A learning path through language models, retrieval, evaluation, and the practice of research.
How do you move from using language models to understanding and building them? These notes connect technical foundations with implementation, experiments, and the judgment needed to ask useful research questions.
The knowledge map behind my preparation for an OpenAI research engineering role: seven connected areas, concrete failure cases, diagnostic exercises, and a method for turning study into usable skills.
How BM25, dense retrievers, SPLADE, and ColBERT represent and match evidence. Worked examples of MaxSim, compression, hybrid fusion, and the difference between scoring quality and retrieval quality.
ethical and reliable evaluations of llm is critical
从 Facebook 早期校园网络、增长团队、S-1 和 Alex Schultz/Andy Johns 等材料出发，拆解可迁移的增长原则：留存优先、magic moment、边际用户、增长系统，以及它们对 ChatGPT 类产品的启发。
what would happen if AI does all the jobs for us as human?
what does the people working in Silicon Valley say about LLM?
Retail + ML = New Commerce
You should at least think this through when considering the use of LLM in RecSys
2 kings in the era of large language model; one is close source, the other is open source
Is the giant Google sleeping?
Will OpenAI create the next-gen cloud computing platform
AIGD is what we need.
search and learning are the two most scalable and future-ready algorithms
The groundbreaking study on ChatGPT reveals a systematic left-leaning political bias, raising critical questions about AI's influence on public opinion and democracy, and calling for collaboration between technology leaders, lawmakers, and society to uphold human dignity in the algorithmic age. 🤖💬🌍
Explore the evolution of Large Language Models (LLMs) and the shift from GPUs to alternative hardware. Dive into the world of FPGAs, ASICs, and optical computing, and discover what the future holds for AI technology.
RAVEN, a groundbreaking AI model, utilizes retrieval-augmented mechanisms to excel in in-context learning, outperforming the ATLAS model and opening new possibilities in AI. 🚀
finetuned llama2 means a lot of things
why open source community matters
a fine-tuned llama2 can help answer some of your medical questions
a new approach to train language models
go cheaper, go cheapest
math tools enhance GPT4 capabilities
google found that synthetic-data intervention to reduce models' tendency to agree with users' incorrect opinions
when database meets LLM Agent
if you need an open agi projects, take a look at here
ChatGPT vs. Stack Overflow study reveals 52% of ChatGPT's answers are incorrect, yet preferred 39% of the time for their comprehensive language.
A universal language for state-of-the-art audio synthesis
a multi-agent assembly line approach, mirroring human workflows for complex problem-solving
you need personalization, simplicity, integrity and extensibility
llm agents are used to simulate societal-economical events
how to generate ML/AI ideas in your job and manager afterwards
what careers are you looking for in the ML/AI community
intro to ML/AI zoo
how is it different to build LLM for non-english languages
Navigating The Core Trio: Search, Recommendation, and Advertising Algorithms in the Digital Realm
what does the role of machine learning manager at FAANG look like?
If a system returns what you need, does it matter how it was produced?
How can we find the intersections between information generation, distribution and consumption?
human are usually wrong, especially on the prediction of future.
This blog post provides a comprehensive exploration of Altman's proposal, drawing from various disciplines and historical lessons. It's a complex and multifaceted issue that warrants continued discussion and exploration. The future is bright, but the path we choose will determine how we get there.