AI Coder Outshines Humans in Prestigious Programming Contest

In a surprising turn of events, an AI named Gemini 2. 5 has shown its coding prowess by securing a gold medal at the International Collegiate Programming Contest (ICPC) World Finals. This contest is a big deal, attracting thousands of top college coders who tackle tough algorithmic puzzles under time pressure. Gemini 2. 5 didn't just participate; it dominated. It solved 10 out of 12 problems correctly, a feat only four human teams achieved. What's more, it did this using the same model that powers other Gemini applications, just tweaked to handle the competition's demands. The contest is no walk in the park. Teams have just five hours to crack complex problems. Gemini started 10 minutes late but quickly caught up, solving eight problems in just 45 minutes. By the end, it secured a second-place finish overall. One problem, in particular, stood out. Problem C was a multi-dimensional optimization puzzle that stumped every human team. But not Gemini. It approached the problem by assigning priority values to each reservoir, using a dynamic programming algorithm to find the most efficient configuration. After 30 minutes of intense computation, it nailed the solution. Google, the company behind Gemini, believes this performance shows AI's potential in fields like semiconductor engineering and biotechnology. The ability to tackle complex, multi-step problems could make AI models like Gemini invaluable in these industries. But there's a catch. All this computing power comes at a cost. While Google hasn't revealed the exact energy consumption, it's safe to assume it's substantial. Even simpler AI models are expensive to run, but the ability to solve previously unsolvable problems might justify the high cost.