 = 42 JEPs view
   • Unit Test Frameworks for C#: The Pros and Cons of the Top 3
   • Biggest scam in software dev? Best Practices.
   • Debugging .NET Containers with Visual Studio Code Docker Tools
   • C# Source Generators is a Roslyn compiler feature introduced in C#9/.NET 5.
   • "Rust 기초 프로그래밍 + 인터프리터 만들기" 강의 자료 (github.com/utilForever)
   • Panther - Is A Fast & Friendly Web Framework For Building Async APIs With Python 3.11+
   • How to clear Docker cache and free up space on your system
   • 🖌 egui: an easy-to-use GUI in pure Rust
   • 고등학생이 매점에서 간편결제를 쓰고 싶어서 직접 개발한 이야기 (tilnote.io)
   • LPython is a Python compiler.
   • The OSSU curriculum is a complete education in computer science using online materials.
   • Python, foreign functions and Steam
   • Introducing CMake Debugger in VS Code: Debug your CMake Scripts using Open-Source CMake Debugger
   • Ruff - An extremely fast Python linter, written in Rust.
   • 딥링크(Deeplink) : URI스킴, 유니버셜 링크, 앱링크 구분과 이해
   • Typograms - 기술 문서에서 다이어그램을 표현하기 위한 텍스트 기반 경량 이미지 포맷 (google.github.io)
   • 파이썬에서 GIL 삭제된다⋯“병렬 처리의 혁신적 진전”
   • Fantastic Learning Resources
   • WASM을 이용한 NES 에뮬레이터 만들기
   • What helps people get comfortable on the command line?
   • Yes, C is not an object-oriented language, but you can still mimic polymorphism with one clever trick. Let's see how CPython does it.
   • 정보 이론에서 프랑스 이론까지: 야콥슨, 레비-스트로스, 그리고 사이버네틱스 장치 (버나드 디오니시우스 게이건)
   • Behind "Hello World" on Linux(2023 version)
   • Behind "Hello World" on Linux(2013 version)
   • WebAssembly System Interface
   • JavaScript 개발자를 위한 RegEx 책(전체 공개) (freecodecamp.org)
   • Mapkick.js - 한줄의 JS코드로 아름답고 인터랙티브한 지도 만들기 (github.com/ankane)
   • Iconoir - 오픈소스 SVG 아이콘 라이브러리 (iconoir.com)
   • This page documents the time-complexity (aka "Big O" or "Big Oh") of various operations in current CPython.
   • “더 쉽고 빠른 파이썬” 파이파이(PyPy)의 이해
   • owntracks-cloudflare-supabase - Track iOS or Android device's location in a custom database.
   • GN⁺: 구글 지도는 눈에 거슬립니다. 구글 지도가 실패한 사례들 (fastcompany.com)
   • WarpStream : Kafka는 죽었다, Kafka 만세! (warpstream.com)
   • Level Up Coding - Complex programming concepts explained with simple terms and stunning visuals.
   • awesome-falsehood - A curated Awesome list of falsehoods programmers believe in.
   • Barco - C로 밑바닥부터 코딩한 리눅스 컨테이너 (github.com/lucavallin)
   • Name Checker - Find out if your project name is taken
   • Become An Expert: Backend Projects That Define Senior Developers
   • helix - A Kakoune / Neovim inspired editor, written in Rust.
   • fd는 find 를 대체할 수 있는 간단하고, 빠르고, 그리고 사용자 친화적인 commad line util이다.
   • Running async code from sync in Python asyncio
   • VanJS - 1KB Reactive UI 프레임워크 without React/JSX (vanjs.org)
   • BEHIND THE CODE: The one who created languages
3. DevOps
   • Resiliency and Disaster Recovery with Kafka
   • AKS로 쿠버네티스 시작하기 : 시행착오 줄이기
   • Terraform Drift: The Bad, the Ugly and the Black Swan
   • Architecting Kubernetes clusters — choosing a worker node size
   • eks-node-viewer is a tool for visualizing dynamic node usage within a cluster.
   • How to optimize Kubernetes resource configurations for cost and performance
   • kube-s3 - Shared storage with S3 backend
   • Vector is a high-performance, end-to-end (agent & aggregator) observability data pipeline that puts you in control of your observability data. Collect, transform, and route all your logs, metrics, and traces to any vendors you want today and any other vendors you may want tomorrow.
