【Zhiyuan AI Daily】Speed reading for 5 minutes a day, AI important things not to be missed! Welcome to click here to follow and subscribe to Zhiyuan AI Daily.
Inventory | The most comprehensive introduction to Stable Diffusion, the most powerful AI art generation model at the moment
: Programmer and cartoonist Xi Qiao, including: SD model effect display, registration method, parameter setting, charging standards, copyright issues, improvement methods and Introduction of SD R&D team, technology, highlights and related resources.
Frontiers | Richard Sutton, a pioneer in reinforcement learning: will develop new types of computational agents
Aspect: The three leaders of reinforcement learning jointly elaborated on the Alberta Plan, which is dedicated to understanding and creating long-term interactions with the complex world and predicting and controlling its sensory input signals Lifetime calculation agent.
Information | Google proposes OptFormer: Use Transformer to predict the best hyperparameters, and you can also know the performance
points in advance: it is one of the first Transformer-based hyperparameter tuning frameworks, which uses flexible text-based representations from large-scale optimization Learn from data.
Inventory | Highlights of the domestic high-level talent data guide in the field of artificial intelligence
: In the list of AI 2000 scholars, the number of American scholars ranks first, accounting for 57.3%, and the number of Chinese scholars ranks second, accounting for 11.6%; Tsinghua University, Hong Kong Chinese Universities, Peking University and other colleges and universities, as well as Alibaba, Tencent, and JD.com rank among the top in the number of talents, and cities and regions such as Beijing, Hong Kong, Macao, Taiwan, and Hangzhou rank high.
Awards | Academician Tan Tieniu won the highest award in international pattern recognition
Highlights: This is the first time since the award was established in 1988 that it has been awarded to scholars outside North America and Europe.
News| Google Blog: Using Machine Learning to Boost Participation in India's Maternal and Child Health Program
What to watch: This program uses historical data from NGOs to predict which beneficiaries will benefit the most.
Information | DeepMind proposes a causal influence diagram: inferring the agent's motivation to make it consistent with the designer's requirements Aspect
: The causal influence diagram (CID) is a method for modeling decision-making situations that allows us to model the agent's intent to infer.
Information | OpenAI released a collection of Alignment research work: three pillars to promote AGI alignment research
highlights: three pillars: using human feedback to train AI systems, training artificial intelligence systems to assist human evaluation, training artificial intelligence systems for alignment research
Interpretation | Eliminate artifacts, use Lipschitz constraints to stabilize the reasoning
point of "loop video super-resolution": Video super-resolution, Video super-resolution, that is, recovering high-quality video sequences from low-resolution videos, is better than images Super-resolution is a more difficult task.
Download | The latest "Graph Neural Network" report, University of Trento, Italy 29-page ppt modeling explains GNN and GAT and other
highlights: discusses the application of graph neural networks in various fields, and summarizes its open source code and benchmark data sets and model evaluation.
Summary | AI-6G and Metaverse? The latest review of "Artificial Intelligence and 6G and Metaverse: Fundamental Principles, Challenges and Future Research Trends" by Dongli Ireland, the 39-page pdf covers 263 documents Highlights
: This article first introduces augmented reality (AR), virtual reality (VR), The background and progress of mixed reality (MR) and spatial computing, and then introduce the role of AI and 6G technology in the Metaverse, related use cases and projects, etc.
