AIGC周刊 | 第25期
时间： 2023.4.24-2023.5.1
本周大事记
1. AudioGPT
全新的音频理解与生成系统AudioGPT
AudioGPT: Understanding and Generating Speech, Music, Sound, and Talking Head
论文: arxiv.org
2. Stability AI 发布 DeepFloyd IF
Stability AI发布全新的文生图扩散模型
主页：deepfloyd.ai
代码：github.com
3. Huggingface发布HuggingChat
huggingface发布对标ChatGPT的chat产品HuggingChat
最新技术：
ReCEval: Evaluating Reasoning Chains via Correctness and Informativeness
论文：arxiv.org
TextDeformer: Geometry Manipulation using Text Guidance
论文: arxiv.org
Patch-based 3D Natural Scene Generation from a Single Example
论文: arxiv.org
主页: weiyuli.xyz
Answering Questions by Meta-Reasoning over Multiple Chains of Thought
论文: arxiv.org
Stable and low-precision training for large-scale vision-language models
论文: arxiv.org
Patch Diffusion: Faster and More Data-Efficient Training of Diffusion Models
论文: arxiv.org
BadGPT: Exploring Security Vulnerabilities of ChatGPT via Backdoor Attacks to InstructGPT
论文: arxiv.org
Enhancing Chain-of-Thoughts Prompting with Iterative Bootstrapping in Large Language Models
论文: arxiv.org
WizardLM: Empowering Large Language Models to Follow Complex Instructions
论文: arxiv.org
Score-Based Diffusion Models as Principled Priors for Inverse Imaging
论文: arxiv.org
Are Emergent Abilities of Large Language Models a Mirage?
论文: arxiv.org
A Cookbook of Self-Supervised Learning
论文：arxiv.org
Speed Is All You Need: On-Device Acceleration of Large Diffusion Models via GPU-Aware Optimizations
论文：arxiv.org
大语言模型的进化树，这是一份超详细ChatGPT「食用」指南
向量嵌入：AutoGPT的幻觉解法？
课程：
Andrew Ng的deeplearning.ai和OpenAI联合推出的ChatGPT视频课程：ChatGPT Prompt Engineering for Developers
为大语言模型建立红队对抗
商业：
1. AI 浪潮：自主性是人类智能的最后堡垒吗?
2. 周枫：当我们谈论大模型时，应该关注哪些新能力？
3. 到底有没有人在投 AI？AI 赛道第一线实操体感分享
案例：
你画一笔AI画整张图！AI实时作画方式被打开，颜色细节自动补全，一个脚本即可实现
神奇LLM引擎上线：帮你把GPT-3直接调成ChatGPT
我们做了个能对话的AI派蒙，免费给大家玩！