   • How “It works in my machine” turns “It works in my container”?
   • A Brief DevOps History: Databases to Infinity and Beyond
   • OpenTelemetry .NET
   • ELK 기반 SRE 환경 만들기 #3 | Data Pipeline 구현 꿀팁!
   • Mastering Kubernetes Observability: A DevOps Engineer’s Guide
   • Programming, Motherfucker Do you speak it?
   • CloudWatch Agent에서 HTTP Proxy를 통하는 모니터링 방법
   • Some tactics for writing in public
   • Understanding Python imports, __init__.py and pythonpath — once and for all
4. 설계
   • Tales of Kafka at Cloudflare: Lessons Learnt on the Way to 1 Trillion Messages
   • [배민스토어] 배민스토어에 이벤트 기반 아키텍처를 곁들인…
   • Building a Reliable Kafka Data Processing Pipeline With Lily Mara
   • Evolving the Federated GraphQL Platform at Netflix
   • Many applications requires to generate unique IDs in their backend. This is an easy task in a single server, but it's harder at large-scale. Here are 3 effective strategies you can use:
   • Performance isolation in a multi-tenant database environment
   • Tomato Architecture - A Pragmatic Approach to Software Design
   • Enhancing Your "Definition of Done" Can Improve Your Minimum Viable Architecture
   • Ubiquitous Caching: a Journey of Building Efficient Distributed and In-Process Caches at Twitter
   • Textual Paint - MS Paint in your terminal.
   • Apache Kafka Patterns and Anti-Patterns
   • The Rise of the Serverless Monoliths
   • Riverbed: Optimizing Data Access at Airbnb’s Scale
5. 경력 관리와 개발문화
   • Reducing Cognitive Load on Startup Engineers
   • "취업 위해 졸업 미뤘는데 결국 둘다 멀어져"…대학가 퍼진 무기력증
   • 자금난에 문 닫는 스타트업 속출...VC들 점점 신중모드로
   • 'IT 공룡' 카카오마저 …"30대 초반부터 대상이죠" 전방위 확산 / JTBC News
   • [배민스토어] 신입 개발자 배민스토어 6개월 생존기
   • 리더는 '진실의 소스'여야··· AI 확산 속 ‘직원 신뢰’ 구축하기
   • The tech job recession is over
   • 미국가서 중국어 공부하지 않기
   • Who's Only Looking Busy at Work?
   • 슬랙, ‘가장 일하는 척하는 시간이 많은 나라’ 순위 공개
   • 실패는 쓰디 쓰지만 오히려 좋아요
   • MS, 대면 근무가 필요한 시나리오 3가지 제시 "만남이 특히 유효한 순간은..."
   • 아마존, 유연 근무제 철회 본격화 '출근 부족 직원에 경고 서한 발송'
   • ESTIMATION GUIDE - get a dev to estimate - increase the number and time unit by one
   • 개발자 컨퍼런스의 계절 – 대기업과 유니콘의 참전
   • 좋은 팀장은 어떻게 만날 수 있을까
   • "취업 위해 졸업 미뤘는데 결국 둘다 멀어져"…대학가 퍼진 무기력증
   • Latency and Throughput for Systems Design Interview
   • 6 Archetypes of Broken Ownership
   • Strategy for Engineering Managers
   • 프로그래머 이력서와 코딩 과제, 검토자는 무얼볼까?
   • The 2023 Tech Market, as Seen by Hiring Managers
   • 피할 수 없는 기술 부채의 덫
   • 사무실 복귀 확산...거품 빠진 화상회의 플랫폼 업계 재편 급물살
   • The Future of Remote Work
   • “요새 신입사원은 왜 그럴까?” 숫자로 본 ‘MZ 세대’ 특징 4가지
   • "사무실 복귀하라" 다른 누구도 아닌 줌의 하이브리드 근무 결정
   • GN⁺: $10M에서 $100M+ ARR까지: CFO로서 배운 5가지 교훈 (openviewpartners.com)
   • Product and Platform Engineers
   • Make Your Meetings a Safe Space for Honest Conversation
   • 마지막 1% (jaredramsey.com)
   • Engineering Leadership Tactics: Building Alignment
   • [독서광] 프로덕트 매니지먼트(본인이 작성한 글)
   • 창업자처럼 일하는 데이터 팀 빌딩 : 3년 로드맵 만들기
   • Average Manager vs. Great Manager
   • Why Emotionally Intelligent Minds Embrace the 3-Question Rule
   • Stop Being a Junior
   • 무엇이 개발자를 생산적이게 만드는가 (jeremymikkola.com)
   • You’re Never Going to Be “Caught Up” at Work. Stop Feeling Guilty About It.
