It was hard to miss the recent headlines about Facebook’s decision to shut down a group of its experimental AI bots. Most of the articles and posts on the topic took an alarmist tone, portraying the bots as some new kind of sentient intelligence that had emerged accidentally and was caught before escaping from the laboratory.
Many of the published stories gave readers the impression that the errant bots had created their own language and were conversing secretly amongst themselves.
The truth, as usual, is both less alarming and more interesting. As most of us are aware, Facebook is playing catch-up with Google, Amazon, Microsoft, and Apple in the race to develop a truly user-friendly virtual assistant. The bots in question were part of an experiment to push the envelope on speed and utility.
The goal of the experiment was encouraging the bots to teach themselves the best and fastest ways to provide useful answers to questions asked by regular people in everyday common language.
Evidently, the experimenters forgot to instruct the bots to use only formal English grammar. As a result, the bots quickly ditched any semblance of grammar and focused instead on speed. Since the bots hadn’t been specifically told to use perfect grammar, they didn’t. You can’t really fault the bots for taking the shortest and fastest route to achieving their goal.
Moreover, the researchers didn’t panic when they realized that the bots were chatting in their own shorthand. The programmers apparently decided to pause the test and rewrite the code, this time including constraints that would prevent the bots from going “off script.”
Frankly, this is a great example of how the simple truth is almost always more fascinating than a batch of breathless headlines written purely as click-bait.
I applaud Facebook’s efforts to push past its competitors and to develop a new generation of AI bots. This is exactly the kind of raw competition our industry needs to keep growing, and it’s heartening to see a large company like Facebook investing in experimental new technologies.
As IT leaders, it’s our responsibility to create a sense of perspective and to explain to non-technical people why it’s important to keep experimenting, even when the initial results appear odd or disconcerting. That’s the nature of experimentation – you keep trying until you find something that works.
I recommend reading an excellent article in Snopes
that includes interviews with the actual Facebook researchers who developed the bots. Unlike many of the other articles about the bots, this one is both enlightening and useful.