The idea that you can detect lies by analyzing "microexpressions" has absorbed billions in spending by police forces and security services, despite the fact that it's junk science that performs worse than a coin-toss.
What could be griftier than selling latter-day phrenology to law enforcement? Adding "machine learning" to the mix!
The UK startup Facesoft used "a database of 300 million images of faces, some of which have been created by an AI system modeled on the human brain" (that is: "modeled" on the "human brain"), to "identify emotions like anger, fear and surprise based on micro-expressions which are often invisible to the casual observer."
The problems with replicating and falsifying this proposition don't matter when you're building a Weapon of Math Destruction whose predictions aren't compared to objective truth once they're deployed (see also: automated sentencing, predictive policing, etc).
The company was co-founded by a plastic surgeon named Allan Ponniah who claims that the police in Mumbai might use his product "for monitoring crowds to detect the evolving mob dynamics."
You know why this won't work? Because they haven't put it on the blockchain. Everyone knows that machine learning junk science products can only be salvaged by putting them on the blockchain.
"If someone smiles insincerely, their mouth may smile, but the smile doesn't reach their eyes — micro-expressions are more subtle than that and quicker," co-founder and Chief Executive Officer Allan Ponniah, who's also a plastic and reconstructive surgeon in London, told the newspaper.
Face-Reading AI Will Tell Police When Suspects Are Hiding Truth [Ellen Milligan/Bloomberg]
(via Naked Capitalism)
(Image: Cryteria, CC-BY)