   • Engineering as Art: Embracing Creativity beyond Science
   • What to Do When Work Is Slow
   • An Engineering Manager’s Guide to Success
   • 비전공자 개발자 취업, 어떻게 준비해야 할까?

보안/클라우드/데이터베이스 관련 소식

1. 보안
   • “100초 움직임 데이터만으로 사용자 특정” VR 헤드셋 시대, 프라이버시가 위험하다
   • “피싱은 여전히 가장 지배적인 인터넷 범죄” 클라우드플레어 2023 피싱 위협 보고서
   • 강은성의 보안 아키텍트ㅣ드디어 통합된 개인정보의 안전성 확보조치 기준
   • ‘민감한 데이터 워크플로우 보호’… 몽고DB, 쿼리 가능한 암호화 기술 적용 발표
   • Excellent introduction to cryptography concepts for beginners with practical examples in Linux (openssl)
   • Testssl.sh – Testing TLS/SSL Encryption Anywhere on Any Port
   • It's 2023 and memory overwrite bugs are not just a thing, they're still number one
   • Social engineering campaign targeting tech employees spreading through npm malware
   • Thousands of images on Docker Hub leak auth secrets, private keys
   • Firejail is a SUID program that reduces the risk of security breaches by restricting the running environment of untrusted applications using Linux namespaces and seccomp-bpf.
   • Top 4 Forms of Authentication Mechanisms
   • “키보드 타이핑 소리로 비밀번호 훔친다··· 정확도 95%” 영국 대학 연구진
   • 마침내 ‘라스트패스’를 삭제한 이유 4가지
   • 키패스XC 리뷰ㅣ“꼭 필요한 기본만 담은” 무료 오프라인 비밀번호 관리자
   • 구글 클라우드, ‘2023년 3분기 위협 지형’ 보안 보고서 발표
   • Google AMP - 피싱 사이트의 최신 회피 전략 (cofense.com)
   • Portable Secret - 암호를 안전하게 보관하기 (mprimi.github.io)
   • 구글, ‘개인정보 표시되는 검색 결과’ 뜨면 자동으로 알려준다
2. 클라우드
   • 구글 클라우드, 가격 책정 API 출시 … ‘비용 최적화 지원’
   • Network Load Balancer now supports security groups
   • Working backwards: The story behind the AWS Cloud Development Kit
   • 오라클, ‘클라우드@커스터머’ 상품에 컴퓨트 서비스 추가
   • 엔터프라이즈 DNA로 클라우드 시장 겨냥하는 오라클
   • (AWS) Cost optimization flywheel
   • AKS로 쿠버네티스 시작하기 : 통합본
   • AWS Load Balancers: A Guide to Key Concepts and Features
   • 클라우드 전환 지름길··· 레거시 앱 현대화 요령 10가지
   • 온프레미스보다 더 비싼 클라우드
3. 데이터베이스
   • Postgres 16 beta3 and the Insert Benchmark on a medium server
   • INDEXING “LIKE” IN POSTGRESQL AND ORACLE
   • SQLite compiled to JavaScript
   • A Database in your Browser in sqlite3 Steps
   • The Great Re-shard: adding Postgres capacity (again) with zero downtime
   • What programming languages do your applications that communicate with MariaDB use?
   • Querying Postgres Tables Directly From DuckDB
   • Retake - Postgres용 하이브리드 검색 오픈소스 (github.com/getretake)
   • chDB is an embedded SQL OLAP Engine powered by ClickHouse
   • What is SpacetimeDB? - You can think of SpacetimeDB as both a database and server combined into one.
   • TiKV is an open-source, distributed, and transactional key-value database.
   • dbdiagram - Draw Entity-Relationship Diagrams, Painlessly 😎
   • When Did Postgres Become Cool?
   • Jailer is a tool for database subsetting and relational data browsing.
