What You May Have Missed #28
R&D and products (ToT, DragGAN, LIMA, ChatGPT iOS) / Essays (Best: prof falsely accused students of using ChatGPT) / Miscellanea (fake ID hacks, LeCun's snarky takes) / Senate hearing on AI oversight
R&D and products
Tree of Thoughts: Deliberate Problem Solving with Large Language Models (DeepMind and Princeton): “[Tree of Thoughts, an improvement over Chain of Thought] allows LMs to perform deliberate decision making by considering multiple different reasoning paths and self-evaluating choices to decide the next course of action, as well as looking ahead or backtracking when necessary to make global choices.”
Drag Your GAN: Interactive Point-based Manipulation on the Generative Image Manifold: “Through DragGAN, anyone can deform an image with precise control over where pixels go, thus manipulating the pose, shape, expression, and layout of diverse categories such as animals, cars, humans, landscapes, etc … these manipulations … tend to produce realistic outputs even for challenging scenarios such as hallucinating occluded content and deforming shapes that consistently follow the object's rigidity.”
“LIMA: Less Is More for Alignment” (Meta): “In a controlled human study, responses from LIMA are either equivalent or strictly preferred to GPT-4 in 43% of cases; this statistic is as high as 58% when compared to Bard and 65% versus DaVinci003, which was trained with human feedback.”
Introducing the ChatGPT app for iOS (OpenAI blog). “The ChatGPT app is free to use and syncs your history across devices. It also integrates Whisper, our open-source speech-recognition system, enabling voice input. ChatGPT Plus subscribers get exclusive access to GPT-4’s capabilities, early access to features and faster response times, all on iOS … We're starting our rollout in the US and will expand to additional countries in the coming weeks.” What’s going to happen with all the existing non-official ChatGPT apps?
LLMs can be used to effectively scale spear phishing campaigns. Ethan Mollick: “Cybersecurity is about to get rough. As suspected, AI is great at creating spearphishing attacks, emails that feel real but compromise security. And they can do it at scale, where previously it took a lot of effort. Open source models seem capable, too.”
Ultimate Guide: 86 ChatGPT Plugins (and the prompts to use with them).