With Valentines Day upon us we decided to ask the question you've all been pondering! Yes you were wondering can a robot fall in love? Well we asked our resident expert in Artificial Intelligence, Dr Luca Longo!
We might have an idea of what a robot is, but what is love? This is the first question that should be tackled" says Dr. Luca Longo of TU Dublin Computer Science. "Defining the concept of love is not trivial. Love is a psychological construct that cannot be precisely defined". But if we could, how would we go about it? Love encompasses a number of cognitive states and emotional properties. It is subject-dependent, sometimes context-specific and many factors can affect it. How do we form a list of these factors is the first real problem. Since love is subject-specific, a significant number of people should be sampled, monitored over time and a robust and wide data-gathering phase must be conducted.
Provided that a definitive list of factors influencing love can be formed, the second goal, in order to build a robot that can fall in love, is to represent each of these factors computationally. Knowledge Representation is a sub-field of Artificial Intelligence from which we can borrow some help. Thirdly, provided we can represent each of these factors formally, the next goal is to put them together towards an index of love. Machine Learning, a sub-field of Artificial Intelligence can help us with this task, and automatically learn from gathered data.
However, modelling love in this way requires a significant amount of evidence and the resulting model might be difficult to be interpreted and explained to humans. Intuitively, love is not a linear concept, thus a linear combination of these factors is likely not going to work. Here, Automated Reasoning from Artificial Intelligence can offer some help by providing us with some mechanism to deal with vague and fragmented knowledge, uncertainty, contradicting pieces of evidence and it can offer strategies for aggregating them in a non-linear fashion.
Provided a strategy can be found, it remains the ultimate problem of validating our formal representation of love. How can we be sure we are actually measuring love? Is our representation valid, reliable, general, sensitive and accurate? As the reader can imagine, defining love is far from being a simple process. "Thus, creating robots that can actually fall in love is a very difficult research task and I am sceptic on this," says Luca. "In line with my expectations of achieving strong Artificial Intelligence, sometimes referred to as general AI, any time soon.
Happy Valentines Day from TU Dublin Computer Science! :)
TU Dublin Computer Science February 2019.
For more information contact Dr Luca Longo (firstname.lastname@example.org)