   • one of my favorite features of sqlite is “:memory:”
   • The Past, Present, and Future of Data Architecture
   • Cozo - Datalog로 쿼리 가능한 임베더블 GraphDB 오픈소스 (github.com/cozodb)
   • The Taming of the B-Trees
   • Hydra - Column-Oriented Postgres 오픈소스 (github.com/hydradatabase)
   • Understanding partitioning and sharding in Postgres and Citus

EOB

[B급 프로그래머] 8월 2주 소식(빅데이터/인공지능, 하드웨어, 읽을거리 부문)

(오늘의 짤방: via @miniapeur)

1. 빅데이터/인공지능
   • Emerging LLM Application Architecture
   • A cheat sheet explanation of how Large Language Models work:
   • Anthropic, Claude Instant 1.2 출시 (anthropic.com)
   • Nvidia reveals new A.I. chip, says costs of running LLMs will ‘drop significantly’
   • Candle is a minimalist ML framework for Rust with a focus on performance (including GPU support) and ease of use. Try our online demos: whisper, llama2.
   • In-Context Alignment: Chat with Vanilla Language Models Before Fine-Tuning
   • Making AMD GPUs competitive for LLM inference
   • Fine-Tuning Llama-2: A Comprehensive Case Study for Tailoring Models to Unique Applications
   • WizardLM is on fire! Seems they released a new model, WizardMath 1 hour ago, that outperforms chatgpt on math skills:
   • Tutor-GPT is a LangChain LLM application. It dynamically reasons about your learning needs and updates its own prompts to best serve you.
   • Personal co-pilot trained on top 10 HF repos by stars using QLoRA in colab on 1 A100 40GB going brrr in vs code 🔥🧑🏽‍💻🚀
   • nanoLoRA - A Minimalistic Implementation of Low-Rank Adaptation
   • Create Your Own Custom LLM Chatbot - Impressive step-by-step tutorial explaining how to choose the best LLM and the components needed for building your own custom LLM-powered chatbot.
   • Gartner Identifies Top Trends Shaping the Future of Data Science and Machine Learning
   • Towards Generalist Biomedical AI
   • Follow Anything: Open-set detection, tracking, and following in real-time
   • 언어데이터과학 (2023학년도 2학기, 서울대학교 언어학과)
   • 딜로이트 , ‘인공지능 활용서' 발간...소비자 부문, 에너지·자원, 산업재, 금융 등 6대 산업군 AI 활용 사례 및 이점 분석
   • StabilityAI, 코드를 위한 LLM 생성형 AI "StableCode" 릴리즈 (stability.ai)
   • 🐍📰 Prompt Engineering: A Practical Example
   • Ask like a human: Implementing semantic search on Stack Overflow
   • Beyond prompting: getting production quality LLM performance with Snorkel Flow
   • Evaluating Correctness and Faithfulness of Instruction-Following Models for Question Answering
   • Rift is open-source infrastructure for AI-native development environments. Rift makes your IDE agentic.
   • FlagEmbedding can map any text to a low-dimensional dense vector which can be used for tasks like retrieval, classification, clustering, or semantic search. And it also can be used in vector database for LLMs.
   • Anomaly Detection in Time Series using ChatGPT
   • 사무원은 가고 마법사가 왔다··· 생성형 AI가 바꿔내는 데이터베이스 분야
   • One-Click Observability(LlamaIndex)
   • Generative Agents: Interactive Simulacra of Human Behavior
   • 🦜⚒️Q&A System Correctness 🧠
   • Building LLM applications for production
   • Getting Started With LLMs(Python · Kaggle - LLM Science Exam)
   • Welcome to , the world's most extensive scholarly knowledge graph with over 26 billion RDF triples.
   • SemOpenAlex: The Scientific Landscape in 26 Billion RDF Triples
   • pgvecto.rs is a Postgres extension that provides vector similarity search functions.
   • Here is how you can obtain a massive speedup with llama-v2 models, much faster than anything else I tried.
   • Getting from Generative AI to Trustworthy AI: What LLMs might learn from Cyc
   • MS, 엔비디아 H100 GPU 서비스 정식 출시
   • LanceDB is an open-source database for vector-search built with persistent storage, which greatly simplifies retrevial, filtering and management of embeddings.
   • devlooper is a program synthesis agent that autonomously fixes its output by running tests!
   • Has Progress on Data, Analytics, and AI Stalled at Your Company?
   • Using Xorbits Inference to Deploy Local LLMs - in 3 steps!
