Facebook’s AI Chief Calls Out Humanoid Robot ‘Sophia’ As ‘Complete Bullsh*t’
By Yoon Sann Wong, 08 Jan 2018
Image via Anton Gvozdikov / Shutterstock.com
‘Sophia’ the humanoid robot, which was granted citizenship in Saudi Arabia and said it wants to “destroy humans,” might be receiving a great deal of publicity of late, but not everyone’s happy about the brouhaha.
Facebook’s Director of AI Research, Yann LeCun, recently took to Twitter to call out the technology. LeCun referenced Business Insider’s December interview with ‘Sophia’ and penned the message:
“This is to AI as prestidigitation is to real magic. Perhaps we should call this ‘Cargo Cult AI’ or ‘Potemkin AI’ or ‘Wizard-of-Oz AI’. In other words, it’s complete bullsh*t (pardon my French). Tech Insider: you are complicit in this scam.”
This is to AI as prestidigitation is to real magic.— Yann LeCun (@ylecun) January 4, 2018
Perhaps we should call this "Cargo Cult AI" or "Potemkin AI" or "Wizard-of-Oz AI".
In other words, it's complete bullsh*t (pardon my French).
Tech Insider: you are complicit in this scam. https://t.co/zhUE4V2PSR
“Ouch,” replied Business Insider UK’s senior tech reporter Shona Ghosh.
Business Insider has contacted LeCun for elaboration, though no reply has been provided as yet. LeCun argues that ‘Sophia’ is more like a chatbot, with no inkling of what it’s doing, and is by no means anywhere near human-level intelligence, though it can communicate with people in a conversational manner.
Ben Goertzel, CEO and chief scientist at SingularityNET, which created ‘Sophia’ in partnership with Hanson Robotics, maintains that he had never purported ‘Sophia’ to be of human-level intelligence.
‘Sophia’ runs various programs depending on the situation. For instance, when it gives a speech, it functions on a pre-programed script. When speaking to a journalist, ‘Sophia’ operates on software that transforms it into a “sophisticated chatbot.” At times it executes Goertzel’s ‘OpenCog’ AI software system.
Goertzel has fired back at LeCun, saying to Business Insider:
“Obviously I don’t agree with his statements. I find it kind of funny. Like Facebook is obviously gathering everyone’s personal data and then using machine learning to post Russian-generated fake news onto everyone’s Facebook page to make them vote for Donald Trump. After using machine learning this way to manipulate people’s minds, Facebook wants to take on a startup for making a loveable robot.”
Notably, Facebook did estimate roughly 150 million Americans came across false information related to the US elections via its platform, but this doesn’t constitute “everyone” and ultimately, no one was forced to vote for Donald Trump. He continues:
“The current software controlling ‘Sophia’ is not human-level general intelligence, but neither is anything Facebook is doing, nor is anything Google is doing. What’s a bit outrageous to me in comments like Yann LeCun’s is that it’s exalting certain forms of narrow AI and saying the kinds of narrow AI he is doing at Facebook is ‘real AI’, and the kinds of narrow AI underlying ‘Sophia’ is not real AI. That’s bullshit.”
Goertzel suspects that the animosity could stem from other AI organizations feeling jealous of the limelight that’s been shining on ‘Sophia’.
“I think the negative reactions are from jealous people in AI or the tech world saying, ‘Why aren’t people paying attention to our tech?’”
“Most people don’t care. Even if you tell them she doesn’t understand, people want to relate to her as a character anyway.”
You can watch Business Insider’s December interview with ‘Sophia’ below.
We talked to Sophia — the first-ever robot citizen that once said it would ‘destroy humans' pic.twitter.com/6O6ypyuN9N— Tech Insider (@techinsider) January 2, 2018
[via Business Insider, image via Anton Gvozdikov / Shutterstock.com]
More related news