Last Week in AI Mini Episode: TikTok, Cheap Deepfakes, AI in 2020, and Deference
Aug 9, 2020
Discover the intriguing world of TikTok and Microsoft's acquisition strategies. Dive into the growing accessibility of deepfake technology and its societal implications. Explore how AI has struggled to adapt since 2020, highlighting the need for updated data to combat biases. Plus, learn about an MIT study showcasing AI's ability to defer to human judgment in medical decision-making, shedding light on the complexities of human-AI collaboration.
AI Snips
Chapters
Transcript
Episode notes
Microsoft's TikTok Acquisition Strategy
- Microsoft's potential acquisition of TikTok aims to grant access to valuable video data.
- This data can enhance their AI systems, particularly in computer vision.
Deepfakes on a Budget
- Philip Tully used open-source AI software and under $100 to create fake images of Tom Hanks.
- This demonstrates the increasing accessibility and affordability of deepfake technology.
AI's 2020 Vision Problem
- AI models are struggling to adapt to the rapid societal changes of 2020.
- Computer vision models misclassify images of new norms, like working from home.