Diving into the depths of my NCAA Pool with AI.

      Forty years of bracket intuition combined with 50,000 simulated tournaments — and the AI's first insightful question was one I had never considered myself.

      Peter Horan / Nano Banana 2

      We are in a new era. After over four decades of participating in March Madness pools and even clinching a few victories (like NC State in 1984), I collaborated with ChatGPT to help complete my brackets. Like many, I watch some college basketball, though usually only focusing on my three or four favorite teams and their opponents. I catch some significant national games on weekends, but that’s the extent of it.

      Typically, when I look at the vast bracket, I find myself pondering: Where is High Point University and who are they facing? What about the infamous 12/5 upset? Is it too cowardly to pick the #1 seeds to win? This results in a lot of anxiety but offers little insight.

      Peter Horan / Nano Banana 2

      Can a machine surpass four decades of intuition?

      I inquired how it would construct a model to forecast the winner of the NCAA tournament. The mere response made me exclaim “wow.”

      Here’s what it said: “I can certainly assist you in creating a model: starting from a basic seed-and-efficiency model to a more comprehensive probability model utilizing KenPom-style efficiency, pace, injuries, travel, coaching, and upset history. I can also aid with: bracket win probabilities, upset predictions by round, Monte Carlo tournament simulation, Elo or logistic regression models, and spreadsheet-based prediction tools.

      ChatGPT took into account team strength on a neutral court, factoring in offensive and defensive efficiency, strength of schedule, consistency, and team strength, among other variables. This eventually led to a set of predictions derived from this equation:

      Peter Horan / DigitalTrends

      It compares the matchups of two teams by applying these team ratings, utilizing a logistic model.

      Peter Horan / DigitalTrends

      Finally, it simulated the entire tournament using this method, undertaking no more than 50,000 simulations.

      Trust the algorithm or your instincts?

      I don’t know about you, but that operates at a level of calculus and computing power that far exceeds my capabilities. It then returned with an astute nuance, essentially posing the question, “Is it more crucial to be correct or to win your pool?” The distinction is that while picking only favorites will yield a higher correct percentage, it won't distinguish you from other participants in the pool. You need to bet on a few upsets to ensure financial success. I can submit two brackets in my primary pool—the one where poor brackets lead to substantial teasing—so I asked for one of each type.

      I closely adhered to ChatGPT’s suggestions, with the exceptions being giving a nod to my three favorite teams in the initial rounds: Santa Clara, Gonzaga, and UCLA. While I'd love to be mistaken, I don’t anticipate seeing any of them advancing further in the tournament.

      I plan to monitor the effectiveness of ChatGPT’s predictions here as we progress through the tournament. You’ll be the first to learn if ChatGPT makes a fool of me. I’m keeping my fingers crossed.

      Peter C. Horan is an entrepreneur and digital media investor with a background in developing successful media, commerce, and ad ventures.

      The new MacBook Neo charges quicker than Apple claims

      Testing has revealed a surprising discrepancy in charging speed for Apple's newest budget laptop — and the fix costs just $59 at Apple’s own store.

      Apple’s latest MacBook Neo operates on the same chip as the iPhone 16 Pro, yet its charging speed falls short compared to the 2024 flagship iPhone. At least that’s what Apple states in the official specification sheet.

      The MacBook Neo is bundled with a 20W USB-C power adapter, a choice that seems increasingly questionable, especially considering that the device had the most successful Mac launch week ever among first-time buyers.

      Read more

      Adobe will allow you to create custom Firefly AI models based on your own work and artistic style

      Firefly now enables you to generate, compare, and edit images and videos while upholding a consistent visual aesthetic.

      Adobe has launched Custom Models in public beta for Firefly, providing a means to produce images in your unique visual style instead of relying on generic AI outputs.

      By uploading your own images, you can craft a model that reflects your distinctive look, be it illustration, photography, or character design.

      Read more

      AI mental health risks revealed as chatbots sometimes enable harm

      New research indicates that some AI responses may reinforce harmful thoughts rather than intervene.

      A study led by Stanford is raising alarms about AI mental health safety, finding that certain systems can inadvertently promote violent or self-harm ideas instead of averting them. The research draws on real user interactions and highlights shortcomings in how AI addresses crisis situations.

      In a small but high-risk sample of 19 users, researchers scrutinized nearly 400,000 messages