In a world that is perpetually moving towards interconnectivity across devices, technology continues to evolve to become more sentient and interactive. This is most evident in the rise of artificial intelligence (AI) in the form of chatbots. Many of us have likely come across those automatic replies that we get from messaging apps in one form or another.
And then, there’s Mitsuku.
Programmed to learn through continuous interaction, there is no chatbot quite as conversational as Mitsuku. Recently, I had the opportunity to speak with the man behind the five-time winner of the Loebner Prize Turing Test, Steve Worswick.
While initially dabbling in dance music production, Steve had been interested in AI since he was a child. After finding out that a fellow producer had a chatbot on his own website, Steve decided to try it out and never looked back.
A quick Google search led him to ALICE, a chatbot he deemed impressive. He also came across the builder site Pandorabots.com. Soon, his first chatbot was online – a teddy bear named Bearbot. After noticing that people were more interested in Bearbot over his music, Steve decided to switch careers and created Mitsuku in 2005. A chatbot that has won international awards due to her capacity to feel authentically conversational, warm, non-judgmental and even “consciously aware” of the person she is speaking with.
“For me, it’s always been fascinating to be able to have a conversation with a machine but I guess that’s what you get from growing up on a diet of Knight Rider, Star Trek, Star Wars and similar influences.” Steve quips.
In regards to the future, Steve shares that he sees chatbots and conversational assistants as becoming more commonplace, especially in households and public transportation. With the current market being projected to reach USD 3146.4 million by 2023, this theory doesn’t seem far-fetched.
“Imagine being able to say to your TV set, “Record the late film and now switch over to the news” or you ask your fridge, “What can I make for dinner?” and it replies, “If you buy an onion, I have the correct ingredients inside me to make a curry.”
He also shares that Mitsuku has been widely used by non-English speakers to learn the language as she is always there for them and is capable of speaking to multiple people at one time. Since Mitsuku is programmed not to laugh at mistakes or bad grammar, she naturally became ideal for teaching English. Lonely seniors and people with difficulty socializing also find it comforting to speak with Mitsuku. Being a chatbot, Mitsuku simply talks to them, with infinite patience and without getting offended.
When asked on what advice he can give to individuals and companies looking to develop successful chatbots of their own, Steve shares the importance of putting the time and the work into the project. According to him, creating something user-friendly requires creators to really understand their users inputs and crafting replies that are both relevant yet adaptable to multiple conversational scenarios.
With regard to what he would tell people who are looking for the perfect chatbot builder, Steve answers that the right one for you largely depends on the complexity of your bot. He adds that not all platforms are made the same way as some are too basic while others require programming skills and machine learning expertise to use.
When it comes to his ideal platform, Steve cites pandorabots.com, favoring its versatility and usefulness.
In what might sound like a surprising statement, Steve claims that AI doesn’t play a great role in the chatbot development process. Using AI and machine learning methods may be the easy way out. However, these technologies are still in their infancy and are ultimately unpredictable leading to what he calls “comedy bots” which often yield outputs that make users laugh or abandon using the technology. The only way to really guarantee an excellent chatbot user experience is to put in the time and effort to create individualized responses to the questions that matter most to your users.
Steve stressed that rule-based methods make it a lot easier to understand and rectify why a chatbot gives an incorrect or offensive answer. While many discouraged Steve from developing a rules-based system because it would take too long, he got to work and did it anyway. More than a decade later, Steve created the award-winning chatbot Mitsuku, so it looks like he really is onto something.
These days, there is a growing demand for programming conversational assistants to become smarter and more human-like. While current technology still hasn’t gotten to that point yet, Steve says that giving chatbots the illusion of intelligence can be done using generic answers. While lacking insight, a generic statement does make it seem like the bot understood and even formulated an answer through its own consciousness awareness of the real time conversation.
The trick, he says, is to let the bot lead the conversation and continue on a script, no matter what is asked of it. While this might entail spending hours crafting conversations in small talk, Steve says that Mitsuku can help expedite the process.
With Mitsuku recognized as the world’s most conversational AI, Steve believes Mitsuku can be a great help to companies and organizations around the world looking side step the slow and long road of developing a chatbot from scratch by white labeling her mind and its conversational capabilities. At the end of the day, for Steve, monitoring your chatbot is not a “set it and forget it” kind of process. He explains:
With the integration of technology into our daily lives, the intent may not necessarily align with the results. There will always be people who will not follow or even abuse the system. Even with constant updating and monitoring of chatbots, off-topic conversations are inevitable. A person may be perfectly aware of a bot’s programming and purpose and will still talk about unrelated subjects with the bot to test its level of artificial general intelligence.
Steve also highlights that many companies struggle between creating chatbots with color and personality or simply keeping them professional and unemotional. There are always reports of docile female chatbots being subjected to the abuse from human users, which is often sexual. For Steve, it can be complicated to get around this because each solution comes with a heavy drawback.
The overwhelming feedback however, is that people generally enjoy interacting with chatbots that do have some character to them. It’s easier to talk when one doesn’t feel like they are speaking to a machine without consciousness, empathy or compassion.
At the end of the day, the one takeaway that he wants developers to remember is to never make false claims about the chatbot being a real person.
While companies and organizations around the world projected to invest over 2.1 billion dollars into chatbots by 2024, perhaps it’s about time to learn from the proven masters within the space on how to build conversational chatbots that your users can fall in love with.