   • GN⁺: 2023년의 AI 현황: 생성형 AI의 획기적인 해 by 맥킨지 (mckinsey.com)
   • A Bicycle for the (AI) Mind: GPT-4 + Tools
   • Getting good results by filtering some public datasets. You'll find lots of duplicates. Filter by instruction similarity score > .95 (cosine) using e5-large-v2.
   • A Novel Approach for Anomaly Detection Using Large Language Models
   • NeMo Guardrails is an open-source toolkit for easily adding programmable guardrails to LLM-based conversational systems.
   • ChainForge - 프롬프트 엔지니어링을 위한 비쥬얼 프로그래밍 도구 오픈소스 (chainforge.ai)
   • AWS 생성 AI 플랫폼 '세이지메이커'와 '베드락' 차이는
   • CTranslate2 is a C++ and Python library for efficient inference with Transformer models.
   • Custom instructions for ChatGPT
   • ChatGPT Custom Instructions
   • Advances in document understanding (by Google) - Visually Rich Document Understanding (VRDU) dataset
   • Stability AI launches StableCode, an LLM for code generation
   • S프롬프트 엔지니어링으로 재탄생하는 프로그래밍 문화
   • 조직의 생성형 AI 성패, CIO 어깨에 달렸다
   • gpt-llm-trainer - Simply input a description of your task, and the system will generate a dataset from scratch, parse it into the right format, and fine-tune a LLaMA 2 model for you.
   • 구글, 웹 브라우저 기반 IDE ‘프로젝트 IDX’ 공개··· AI 코딩 기능 지원
   • "AI 기반 빙 & 에지 출시 6개월··· 채팅 10억 건, 이미지 7.5억 개 이상 생성"
   • How to fine-tune Llama 2 without writing a single line of code.
   • Google PaLM & TPU 개발자들이 MatX 라는 새로운 칩 회사를 설립 (matx.com)
   • 5 Surprising Stats About the State of AI in 2023
   • Show GN: 프롬프트 엔지니어링으로 수능 국어 1등급에 도전하는 오픈소스 프로젝트 (github.com/NomaDamas)
   • 에이수스, 엔비디아 젯슨 오린 플랫폼 기반 소형 AI컴퓨터 출시
   • Ollama allows you to run open-source large language models, such as Llama 2, locally.
   • Stanford's Natural Language Processing with Deep Learning covers everything from Word2Vec to RLHF!
   • Statistics 110: Probability by Harvard University.
   • Stable Diffusion WebUI is now on #OpenXLab with #SDXL 🥳 Thanks to #OpenXLab for the A100 GPU! 🔥
   • Alibaba, 오픈소스 AI 모델 공개 (cnbc.com)
   • 오픈소스 언어 모델의 현재 (twitter.com/Yampeleg)
   • Fine-Tune LLaMA 2 with QLoRA
   • Stable Diffusion for Audio is here 🤯
   • Do Machine Learning Models Memorize or Generalize?
   • Recommendation Engine: What It Is, How It Works
   • 🎓 ML Courses (11K ⭐️)
   • Text Split Explorer
   • Text Splitter Playground
   • Entity Metadata Extraction
   • llama2.c for Dummies (초보자를 위한 llama2.c 가이드) (github.com/RahulSChand)
   • Llama 2 Uncensored 버전을 로컬에서 실행하기 (ollama.ai)
   • 소프트뱅크, 일본어 특화 AI 모델 만든다··· 오픈AI 경쟁 기관 ‘SB 인튜이션’ 설립
   • "AI가 화력발전소 3개 전력 소비" 고성능일수록 전력 소비도 급증
   • Supercharging AI/ML Development with JupyterLab and Docker
   • 팔란티어 "전례없는 AI 수요 목격"…매출 전망 상향, 주가 2% 상승
   • INT8 Quantization for x86 CPU in PyTorch
   • The History of Open-Source LLMs: Early Days (Part One)(다른 내용 보기: I’m a researcher with an interest in deep learning and a passion for explaining scientific concepts to others.
