I recently met with some brilliant Russian mathematicians who showed me a way for artificial intelligence models to communicate via something akin to machine telepathy. The startup Mostik, named after the Russian word for bridge, uses mathematical values from their weights to allow different models to interact. This approach has led to a model that has surpassed expectations in a difficult competition and created a hybrid system that costs one-twentieth of the full GLM model. As Karl Tuyls, a former computer scientist at Google DeepMind, points out, this method can provide substantial improvements with just a smaller model running alongside.

The Mostik team’s approach could increase the value of open-weight models, allowing them to better compete with the closed, proprietary models offered by frontier labs like Anthropic and OpenAI. Smirnov, the chief scientist, says that combining lots of different models may turn out to be a better way to advance AI. As Malysheva, the CEO, says, 'I personally do not think we will have a monolithic model or that the capabilities of models will come from scaling.'

Malysheva’s journey from being told she couldn’t solve Math Olympiad problems to becoming the CEO of Mostik, and her determination to prove those who doubted her wrong, is a testament to her resilience. This innovative approach could revolutionize how we think about AI and its interactions, potentially even revealing new insights into the workings of the human brain.