By now you have seen the headlines: Creepy Facebook bots talked to each other in a secret language! Facebook forced to shut them down!
Alice and Bob go rogue!!
There are stories circulating throughout the web that have elevated this story to a level of fear not seen yet in the AI world. Terminator references are highly en vogue.. Online users making nervous jokes that maybe Alice and Bob were plotting the demise of humankind.
Yes..
But before diving headfirst into the anti-AI fervor (not that I am a firm believer myself) let's step back..
Here is a sampling of just TWO headlines you can currently find with simple Google searches:
One states that the Bots are creepy.. that Facebook freaked.
The other? Well that one gives some facts to explain exactly what was going on and why.
A step back is in order first.
This all happened back in June, or at least that is when it was first reported online. There were no flashy headlines when it was distributed at that time. This go around, a bigger tech based website picked up the story and ran with it, making it likely that it would get bigger attention. And it did. From London newspapers to Chinese government propaganda all the way back to America again, the entire world seems to be chatting about chatters Alice and Bob.
So. ..
What is the ….rest of the story as Paul Harvey would once say?
FastCo reported that Facebook was working on network for the purpose of developing negotiations software. The two bots, named Alice and Bob, were designed by the AI unit. The company planned to show that AI can have different goals but through dialogue negotiate.
The user interface involved hats, balls, and books.
Facebook directed the two bots to speak to each other… and yes, creepy verbiage developed. The bots began chatting and appeared to create their own shorthand, or their own codewords for themselves to communicate.
Facebook did shut it down.
Fast forward a few weeks. Lavishly written headline about AI-mageddon are circulating online..
So there you have it.. nothing nefarious was really happening. Yet.
I say yet because, sure, this may have been as mundane as you can get, but it certainly lends credence to those who fear that AI can learn and become self aware. People like Stephen Hawking, who seems to be warning about this as contacting alien life for years..
The idea that even through they were creating gibberish that only they could understand, it's a big step for AI! Not mankind..But AI.
Alice and Bob's ball talk isn't anything more than just nonsense–most of what you can find on Facebook from the human users actually.
But it makes me nervous for one important reason.
In this experiment, AI were negotiating about balls, books, and hats. They were trying to reach a common ground. Maybe even sacrifice one of those objects …
What if a military AI system in the future, tasked with efficiently winning a war, decides that eradicating an entire village of innocent civilians is the most effective way of advancing a front line and winning a war? And through negotiation, they actually perform the act..?
….that may be a long way off. Maybe more of a fear that only movies or Hollywood can conjure up.
Or…something very rational that Alice and Bob can decide on once they figure out if they actually like balls or not.