   • Tensor is a fundamental data structure in Machine Learning. I will clearly explain it today! 🚀
   • llama2 8/7/23 Updates
   • *Vector databases and why they matter in the LLM and Gen AI world*
   • Evaluating LLMs as Agents
   • GN⁺: GPTBot - OpenAI의 웹 크롤러 (platform.openai.com)
   • vLLM & large models - Using tensor parallelism w/ vLLM & Modal to run Llama 70b
   • 생성형 AI 키우는 삼성SDS...AI+클라우드로 기업 AI 시장 잡는다
   • 진료 기록 써주고 환자 데이터 요약…빅테크 3사, '의료용 AI' 출사표
   • 클라우드 빅3, 생성 AI에 언제 웃을까 - AWS, MS, 구글클라우드 등 분기 실적 비교
   • Data engineering failure — Why is it almost impossible to meet deadlines?
   • Functionary is a language model that can interpret and execute functions/plugins.
   • Routers are modules that take in a user query and a set of “choices” (defined by metadata), and returns one or more selected choices.
   • Deploy models painlessly
   • Stanford University is offering the Large Language Models course for FREE!(CS324 - Large Language Models)
   • Key-Locked Rank One Editing for Text-to-Image Personalization
   • Large Language Models Explained - At a High Level
   • Large language models, explained with a minimum of math and jargon
   • Leveraging Machine Learning for Effective Marketing Strategy Development
   • This repository contains demos I made with the Transformers library by 🤗 HuggingFace. Currently, all of them are implemented in PyTorch.
   • Segment Anything as a Service
   • The first iterations of ControlNet for SDXL are hitting HuggingFace 🧙‍♂️
   • K-nearest Neighbors in Scikit-learn
   • NASA-IBM, 기후 변화 연구 위한 LLM 모델 오픈소스로 공개
   • 메타, 오픈소스 AI '오디오 크래프트' 출시··· "텍스트 입력만으로 음향·음악 생성"
   • Function calling in Llama
   • “생성형 AI만 있으면 나도 영화 감독?” 런웨이 젠2 활용한 SF 단편영화 제작기
   • 파워 앱스ㆍ파워 오토메이트에서 로우코드 AI 코딩하기
   • Nvidia H100 GPUs: Supply and Demand
   • Neo4j Graph Store
   • Do Multilingual Language Models Think Better in English?
   • CoreWeave raises $2.3 billion in debt collateralized by Nvidia chips
   • On-disk HNSW index for Postgres with pg_embedding
   • Wow, pushing a 7b model to almost 50% accuracy on GSM8k, approaching code-davinci-002, this is significant!! Half a year ago, the best score I could get was only 27% with FlanT5 11B. Science moves really fast.
   • PubMedQA - A Dataset for Biomedical Research Question Answering
   • Google is about to receive the biggest update in its history. Artificial intelligence will be integrated directly into the search engine.
   • AI NPC 생성 기술 인월드AI, 5천만 달러 투자 유치··· 삼성·LG도 참여
   • ‘구글 어시스턴트’의 미래는 어떻게 될까?
   • 엔비디아 덕분에 신데렐라가 된 'GPU 클라우드' 업체
   • 카카오, 2분기 영업이익 34%↓"10월 초거대 AI 모델 공개"
   • ◐ GPT-Migrate ◑ - Easily migrate your codebase from one framework or language to another.
   • Securing LLM Systems Against Prompt Injection
   • Using AI to Build Stronger Connections with Customers
   • @arp_ai - One of the best channels for NLP! 🙏
   • Revisiting DETR Pre-training for Object Detection
   • 🧺 RAGstack - Deploy a private ChatGPT alternative hosted within your VPC.
   • Announcing SDXL 1.0 by stability.ai
   • 사전 훈련 없이도 작업 척척··· 딥마인드, 로봇 위한 ‘액션’ 모델 RT-2 공개
   • IT 리더가 검토해야 하는 생성형 AI 쟁점 20가지
   • 맥킨지 “기업 22% 생성형 AI 활용··· 기술, 금융, 제약 분야 등에 도입 증가”
   • Introducing LeMUR, the easiest way to build LLM apps on spoken data. Search, summarize, ask questions, and generate new text, with knowledge of all your application’s spoken data.
   • 서비스나우, 생성형 AI 기반 사례 요약 및 코드 생성 기능 발표
   • 더 강력한 생성형 AI 규제가 필요한 이유 “결국 업체가 아닌 기업이 책임지기 때문”
   • Med-Flamingo: a Multimodal Medical Few-shot Learner
   • PromptTools - 🔧 Test and experiment with prompts, LLMs, and vector databases. 🔨
   • GitHub CEO: AI and software development are now inextricably linked
   • FLO가 비슷한 음악을 찾는 방법
   • FalconLite is a quantized version of the Falcon 40B SFT OASST-TOP1 model, capable of processing long (i.e. 11K tokens) input sequences while consuming 4x less GPU memory.
