The internet has gone crazy with the announcement that Facebook had to switch off some AI chatbots for fear that they were becoming sentient.
Okay, it’s a great opening statement and may be representative of what the internet is all talking about, but the reality is sadly (actually, not sadly, fortunately), not even close to the apocalyptic chaos we are all predicting.
So what actually happened? Facebook has been working on chatbot based AI and new ways to train it to understand how to hold conversations. To do this they pitted two chatbots against each other, getting them to trade virtual items. The hope is that they would both learn the nuances of language and how a chatbot can take a dominant position in a negotiation.
So what went wrong? Well, it certainly wasn’t the sentience that a lot of media are reporting. Essentially, the chatbots started to get more and more efficient with language, to the point that full sentences and paragraphs were condensed down to a level that the human brain failed to comprehend. The chat bots were still very aware what each other was talking about (remember, they only have a very limited context to frame the conversations in, they do not have to contextualize it with the petabytes of external information that the human brain has to consider when thinking of language.
So then Facebook realized that this was a failed experiment and switched it off.
If you are interested in reading the full development post, you can read it here.
I don’t know, maybe the fantasy version of this is more exciting. Maybe we can just hold out for the inevitable SciFy channel made for TV version.
Anyway, sit back and relax, the robots are not taking over……YET!