This week's edition explores AI's role in healthcare prediction, the science of effective human-AI teamwork, new agent capabilities from OpenAI, and creative technology for family connection.
🧠 New AI Tool Predicts Cognitive Decline Years Before Symptoms Appear
The technology: Sleep monitoring has long been used to assess health conditions, but researchers at Mass General Brigham have taken this concept further. Using electroencephalography (EEG) - a non-invasive technique that measures electrical activity in the brain through scalp sensors - their AI system can detect subtle patterns during sleep that predict future cognitive decline.
The research findings: In a groundbreaking study published in the Journal of Alzheimer's Disease, researchers tracked 281 cognitively normal women for five years. The AI system analyzed their sleep EEG data and successfully identified 85% of individuals who later developed cognitive impairment, achieving an overall accuracy of 77%. The technology detected changes in gamma band frequencies during deep sleep that signal future cognitive decline, even when no symptoms were present.
Why it matters: This innovation creates a potential five-year window for intervention before cognitive symptoms appear. For individuals, this means valuable time to implement lifestyle changes that may preserve brain health - like regular exercise, cognitive stimulation, and dietary improvements. The technology could transform our approach to dementia prevention by shifting focus to the presymptomatic stage, when interventions might be most effective. Researchers are now designing clinical studies to test whether electrical stimulation can alter these EEG patterns during sleep, potentially reducing cognitive decline risk.
Citation: Haghayegh S, Herzog R, Bennett DA, et al. Predicting future risk of developing cognitive impairment using ambulatory sleep EEG: Integrating univariate analysis and multivariate information theory approach. Journal of Alzheimer’s Disease. 2025;0(0). https://doi.org/10.1177/13872877251319742
🤝 MIT Research Reveals When Humans and AI Work Best Together
The research: The MIT Center for Collective Intelligence studies how humans and technologies can work together more effectively. Their researchers recently published findings in Nature Human Behaviour that analyze when humans and AI perform better together versus when either performs better alone.
Key discoveries: Contrary to popular belief, humans and AI working together don't automatically outperform either working independently. The meta-analysis of 370 experiments revealed that human-AI collaboration works best in specific scenarios: when humans already outperform AI at a task, when creating content (rather than making decisions), and especially when using generative AI for creative work. For example, in bird image classification, humans achieved 81% accuracy and AI 73%, but together they reached 90% accuracy.
Why it matters: For your everyday AI interactions, understanding these patterns helps you make better choices about when to rely on AI assistance versus your own judgment. When working with tools like ChatGPT or Midjourney, you'll likely get the best results on creative projects where you provide direction and refinement while the AI handles initial generation and variations. However, for decision-making tasks where the AI consistently outperforms humans (like detecting fake reviews), you might be better off trusting the AI's recommendation rather than second-guessing it. The key is recognizing your strengths (contextual understanding, emotional intelligence) and the AI's advantages (processing large amounts of data, generating variations quickly) to create effective collaboration.
Citation: Vaccaro, M., Almaatouq, A. & Malone, T. When combinations of humans and AI are useful: A systematic review and meta-analysis. Nat Hum Behav8, 2293–2303 (2024). https://doi.org/10.1038/s41562-024-02024-1
More AI everywhere doesn’t mean better results. Small teams of AI experts beat mass licenses. Attackers now use AI to run cybercrime at machine speed. Leaders can’t keep up. Tools still forget us. The winners? Focused expertise and hybrid AI, not AI everywhere.
You've built something and you need to know if it works. So you do what's sensible—you ask an LLM to grade it. Factual accuracy, code quality, agent outputs. The machine judges the machine, and you get a number you can act on. Except that number
Imagine you have a bunch of teams, some with AI, some without, and some where everyone gets their own AI. Researchers ran a big experiment with over 400 people to see what actually happens when you mix and match humans and AI in different ways.
Here’s what they found:
