Last Week in AI Making AI Less Racist and Terrible, AI for Wildfires and Reading Lips, Fun AI Facts about Fun Guys
Jun 24, 2021
Discover how AI is being harnessed to spot wildfires faster than humans and the ongoing efforts to make text-based models less biased. The discussion dives into the ethical implications of surveillance technology, especially in workplaces, as well as the quirky world of AI-generated humor. Get ready for fun facts about unusual traits, and explore the interplay of innovation and ethics in artificial intelligence, making for a fascinating blend of serious and lighthearted topics.
AI Snips
Chapters
Transcript
Episode notes
Controllable GANs
- New research aims to improve control over AI image generation using GANs.
- Bounding boxes can specify object placement, leading to more predictable and controllable image synthesis.
Making LLMs Less Toxic
- OpenAI researchers are working to make large language models like GPT-3 less racist and terrible.
- By feeding the model encyclopedia-like samples on sensitive topics, they aim to improve the appropriateness of its output.
AI for Wildfire Detection
- AI is being used to spot wildfires faster than humans by analyzing satellite imagery.
- Sonoma County and New Mexico are testing and deploying these AI-powered wildfire detection systems.