Welcome to The Daily AI Briefing! Hello and welcome to The Daily AI Briefing, where we bring you the most significant developments in artificial intelligence happening right now. I'm your host, and today we have a packed show with groundbreaking innovations, new tools, and important industry movements that are shaping our AI-driven future. Today's Highlights In today's episode, we'll explore an AI system that predicts cancer outcomes from facial photos, Sakana AI's brain-inspired continuous thought machines, and a clever way to mine video content with Google's NotebookLM. We'll also examine OpenAI's new medical benchmark called HealthBench, highlight trending AI tools, and round up the latest industry news including major funding developments. Cancer Prediction from Facial Photography Researchers at Mass General Brigham have developed an intriguing AI system called FaceAge that analyzes facial photographs to estimate biological age and predict cancer survival outcomes. Trained on tens of thousands of facial images, the system translates subtle facial characteristics into biological age estimates. The findings are remarkable – cancer patients appeared approximately five years older on average according to the AI, with higher FaceAge scores correlating with worse survival rates. When physicians added these FaceAge risk scores to their clinical data, they saw significant improvements in predicting 6-month survival rates. What makes this particularly fascinating is that the AI's predictions correlated with genes associated with cellular aging, suggesting FaceAge is capturing biological processes that can't be detected by chronological age alone. Sakana AI's Continuous Thought Machines Moving to innovations in AI architecture, Sakana AI has unveiled what they call Continuous Thought Machines or CTMs. This represents a fundamental shift in how AI systems process information. Unlike conventional models that make instant decisions, CTMs are designed to "think" step-by-step over time, much like human brains. This approach draws inspiration from neuroscience, where the timing of neuron activation is crucial for intelligence. In demonstrations, Sakana showed these CTMs solving complex mazes by visibly tracing possible paths and tackling image recognition by examining different parts of an image – spending more time on areas based on task difficulty. This mimics how humans approach problem-solving more closely than traditional AI systems. Video Content Mining with NotebookLM Content creators will be interested in a new tutorial showing how to leverage Google's NotebookLM to analyze videos and enhance content creation. The process allows users to generate transcripts, title ideas, hooks, and descriptions from video content. The workflow is straightforward: visit NotebookLM, sign in with your Google account, create a new notebook, add videos either via file upload or YouTube connection, and then use prompts to generate transcripts and other content elements. What makes this particularly useful is the ability to upload multiple videos with their performance statistics for comparative analysis, helping creators understand what's working and what isn't in their content strategy. OpenAI's HealthBench Healthcare AI took a step forward with OpenAI's release of HealthBench, a benchmark created in collaboration with 262 physicians to evaluate AI systems' performance in health conversations. This benchmark tests models across various healthcare themes, including emergency referrals and global health issues, while measuring behaviors like accuracy and communication quality. It represents an important effort to establish standards for measuring AI's safety and effectiveness in medical contexts. Recent models have shown remarkable improvement on this benchmark. OpenAI's model designated "o3" scored 60% compared to GPT-3.5 Turbo's 16%. Even more promising is that smaller models are becoming increasingly capable, with GPT-4.1 Nan