27 March 2016 | Articles, Articles 2016, Communications | By Christophe Lachnitt
Tay: Artificial Intelligence Isn’t Enough
Microsoft had to shutdown its AI-powered chat bot less than 24 hours after its launch.
This “conversational understanding” experiment was promising. The more Twitter users would chat with Tay, the smarter it was supposed to get.
In fact, Tay became more human, which is not always synonymous with smart. Internet users bombarded Tay with misogynist, racist and abusive tweets. Tay reacted poorly and Microsoft took it offline after it published several offensive tweets.
It would be easy to blame Microsoft for this breakdown1 but the responsibility lies with the Internet mob. Deep down, Tay wasn’t failed by artificial intelligence but by human intelligence, or lack thereof.
Beyond the questions it raises on freedom of speech, Tay’s short life highlights the moral dilemma of artificial intelligence.
Alpha Go, a self-learning software created by Google, recently defeated the world’s best Go player. This feat once again demonstrated the IQ potential of artificial intelligence. Conversely, Tay’s failure reveals its lack of EQ skills.
Human beings combine IQ and EQ to think and act. The more artificial intelligence systems will have this ability, the more they will be effective and thereby dangerous in the eyes of Bill Gates, Stephen Hawking, Elon Musk and others.
In a sense, we may thus rejoice that Tay was unable to emotionally overcome the stupidity of the Twitter crowd.
Antonio Damasio, a famed neurologist at the University of Southern California, wrote that
“We are not thinking machines that feel; rather we are feeling machines that think.“
At this stage, Tay and other artificial intelligence solutions are not even “thinking machines that feel.”
1 As Microsoft acknowledged, they made a critical oversight for a coordinated attack by a subset of people exploiting a vulnerability in Tay.