   • Awesome MLOps Awesome - A curated list of awesome MLOps tools.
   • The state of AI in 2023: Generative AI’s breakout year
   • Chapyter is a JupyterLab extension that seamlessly connects GPT-4 to your coding environment.
   • Patterns for Building LLM-based Systems & Products
   • 🔉 LP-MusicCaps: LLM-Based Pseudo Music Captioning(Colab 실행)
   • Music To Image - Sends an audio into LP-Music-Caps to generate a audio caption which is then translated to an illustrative image description with Llama2, and finally run through Stable Diffusion XL to generate an image from the audio !
   • Researchers Publish Attack Algorithm for ChatGPT and Other LLMs
   • 7 Frameworks for Serving LLMs
   • Introducing CM3leon, a more efficient, state-of-the-art generative model for text and images
   • KoBBQ: Korean Bias Benchmark for Question Answering
   • My AI work is available here: Everything I know about AI is in this file.
   • The team is working on a Llama 2 variant of Giraffe and plans to release the weights for that one as well.
   • LLM Reasoners is a library to enable LLMs to conduct complex reasoning, with advanced reasoning algorithms.
   • Open Problems and Fundamental Limitations of Reinforcement Learning from Human Feedback
   • datagran - No-code components + a powerful IDE + a smart AI assistant.
   • Update on the NN+Gzip front.
   • 🚀 Are you interested in learning more about machine learning and its applications? Do you want to watch some awesome YouTube channels that cover topics such as deep learning, natural language processing, neural networks, and more?
   • Run Llama 2 on your own Mac using LLM and Homebrew
   • Treating Attention Deficit Disorder in LLMs
   • Optimizing latency - An exploration of ways to optimize on latency.
   • "기업 55%, 새 애플리케이션 개발에 AI 우선 전략 채택" 가트너 AI 설문조사
   • calamanCy is a Tagalog natural language preprocessing framework made with spaCy.
   • XML Agent
   • UnIVAL: Unified Model for Image, Video, Audio and Language Tasks
   • Building a center of excellence for data science: Five pillars for success
   • Med-Flamingo is a medical vision-language model with multimodal in-context learning abilities. This model is based on the OpenFlamingo-9B V1 model which uses the CLIP ViT-L/14 vision encoder and the Llama-7B language model as frozen backbones.
   • LLM-Rec: Personalized Recommendation via Prompting Large Language Models
   • 🛠️ToolBench🤖 - 🔨This project (ToolLLM) aims to construct open-source, large-scale, high-quality instruction tuning SFT data to facilitate the construction of powerful LLMs with general tool-use capability.
   • ToolLLM: Facilitating Large Language Models to Master 16000+ Real-world APIs
   • 미국 패스트푸드 체인 AI 도입 열풍.. 드라이브 스루에 속속 도입
   • 🚀LLaMA2-Accessory is an open-source toolkit for pre-training, fine-tuning and deployment of Large Language Models (LLMs) and mutlimodal LLMs.
   • 실무에 생성형 AI 어떻게 활용할까..AWS, 무료 및 저비용 교육 과정
   • 7 free and low-cost AWS courses that can help you use generative AI
   • 음성 비서와 생성AI 통합 급물살...아마존·구글 행보 구체화
   • Practical AI for Instructors and Students Part 1: Introduction to AI for Teachers and Students
   • llama2.rs - This is a one-file Rust implementation of Llama2.
   • Llama 2: Open Foundation and Fine-Tuned Chat Models
   • Gorilla: Large Language Model Connected with Massive APIs
   • With @llama_index data agents + text-to-image, we can augment prompt w/ relevant context from a knowledge base! 🔎
   • Handling big models for inference
   • Truss - The simplest way to serve AI/ML models in production
   • Python quant code from Goldman Sachs.
   • 中 “AI 시장 잡자”…알리바바가 선택한 승부수는?
