He proposed the Forward-Forward Algorithm and has been toying with it eversince.
While Automatski eliminated Backtracking from its deep learning platform 2 decades ago. We just use forward passes and some adjustments to train the models.
Consequently we don't need any GPUs pretty much. 99% lower energy. 95% lower compute costs. 100X larger models...
Beat that OpenAI, Microsoft, Google, Nvidia, Intel, AMD !!!
We created the most Universal Model for Quantum Deep Learning & Quantum State Preparation using a Universal Ansatz Circuit & Simulator/Computer Instructions.
This delivers Polynomial if not Exponential Quantum Advantage. Which is otherwise not possible.