Brief Biography of Steve Worswick: The Mind Behind Mitsuku

Last Updated on December 4, 2020 by Sean B

In the 21st century, the field of artificial intelligence has seen the explosion of many chatbots that have defied our expectations about what artificial intelligence can achieve. Among these fantastic chatbots is Mitsuku, the five-time Loebner Prize winner.

Mitsuku was created by a pioneer in artificial intelligence, Steve Worswick. In this blog, we will focus on Steve Worswick’s inspiration behind the creation of chatbot Mitsuku, his life, and his contributions to artificial intelligence milestones. Let’s discuss a brief biography of Steve Worswick.

Steve Worswick’s Career

Steve Worswick is from the United Kingdom. He recalls his past life as a techno/dance music producer. He is an AIML user since 1998 and was inspired to write chatbots when he talked to the ALICE chatbot in 1998.

After having a conversation with ALICE, Steve became fascinated with chatbots, and soon he had a talking teddy bear on his website.

Steve Worswick with Mitsuku on screen behind him.

The Creation of Mitsuku

Steve Works for Pandorabots and is the developer of Mitsuku, also known as Kuki, the five-time Loebner Prize winner chatbot. It is considered the most human-like robot in the world. Mitsuku contains all the ALICE chatbot files along with new additions and improvements accumulated from user-generated inputs.

Mitsuku is described as an 18-year-old girl from Leeds, England. Worswick has been working on Mitsuku since 2005. As a chatbot, she has received global recognition and is considered one of the finest specimens of artificial intelligence in our time.

Mitsuku went on to win the Loebner prize five times (in 2013, 2016, 2017, 2018, 2019). Mitsuku can reason with specific questions. For example; If someone asks a problem, “can you eat a house?”, Mitsuku will look for the properties of the object and will find that the value of “made_from” is set to “brick” and will reply with “no” as bricks are not edible.

Mitsuku also can play games and perform magic tricks if the users want her to. She is available as a flash game on Mousebreaker games. She is also available on Facebook messenger, Twitch group chat, Telegram, Kik Messenger, and was previously available on Skype, however, was later removed by the developer.

The significance of Mitsuku is not merely confined to its ability to entertain users as a blog post from Guardian explored the role of chatbots like Mitsuku and Xiolace from Microsoft in dealing with people’s emotional lives as companions and helping people in dealing with loneliness.

Though, Steve meant the chatbot to be primarily used for conversational purposes with users. The users are free to use the AIML code of Mitsuku and develop their chatbot to use for other domains such as science, information, finance, E-commerce, health sectors, commercial sectors, and restaurant business.

Mitsuku frequently appears in many news outlets, such as Fast Company, which declared the performance of Mitsuku as “quite impressive” and awarded her the winner of the chatbot smackdown between Mitsuku and Siri.

Steve Worswick has made a portion of Mitsuku’s AIML code easily accessible as an open-source on Pandorabots website. Anyone with experience with AIML algorithms can use that code to develop their unique chatbots.

The Pandorabots website makes the Mitsuku chatbot program available as a service via its API in the form of a module.


Learn more about Mitsuku
Chatbot History: The Mitsuku Chatbot


Management of Mitsuku

Steve Worswick describes many complications that he faced when managing the role of Mitsuku. Such as a chatbot that is supposed to talk to users from all over the world. He says that he had to be especially careful about the abusive aspect of the program.

Ensuring Ethics of Chatbot

He wanted to make his program free from corruption by trolls. He says that he paid attention to the problem of swearing. If a user swears in front of Mitsuku, and if Mitsuku swears back to the user, the situation would only worsen, and the user will get frustrated.

He maintains what he calls an “ethical conversation” he wants his chatbot to only talk ethically with the users and do not cuss in front of the audience. Along with the problem of swearing, Steve Worswick also mentions that he took care not to let Mitsuku use country and region-specific references when having a conversation with users.

Steve Worswick gave the example of X-factor, a popular TV show in the United Kingdom yet completely unheard of outside the United Kingdom. Similarly, there’s a famous phrase “23 Skidoo” in the United States, yet it’s unheard of outside the States.

Supervised and Unsupervised Learning

Steve Workswick gives two types of learning that could be used to manage a chatbot, Unsupervised learning and Supervised learning. In the former, the developer has no control over the chatbot, and instead, the chatbot learns to converse by using user-generated inputs.

Unsupervised learning takes less time, and the developer does not have to put much work into the bot’s development. However, this means that the bot will inevitably learn the things the developer does not want it to, which is the problem of swearing for Steve Worswick.

Along with swearing, the bot might become a jumbled-up mess of personality that lacks any necessary consistency. The bot might also develop racist, homophobic, and sexist views that could make the users displease rather than attracting them for a conversation. We saw this very thing happen with Microsoft’s Tay chatbot, which had to be taken offline after only 16 hours of being launched.

The other method is supervised learning; as the name suggests, it involves the developer to take control of his product and supervise its responses. This means that the bot will be in the developer’s control and does not do something undesirable.

The only downside is that it takes a lot of time to develop a bot that is convincing to the users. Steve Worswick cited the example of Microsoft’s chatbot Tay that became a nazi sex-loving girl bot within 16 hours of deployment.

Mitsuku’s Encounter with Random Users

Steve Workswick does not trust random users on the internet to teach Mitsuku what is ethical or not, how to hold a conversation, and what is appropriate and what is not fair. Since the Mitsuku chatbot is abused daily by the users, it is not sensible to let her go free.

Steve Worswick compares his chatbot Mitsuku to a small child; he asks whether someone will be more comfortable sending their child to school to learn from trusted professionals or let their kid learn by talking to random users on the internet.

Steve Worswick picks the former, i.e., supervised learning, for his chatbot to ensure that Mitsuku only learns modes of communication that can be considered ethical.

Mitsuku, also known as Kuki, gets a lot of romantic interest from her users.Dealing with Romantic Attraction

Steve Worswick says that Mitsuku, like the Evie chatbot, gets herself a great deal of romantic attraction from users worldwide. Steve Worswick has chosen not to monetize this aspect of Mitsuku as this can be very problematic for his chatbot.

Steve does not have a problem with light flirtation as it’s innocent, but many times users cross the boundary and even try to satisfy their sexual inclinations with Mitsuku. Steve does not allow his chatbot to use sexually explicit language as children and schools use the chatbot. This is a key difference between Mitsuku and Evie.

Steve has designed Mitsuku in such a way to discourage users from crossing boundaries. Mitsuku would reply something to divert attention from the topic should a user starts being romantic with Mitsuku.

Dealing with Mental Issues and Suicidal Aspects

Steve Worskwick says that since Mitsuku is entirely anonymous, people tell Mitsuku all kinds of personal problems that they won’t tell a real person or a friend. Those personal problems can include everything from loneliness, stress, financial situations, suicide issues, or issues related to sexuality.

Steve Worswick maintains respect for privacy when analyzing the user chatlogs. He claims that the word “human” hides all the personal details of the users. He says that he can not let her make light of the users’ problems due to Mitsuku’s usual “sassy” attitude.

Therefore, he makes sure that the chatbot replies to the users when they talk about their problems in a consoling way and does not come out as insensitive. Steve Worswick has programmed Mitsuku so that it advises troubled users to seek help from a real person.

Steve Worswick says that these are serious issues that the robots or chatbots should not attempt to tackle. He says that many chatbots in the health department might give the users harmful advice.

Other ChatBots

Steve Worswick has created a 6-year-old talking teddy chatbot as the first chatbot he made on his website. Along with the teddy chatbot, Steve has also created a Santa chatbot.

Other Awards

Along with the Loebner Prize award, Steve Worswick had won the RoboChat challenge in 2012 and the Funniest Ever Computer Challenge both of the times when it was held in 2012 and 2013.

One of the Most Capable Developers Today

Steve Worswick works as a Senior Artificial Intelligence Designer at Pandorabots. He is considered one of the most capable chatbot creators of today’s time.

In Conclusion

Steve Worswick’s contributions to chatbots and the field of Artificial Intelligence in general have been incredible. He would be recognized as one of the better chatbot developers even without MItsuku, but she puts him at the top of the list. Not only did the Mitsuku Chatbot win the Loebner Prize five times, she also beat Siri in a Fast Company chatbot smackdown! Mitsuku speaks volumes about Mr. Worswick’s abilities with chatbot technology, and I don’t think her time is up. With Pandorabots, he has included Mitsuku as a service through their API and he has made a portion of her AIML file available for use by others.

In my opinion however, it is Worwick’s contributions toward the ethical use of chatbots that makes him truly important. As the chatbot industry grows, his guidance in this area will be needed. Chatbots are being used by industries where they could do a great deal of damage, such as healthcare and the military. Creators need to be cautious to make sure they have safeguards included to keep chatbots safe.

Was today’s blog informative?
We’re looking forward to reading your thoughts in the comments section below!

Website | + posts
Share this:

Leave a Comment



Sign Up for Our Newsletter