Tech that thinks like us takes a step closer
IT has long been confined to the realms of science fiction.
But now advanced artificial intelligence (AI) that can ‘think’ like humans may be a step closer, thanks to breakthrough research by Scots scientists.
The Glasgow University study could pave the way for more dependable AI technology that behaves more like humans and makes fewer unpredictable errors.
The research was published in the journal Patterns and analysed deep neural networks – part of the broader family of machine learning. To visualise how the networks’ processing of information matches that of humans, 3D modelling was used.
Professor Philippe Schyns, senior author of the study and head of the university’s Institute of Neuroscience and Technology, said: ‘When building AI models that behave like humans, for instance to recognise a person’s face whenever they see it as a human would do, we have to make sure that the AI model uses the same information from the face as another human would do to recognise it. If the AI doesn’t do this, we could have the illusion the system works like humans do, but then find it gets things wrong in some circumstances.’
Although deep neural networks are often seen as the best model of human decisionmaking behaviour, even simple visual discrimination tasks reveal inconsistencies from the AI models compared to humans.
Deep neural network technology is used in applications such as face recognition. But scientists still do not fully understand how the information is processed and therefore cannot predict errors. The research team used modifiable 3D faces and asked humans to rate the similarity of randomly generated faces to four familiar identities.
They then tested the deep neural networks – checking not only whether humans and AI made the same decisions, but also whether or not it was based on the same information.
One network that correctly classified 2,000 identities was driven by a heavily caricaturised face. This was said to show the information it processed to identify the faces was very different to that used by humans.