Emotions are our most human quality, but what if we could teach AI to understand our feelings? The goal of artificial intelligence (AI) research has shifted over the years to compute what humans could not, to beat us in specific tasks, and most recently to create an algorithm that can show its working. But while the majority of research has been concentrated on narrow AI that can do one thing very well, others have been working on the alchemic pursuit of making algorithms more human.
Narrow AI has seen tremendous progress in tasks such as computer vision and natural language processing. But while these fields have had staggering success using statistical AI methods, some question whether we are going too far in one direction. Making algorithms think like humans may require us to implement an emotional or instinctual element into AI's thinking, but human-like AI has not been so easy to achieve in the past.
AI may someday be able to work in diverse environments as well as a human, but in situations like this, in the short term, AI is far more likely to augment our labor or do things we simply don't want to do. Helping AI recognizes our emotions may not culminate in general artificial intelligence (AGI) as some suggest, but may just make it better suited to those applications that require interaction or sensitivity without the algorithm actually understanding the emotions at play. Working on the assumption that AI can handle emotional cues as long as there is real human interaction to work from, Cogito provides an AI system to help humans communicate more effectively.
Using voice analysis to improve communication and monitor mental health is an admirable way of leveraging AI for emotional improvement, at least in terms of recognizing intangible elements of our interactions, but it is not the only way to make narrow AI more human.