Would I Lie to You?
In 1950 he invented a test of a machine's ability to exhibit intelligent behaviour equivalent to, or indistinguishable from, that of a human. Turing proposed that a human would judge natural language conversations between a human and a machine that is designed to generate human-like responses. If the human judge could not reliably say which was the machine which was the human then the machine would be said to have “intelligence”.
Although the test has yet to be reliably passed, we seem to be getting close.
The term Artificial Intelligence was first proposed at a 1956 conference in Dartmouth University in the US. There was considerable excitement about the potential of machines exhibiting human behaviour, both good and bad. This was partly captured as HAL, the computer in 2001 A Space Odyssey – or The Terminator in 1984.
By 1997, rules based use of AI had improved to such an extent that IBM’s Deep Blue beat world chess champion Gary Kasparov. In 2011 IBM’s Watson won the TV quiz Jeopardy, which is the same year that Apple launched the first AI assistant Siri.
And natural language AI is the current hotspot. Digital assistants are improving and have even made their way into our cars. Almost a fifth of start-up or early stage businesses in the Cambridge cluster feature AI in some sort of way.
However, we humans don’t always play by the rules. Apparently the average person in the UK tells about 10 lies a week (according to confused.com), mostly about personal vices.
But AI is catching up there too. In a recent report from London’s City University, their AI solution is better at spotting our lies than other humans (70% compared to 54%). The study claims that their algorithm looks at clues such as how sentences are linked together or mirroring the questioner. The work revealed that people who are lying are less likely to use personal pronouns - such as "I", "me", "mine" - and tend to use more adjectives, such as "brilliant" and "sublime". Reasons for this language use could be that liars try to dissociate themselves from the content of a message, while clouding its meaning in unnecessary description.
With all forms of cyber crimes on the increase, the use of AI to detect lies could become pivotal and AI based fraud detection companies are already coming to market. There is even evidence from Michigan of it’s use in the courtroom to separate human lies from the truth.
Perhaps a future test will be whether AI is better at telling lies than a human.
If you’ve liked this commentary why not link to it and see further articles