Links for 2023-03-24
Sparks of Artificial General Intelligence: Early experiments with GPT-4 — “We demonstrate that, beyond its mastery of language, GPT-4 can solve novel and difficult tasks that span mathematics, coding, vision, medicine, law, psychology and more, without needing any special prompting. Moreover, in all of these tasks, GPT-4's performance is strikingly close to human-level performance, and often vastly surpasses prior models such as ChatGPT. Given the breadth and depth of GPT-4's capabilities, we believe that it could reasonably be viewed as an early (yet still incomplete) version of an artificial general intelligence (AGI) system.” https://arxiv.org/abs/2303.12712
ChatGPT Gets Its “Wolfram Superpowers”! https://writings.stephenwolfram.com/2023/03/chatgpt-gets-its-wolfram-superpowers/
ChatGPT + Code Interpreter = Magic: OpenAI is testing the ability to run code and use third-party plugins in ChatGPT. https://andrewmayneblog.wordpress.com/2023/03/23/chatgpt-code-interpreter-magic/ (“For those who aren't aware: you write an OpenAPI manifest for your API, use human language descriptions for everything, and that's it. You let the model figure out how to auth, chain calls, process data in between, format it for viewing, etc. There's absolutely zero glue code.” https://twitter.com/mitchellh/status/1638967450510458882 / “Plugins for processing a video clip, no ffmpeg wizardry required.” https://twitter.com/gdb/status/1638971232443076609 / “OpenAI is effectively using its ecosystem position to get all major services to expose AI-friendly APIs. An AI-controllable world is coming.” — Nat Friedman)
Opera announced GPT-based AI prompts in the browser. You can now interact with ChatGPT from anywhere to: Shorten or explain articles; Generate tweets and content; Find relevant info based on text. ChatSonic gives users the capability to create AI-generated images. https://blogs.opera.com/desktop/2023/03/opera-ai-tools/
Duolingo launches a language tutor powered by GPT-4 https://twitter.com/duolingo/status/1635688521695633408
Zero-1-to-3: Zero-shot One Image to 3D Object: “We learn to control the camera perspective in large-scale diffusion models, enabling zero-shot novel view synthesis and 3D reconstruction from a single image.” https://zero123.cs.columbia.edu/
Rotating without Seeing: Towards In-hand Dexterity through Touch https://touchdexterity.github.io/
Dexterity from Touch: Self-Supervised Pre-Training of Tactile Representations with Robotic Play https://tactile-dexterity.github.io/
It’s Game Over on Vocal Deepfakes https://daringfireball.net/2023/03/game_over_vocal_deepfakes
How bad a future do ML researchers expect? https://www.lesswrong.com/posts/Zq2HaihaDy7sSarMz/how-bad-a-future-do-ml-researchers-expect
Probability estimates of doom from AI by various people from within or adjacent to the rationalist community https://astralcodexten.substack.com/p/why-i-am-not-as-much-of-a-doomer
Discussion with Nate Soares on a key alignment difficulty https://www.lesswrong.com/posts/iy2o4nQj9DnQD7Yhj/discussion-with-nate-soares-on-a-key-alignment-difficulty
“OK, it’s time to freak out about AI” https://nonzero.substack.com/p/ok-its-time-to-freak-out-about-ai
Robot Learns to Master Task Sequencing with Human Expert Demonstrations https://www.youtube.com/watch?v=jON2AhwBp5M
Researchers Pressured Twitter To Treat COVID-19 Facts as 'Misinformation' https://reason.com/2023/03/17/researchers-pressured-twitter-to-treat-covid-19-facts-as-misinformation/
Revisiting some old tweets:
Technically, we are there, or almost there:
2019:
2020:
2021:
Coming soon:
2023: