You guys aren't about decomposing them to "smaller" fuzzy compilers? Or replacing parts inside the fuzzy compilers by deterministic bits? Or turning to implict machine learning, where for example ImplicitLayers/NeuralODE's/DEQ's are becoming factors inside a factor graph giving an implicit representation of an LLM?
You guys aren't about decomposing them to "smaller" fuzzy compilers? Or replacing parts inside the fuzzy compilers by deterministic bits? Or turning to implict machine learning, where for example ImplicitLayers/NeuralODE's/DEQ's are becoming factors inside a factor graph giving an implicit representation of an LLM?