The FAIK Files The Trough of Disillusionment
33 snips
Oct 24, 2025 Google's Gemini 3.0 is making waves with its latest AI updates. A revealing study shows AI assistants misrepresent news content nearly half the time. Discussions on AI's impact on job loss include alarming leaks from Amazon. Fascinating deepfake incidents highlight the techniques used in political manipulation and scams. From voice cloning tricks to well-crafted disinformation, the hosts dissect how deepfakes are reshaping trust in media and communications.
AI Snips
Chapters
Books
Transcript
Episode notes
Agentic AI Aims To Control Interfaces
- Google is pushing multimodal and agentic AI that can interact with user interfaces, not just process text and images.
- These "computer use" models aim to automate tasks like shopping and itinerary planning by visually analyzing pages and controlling UI elements.
Agentic Models Expand Attack Surface
- Agentic capabilities raise new attack surfaces like prompt injection, where malicious context causes harmful actions.
- Early implementations perform unevenly and expose rough edges attackers can exploit.
Early AI Flaws Don't Predict Trajectory
- New AI features often launch publicly in rough form but improve rapidly through iteration and adoption.
- Early failures shouldn't blind us to significant future potential and fast progress.
