Computers can solve murders, but humans still best detectives

COMPUTERS have been trained to solve murder cases – but not as well as humans yet, a study has found.

The artificially intelligent machines “watched” box sets of the TV crime drama CSI before correctly identifying the murderer during the final part of an episode 60 per cent of the time.

Although this lagged behind humans, who guessed correctly 85 per cent of the time, scientists claim it represents a big step forward in human understanding by computers.

The study was aimed at enabling machines to identify a fictional killer – by assimilating information from images, audio, transcribed dialogue and scene descriptions.

Dr Lea Frermann, from the University of Edinburgh’s School of Informatics, said: “Pinpointing the perpetrator in a TV show is very difficult for computers, but our model performed encouragingly

‘Finding the perpetrator in a TV show is very difficult for computers, but our model performed well’

well.”

Her team wanted to show computers – which are increasingly being used for medical treatments – could solve problems that are challenging for humans.

Dr Frermann and colleagues designed their model to solve arbitrary problems based on acquiring data. Such devices could play a role in developing efficient algorithms for realworld tasks that require complex reasoning, said the researchers.

They mapped footage, script and background sounds from the show into a machine-readable format, which was fed into a computer that learned to process the plot as each episode unfolded, continually revising the criminal’s identity.

However humans who watched the same shows were almost one and a half times better – picking those responsible 85 per cent of the time.

Scientists taught machines to approach solving crimes by considering who might be guilty from their behaviour as the plot unfolds. The data comprised 39 CSI episodes which followed a regular plot.

The study was published in Transactions of the Association for Computational Linguistics.