The University of Missouri reports studies show the typical person each day spends 70% to 80% of their waking hours doing some form of communication. These include: listening (45%), speaking (30%), reading (16%) and writing (9%). Whether you are chatting with someone or not, you know that in the digital world chatbots are gaining traction.
Chatbots have been around for a while, but they have always fallen short when it comes to the ability to have natural conversations with people. Until now, that is.
Facebook's artificial intelligence (AI) researchers have been aggressively working on algorithms that increase chatbots ability to respond to people's queries in a manner that seems more like actual conversations. By concentrating its efforts on improving chatbots' abilities to converse in a more human-like manner, Facebook is betting chatbots will not only be able to better communicate, but also eventually be able to handle complex tasks, such as booking plane tickets.
To do this, Facebook's AI researchers have created algorithms that allow chatbots to significantly expand the dictionary of words they understand, as well as the myriad ways words can be used, along with some of the nuances. Providing chatbots with greater context allows them to actually learn how to more naturally interact with people online, particularly when people deviate from the more basic questions and interactions.
These algorithms, known as neural networks, enable chatbots to analyze massive quantities of data, identify patterns and map relationships between certain data points to understand how words tie together. Using this process, chatbots can learn without the need for human intervention.
Facebook's researchers are using massive quantities of existing text to teach chatbots to better understand language. They are building off of research begun in 2018 by Google, whose own researchers created an intensive learning model called Bidirectional Encoder Representation from Transformers (BERT). BERT was the first model to learn a word's meaning within its context. Facebook's model, called RoBERTa, has already reached an 88.5% accuracy level and is being trained so that its knowledge will extend beyond a single area and a specific task.
While it is uncertain when Facebook's advances will be implemented in the form of chatbots that can engage with customers in a human-like way, experts believe that is not far off. Since Facebook is posting the results of its AI research online with the entire research community, it is safe to assume that the banking industry will soon be embracing these advances as well.
Community financial institutions need to stay on top of such advancements too or risk getting left behind. After all, once chatbots can better engage with customers in a more natural way, they will offer yet another way to ensure customers are happy with the service they receive. They will also provide yet another avenue for marketing additional services and products to customers 24x7. For additional information on chatbots, read our previous BID article, Chatting About Chatbots.