Last Week in AI OpenAI’s InstructGPT, Meta’s New AI Supercomputer, China Regulates DeepFakes, AI Playes Tetris
Feb 6, 2022
Discover the latest AI advancements, including Meta's ambitious goal to create the fastest supercomputer. OpenAI unveils new, safer text-generating models. China is stepping up its regulations on deepfakes, raising ethical questions. Learn how AI is revolutionizing mental health monitoring. Plus, see an AI play an incredible game of Tetris that showcases advanced strategies. Google highlights women's contributions to science, aiming to reshape visibility in the field.
AI Snips
Chapters
Transcript
Episode notes
AI Advancements
- Meta is building the world's fastest AI supercomputer, aiming for trillion-parameter neural networks.
- OpenAI's InstructGPT generates less toxic text, while a new Meta AI learns from multiple modalities.
Meta's Supercomputer Specs
- Meta's supercomputer uses 16,000 NVIDIA A100 GPUs and will process exabytes of data.
- Andrei mentions a Google anecdote where employees can freely use 100 GPUs.
ML for Mental Health
- MIT and Harvard are using machine learning to improve mental health, with a 5-year study using wristbands.
- This aims to quantitatively track and predict mental health symptoms, similar to fitness trackers.