https://github.com/ifoxhz/pigame

If we compress LLMs into neuromorphic chips to reduce inference latency, would it be relatively easy to achieve low-cost autonomous driving?