During a long session of face to face meetings, more like a working session for deployment automation or how I used to joke about it, during a mad ‘rocket scientist’ meeting (off course exaggerating a lot the complexity of the problems we were trying to address), I couldn’t get this  out of my head:

We’re having pretty often difficulties in understanding messages that other humans transmit. Yes this could happen because of language barriers, different level of experience or completely different experiences, personal egos or even the amount of attention each and every person gives to the conversation. Imagine how difficult it will be to try to understand AI talking to another AI in their own developed language.

Machines talking to other machines can be understood if you know the standard they are using and the way the code and the infrastructure that executes the communication works. But AI to AI, what if they create their own standard, a different one for each communication, what if they constantly change the code that communicates to the other AI, how humans will be able to keep up with their changes or even understand them in the beginning?