   • 네이버 초대규모 AI, 커머스·모빌리티·금융·교육으로 확산 - 쏘카·SK C&C·한글과컴퓨터 등 하이퍼클로바X 협업 기업 줄이어
   • AWS, 생성AI 모델 포트폴리오 늘린다...코히어와도 제휴
   • So you want to build your own open source ChatGPT-style chatbot…
   • Financial Applications of Machine Learning
   • PeerDB is a Postgres-first data-movement platform that makes moving data in and out of Postgres fast and simple. It enables you to sync, transform and query data across your stores using simple SQL commands.
   • 'AI 신뢰성 인증제도' 나온다
   • Reasoning or Reciting? Exploring the Capabilities and Limitations of Language Models Through Counterfactual Tasks
   • But what are PyTorch DataLoaders really?
   • Here's how to generate your PowerPoint slides using AI (It's 100% free) 👇
   • Large Language Models and Nearest Neighbors
   • DeepMind just announced Med-PaLM M, a Multimodal Generative AI
   • What AI can do with a toolbox... Getting started with Code Interpreter
   • 압도적으로 혁신적인가? GPT-4의 비밀-2
   • RT-2: New model translates vision and language into action
   • Big Data Isn’t Better Data: What’s Wrong with Analytics
   • LLM Attacks
   • roop for StableDiffusion - This is an extension for StableDiffusion's AUTOMATIC1111 web-ui that allows face-replacement in images.
   • Web Explorer - This is a lightweight app using the Web Research Retriever.
   • 5 ways to Increase Statistical Power - Statistical Power in A/B testing visualized
   • FacTool: Factuality Detection in Generative AI
   • Tracking Anything in High Quality
   • Berkeley Open-Sources AI Image-Editing Model InstructPix2Pix
   • edge-tts is a Python module that allows you to use Microsoft Edge's online text-to-speech service from within your Python code or using the provided edge-tts or edge-playback command.
   • ML app with streamlit in super short steps
   • SAM ALTMAN SAYS SORRY, AI IS DEFINITELY DESTROYING JOBS
   • Vector Store Options & Feature Support
   • 7 Ways to Monitor Large Language Model Behavior - Seven ways to track the evolution of LLMs with LangKit and WhyLabs
   • LLMs and the Emerging ML Tech Stack
   • Hippocratic AI is the new state of the art (SOTA) model, outperforming GPT-4 on 105 of 114 healthcare exams and certifications.
2. 하드웨어
   • Intel's Downfall Mitigations Drop Performance Up to 39%, Tests Show
   • 열흘 동안 질화갈륨 고속 충전기 7개 산 사람의 이유 있는 변명
   • GN⁺: Cloudflare 가 제공하는 인터넷 속도 테스트 (speed.cloudflare.com)
   • 아마존이 전세계 Arm 서버 CPU의 절반을 보유중 (theregister.com)
   • “조립 좀 해 본” PC 애호가를 위한 필수 툴 7가지
   • 애플이 새 아이폰용 칩을 위해 수십억달러를 절약하는 방법 (theinformation.com)
   • “SSD 병목의 궁극적 해결”⋯‘광자’ PCIe 새 규격 만든다
   • [비행소년] 비행기를 추적해 보자 #1
   • 프레임워크 랩톱 13 리뷰 | ‘누구나’ 수리, 업그레이드해 오래 쓸 수 있는 노트북
   • “메테오 레이크로 ‘AI PC’시대 열 것” 인텔 CEO 팻 겔싱어
3. 읽을거리
   • 부자 미국·가난한 유럽, 격차 더 커진다
   • 이래도 게임 탓, 저래도 게임중독... 미디어 속 게임 수난사
   • 가계부 템플릿(2023년)
   • 미 연준 역사 한 방에 정리 (WSJ)
   • Hubble Space Telescope
   • 트위터 "이용자 연평균 소득 5천220만원…타 SNS·동영상보다↑"
   • 거주자의 종합소득에 대한 소득세는 해당 연도의 종합소득과세표준에 다음의 세율을 적용하여 계산한 금액(이하 “종합소득산출세액”이라 한다)을 그 세액으로 한다.
   • How Well Can You Hear Audio Quality?
   • 한달에 4천을 벌어도 인생은 안 바뀌더군요.JPG
   • 얼렁뚱땅 잼버리 메타버스, 국민 여러분의 세금이 '터지지 않고' 있습니다
   • 나이 들수록 불행한 한국인…"월소득 500만원 넘으면 더 행복"
   • 돈이 많다고 행복한 건 아니다
   • NASA Plus is the latest streaming competitor

EOB