Ticketing sites for music halls, performance theaters, sports stadiums, retail stores – it’s rare to visit a website these days without having a message pop up on the screen offering to help you find what you’re looking for. But how do they sometimes seem so realistic? And why are they sometimes able to almost convince us they are human?
These computer-driven assistants known as “chatbots” are the focus of Associate Professor Lydia Tapia’s Artificial Intelligence class in the UNM School of Engineering Computer Science Department.
“The premise is really to teach students about the fundamentals of artificial intelligence, what are the basic methods of it, and how to navigate and excel in the field,” Tapia explained.
As part of the Introduction to Artificial Intelligence (CS427/527) class, students design, create and code their own chatbot. A big part of that process is learning how to gauge intelligence by measuring it against human behavior. The objective is to create chatbots that, when interacting with a human, could be perceived as human-like due to their responses.
“The technology isn’t quite there yet, but these are really good attempts at making a human-like personality come from a computer,” Tapia said. “Even what’s state-of-the-art right now is not very good.”
One of the most critical aspects of making a successful chatbot is creating an expansive knowledgebase for it to pull from. By providing the bot with a lot of information and background into its specialized field, there is a greater possibility for it to be able to answer questions and provide expert feedback.
The students put their own flair on the project as they go along, with most of them creating chatbots to talk about their own hobbies or passions.
Vanessa Surjadidjaaj, a UNM Computer Science doctoral candidate, created her chatbot to be an expert on Pokémon – something she’s been involved in for years.
“It was helpful to have an already a fully developed database from my experience to pull from,” she said. “That helped me be able to program into my bot information about a lot of the items and Pokémon knowledge in general.”
Surjadidjaaj says despite already having the Pokémon knowledge herself, one of her biggest challenges was compiling the data and transferring it, or ‘training’, the chatbot.
“It is literally coming from not knowing anything at all to having an entire knowledgebase,” she said. “So, you need to go in a repetitively train it and tell it what is wrong or right and mold it into the expert you want it to be.”
UNM is not the only higher education institution looking to streamline students into the artificial intelligence industry. Amazon is currently in the later phases of a multimillion-dollar challenge to generate more realistic computer-human interaction. The Alexa Prize Socialbot Grand Challenge 3 garnered submissions from around the world, with finalists from Carnegie Mellon University, Czech Technical University in Prague, Emory University, Moscow Institute of Physics and Technology, Stanford University, University of California and University of Michigan.
“The hope is that eventually this technology will provide more human-like feedback on devices like Alexa or Google Home,” Tapia said. “Then users can say things to their devices like ‘I’m lonely’ or ‘How was your day?’ and they can feel like they have a friend in their house.”
Tapia received about $2,800 from a Google Cloud for Education grant, enabling her and her class to use Google resources for their chatbot projects. Not only does that provide students with access to the cloud and Dialogue Flow program, it also integrates their chatbot with Google Assistant.
“Something the computer cannot do is take a problem and decompose it, in order to figure out how to move forward,” Tapis said. “That’s what our students bring to the table.”
Students learn about search and solution paths, like searching for the appropriate response, and translate that search into computer code. When they leave the class, Tapia says they are armed with skills that will be useful to them in the field and marketable in the workforce.
Tapia says in addition, the future of this technology could have huge impacts on society and could eventually be used in fields like talk therapy; but only if researchers are able to perfect the coded human element.
“Artificial Intelligence really is one of the big futures of technology,” Tapia concluded. “It’s really good for UNM students to come out of our program with skills that will not only help them in their jobs and will assist them in solving problems that can’t be solved very easily on a societal level.”