Is this really what it has come to? Do we need to use deep reinforcement learning to make IoT devices useful? Should we really have to deal with such poorly documented proprietary garbage?
Reminds me of the Linux origin story with RMS and the printer[1].
Good luck trying to understand my NB-IoT protocol messages then. Raw sensor values, compressed, encrypted, basE91 encoded. No chance for any RL neither AI to understand that. No need to interoperate neither. IoT is a single private sensor connected to a single private receiver. The medium is not the message here.
If I read this correctly, this is a technique for discovery of arbitrary, unknown API. “iot” is too specific. But the former formulation sounds amazing, doesn’t it?