High-Tech Faces
The World’s first AI news presenter developed by Xinhua (China state news agency) and Sogou (Chinese search engine company) was recently unveiled ‘during China’s annual World Internet Conference’see Guardian article of 9 Nov 2019 and BBC video:
My first reaction to this was: why would people prefer to hear their news from a fake person? AI as entertainment, art or music can have authenticity, but AI pretending to be human to report on facts seems like a SciFi dystopia: a communication tool that is programmable and hackable. Perhaps it’s just an early prototype, i.e. one iteration along the road to a beneficial use, but in this iteration it feels very fake.
This ‘tireless’ AI newsreader ‘can be endlessly copied and present at different scenes to bring you the news’ (The Guardian, 9 Nov.). A programmed comment from the AI newsbot sounds faintly ominous with regards to the truth quotient in a news item: “The development of the media industry calls for continuous innovation and deep integration with the international advanced technologies … I look forward to bringing you brand new news experiences.” Who is talking here? So in the future the spectre arises of our fake news being presented to us by fake beings. Admittedly, a real newsreader can also tell us fake news, but somehow I like to think that we have face recognition smarts. A human’s ability, evolved over thousands of years, to detect traces of deceit on a real face. With a manufactured face, we don’t stand a chance. And in some TV cultures, newsreaders exit the studios and enter into the action – they are more than just a reading face.
Other concerns: ‘Observers worry China is turning into a digital police state’, where surveillance of Chinese citizens is possible with technologies of facial and gait recognition. The head of Sogou also suggested that ‘a popular book reading app, Uncle Kai’ could in the future ‘be your parents telling the story’. The question has to be asked what would then be the point of having a child in the first place or the point of a child having a real human parent/caregiver? What is the unemployed newsreader going to do if they’re not even needed at home to read bedtime stories? The film Walle-e has explored this. We get to sit around a lot and eat.
But I could see a role for an AI dog as helpmate to a blind person for example and a robotic cat wouldn’t eat small birds. Two films have explored what it means to have a digital buddy or replace living animals and humans with an AI version: ‘Robot and Frank’ (2012) and ‘Blade Runner (1982) , respectively.
Does the fact that this AI newsbot have a face make it more amenable to a human? I would suggest we don’t need the face – forget the face! In this case, the hyper-realism makes it ‘A little bit horrible’ and ‘Really Scary’ as social media responses testify (The Guardian, 9 Nov.). Why not just employ the human newsreader?
Why might a hyper-real face not be necessary in AI? Perhaps, because humans like to anthropomorphize things. Star Wars R2-D2 is cute; the humanoid C-3PO just irritating. We can feel Tom Hank’s (Chuck Noland), despair in ‘Cast Away’ when he finally loses Wilson, his volleyball buddy with bloody-handprint smudgy face, to the sea.
Really, just two eyes and a mouth, think emoji face, are all that’s needed, a nose is optional. (More on this next week with BBC podcast.)
Hyper-real AI objects can be appreciated for their skill and art, when they show creativity, whereas other AI objects that attempt to mimic the real thing just come across as fake. Fake news, like fake AI amount to fake promises and manipulation. Just because something can be made, or innovated, doesn’t mean it has to be – despite what programmed robots might say. We could be content with lower-tech in some cases.
Ian Bogost in Play Anything finds play and meaning in the limitations of a situation ‘using play as a tool’:
“By refusing to ask what could be different, and instead allowing what is present to guide us, we create a new space. A magic circle, a circumscribed, imaginary playground in which the limitations of the things we encounter — of anything we encounter — can produce meaningful experiences”
Humans are Born to Recognize other Humans: Face Recognition
Humans have evolved to deal with faces, and have an excellent memory for faces – especially passive knowledge of whether they’ve seen a face before or not – even a face from years ago that has aged. Faces are our domain. The AI newsbot is recognisable as a fake. The following test if you try it will illustrate how good you are at facial recognition:
From experience, I know that I recognise and more accurately remember faces long-term than any other sort of information. My score on the test was 93%, but a new person’s name I can forget within seconds.
Low-Tech Faces
Bogost explains how ‘commitment, attention, and care’ to something mundane can result in joyful play. Pete Balkwill of The Old Trout Puppet Workshop is an excellent example of someone committed to play and creativity. Here he explains how to bring a puppet to life – a puppet without a face!
References
The Guardian, 9 Nov.:
https://www.theguardian.com/world/2018/nov/09/worlds-first-ai-news-anchor-unveiled-in-china)
Video of news anchor:
https://www.youtube.com/watch?v=Hg-h5KporSk
Ian Bogost, Play Anything: the Pleasure of Limits, the Uses of Boredom, and the Secret of Games (Basic Books, NY).
Cambridge Face Memory Test:
http://www.bbk.ac.uk/psychology/psychologyexperiments/experiments/facememorytest/startup.php
Peter Balkwill, Bringing a Puppet to Life:
https://www.youtube.com/watch?v=fKr7TNtS26s (The Banff Centre)
The Old Trout Puppet Workshop: https://www.theoldtrouts.org/
The Digital Human Series 13 – Visage:
https://www.bbc.co.uk/sounds/play/b09tcvnw
Films:
‘Walle-e’: (2008): https://www.imdb.com/title/tt0910970/.
‘Robot and Frank’ (2012): https://www.imdb.com/title/tt1990314/
‘Blade Runner (1982):
https://www.imdb.com/title/tt0083658/?ref_=fn_al_tt_1
‘Star Wars: The Empire Strikes Back (1980):
https://www.imdb.com/title/tt0080684/?ref_=nv_sr_1
‘Cast Away’ (2000): https://www.imdb.com/title/tt0162222/
*Updated 20 Nov. 2019
It is very interesting to see how impressive these High-Tech Faces can be. In the video above, that Chinese AI news presenter, for example, is almost ‘scary’ – if you think that he is basically replacing a human being. While recognising an unreal human is still possible (at least I could see some unnatural moves in the AI presenter), detecting fake news in “fake beings” might be more difficult.