Recently there has been a lot of articles on A.I. and dooms day which it brings with itself. There is not much to say, just the fact that it's not happening anytime soon. For all the paranoid people, you can sleep well without worries.
People who do not work on A.I. research can easily believe what some journalists write, but seriously - facebook's bots did not invent their own language, they diverged and crashed - talking complete nonsense.
Alan Turing, smart man and person who decoded war codes (and saved many people) defined first stage of A.I. by passing Turing test:
As you see from conversation Facebook's robots had, it is easy for anyone to tell it's not humans speaking. And yes, machines might eventually develop their own language, but current A.I. is not good enough even when speaking with us - humans. Do you remember when you asked something Siri / Cortana / Alexa / OK Google and received weird answer? Current A.I. is not capable of providing us good answers and developing their own language is light years away from that stage and more advanced. Would you teach a child calculus if they did not start walking yet? No. And same holds with A.I. - don't expect it to invent their own language if they have not mastered speaking one yet.
Anyone claiming A.I. developed their own language, please go ahead and let the A.I. bots speak for a year and see what amazing they come up with. If it is "balls have zero to me to me", the only thing that is missing is American "yo" at the beginning, or Canadian "eh" at the end.
And just for the record, beating humans in chess and Go game is not intelligence, it is just brute force and trying a lot of different combinations. Intelligence is to make good move without trying all the combinations and exploring only few.