BabyLM Challenge, ChatGPT Limitations, Death of Self-driving Cars, and the Robot Revolution
AI Bytes: Volume CI, Issue #22
Get smarter faster with AI Bytes: 3 articles, 2 podcasts, and 1 video.
Was this email forwarded to you? Sign up here.
READING
III. The Race to Make A.I. Smaller (and Smarter)
A group of young academics working in natural language processing issued a challenge to create functional language models using data sets that are less than one-ten-thousandth the size of those used by the most advanced large language models. The BabyLM Challenge aims to create smaller, more accessible, and more compatible language models that are nearly as capable as high-end models. The challenge is a step toward more accessible, more intuitive AI.
II. OpenAI isn’t doing enough to make ChatGPT’s limitations clear
OpenAI's chatbot, ChatGPT, has often misled users by generating false information, leading to serious consequences. Users deserve some blame, but OpenAI needs to do more to make it clear that ChatGPT cannot reliably distinguish fact from fiction. A simple intervention could be for ChatGPT to recognize when it's being asked to generate factual citations and caution the user to "check my sources."
I. The death of self-driving cars is greatly exaggerated
Despite the failure of many self-driving car projects, Waymo and Cruise are still working on the technology and have already begun testing it in Phoenix and San Francisco. They plan to expand to other cities and increase their passenger rides in the coming years, though profitability remains a challenge. Industry analysts believe that self-driving technology will continue to improve and expand, but progress will be slower than originally anticipated. It usually is.
LISTENING
II. The Robot Revolution | Keerthana Gopalakrishnan of Google Robotics on The Cognitive Revolution
In this podcast episode, Nathan Labenz interviews Keerthana Gopalakrishnan, a Google Robotics researcher, about training robots through supervised learning, integrating language models for commands, and robotics safety. They also discuss the potential future of robotics and AI, including using robots to better understand ourselves and the impact of AI on society.
I. Kevin Kelly on Advice, AI, and Technology
The argument is that while there is a possibility of AI evolving beyond our control, it is not worth worrying about in terms of day-to-day policy. The idea of a superhuman AI is based on belief rather than evidence, and there is no evidence that intelligence is a single dimensional or that it could expand exponentially. While some people should be dedicated to ensuring AI safety, it should not be a significant influence in our society's decision-making at this time.
WATCHING
NVIDIA'S HUGE AI Breakthroughs Just Changed Everything (Supercut)