An image showing a chatbot at our current state

The Evolution of Chatbots

Most of us recognize IBM’s Watson, but many may not realize that it has evolved from a chess-winning competitor of the early 21st century to a chatbot that uses natural language processing and machine learning. Chatbots became a part of a household with Apple’s Siri and Amazon’s Alexa. Today, chatbots use conversational artificial intelligence to interact more naturally, delivering a more sophisticated human experience.

Although it may seem as though these automated conversationalists appeared overnight, the evolution of chatbots has taken over 50 years. Beginning in the early 20th century, Alan Turing developed a theory that the human brain is a digital computing mechanism that learns over time to become a universal machine. He is known for his pioneering work in artificial intelligence and cognitive science. In 1950, he created the Turing test to determine whether a computer was thinking.

Evolution of chatbots.

After Turing, researchers continued to work on what we know today as chatbots. They used various technologies such as natural language processing and artificial intelligence to produce the most artificial human experience.

Natural Language Processing (NPL)

Joseph Weizenbaum, an MIT computer scientist, began work on ELIZA in 1964. By 1966, ELIZA appeared to be conversing with humans. In reality, ELIZA, named after Eliza Doolittle in Pygmalion, was mimicking the words of the people conversing with her. She would substitute a human’s words and use them in her responses, making it appear as though she was talking with them. 

Almost ten years later, Kenneth Colby advanced the underlying principles of ELIZA by creating PARRY that used a more conversational strategy. In 1973, a conversation occurred between ELIZA and PARRY. These were early attempts at using natural language processing in chatbots.

Artificial intelligence (AI).

Jabberwacky, named after the nonsense poem Jabberwocky by Lewis Carroll, was created in 1988 by Rollo Carpenter. It is considered one of the earliest efforts of creating artificial intelligence through human interaction. According to Carpenter, it was designed to “simulate natural human chat in an interesting, entertaining and humorous manner.”

In 1995, Artificial Linguistic Internet Computer Entity or ALICE was released. In its original form, it used natural language processing. Since its appearance in 1995, ALICE has undergone several changes, migrating to an artificial intelligence language.

Conversational AI.

At the beginning of the 21st century, chatbot technology struggled to deliver a reliable and sophisticated experience. Some implementations took too much time to deliver. Others were less reliable, spewing inappropriate or offensive remarks. In 2010, with the launch of Apple’s Siri, chatbot deployment changed. Since 2010, the following chatbots have become household words:

  • Amazon’s Alexa in 2015
  • Microsoft’s Cortana, 2015
  • Facebook Chatbots, 2016

As of 2018, Facebook has over 300,000 different chatbots operating across its platform, although not all are AI-based.

Chatbots come in two forms. Rule-based chatbots use linguistic-based directives to determine how the computer interacts with a human. Machine-learning chatbots ingest data to learn how to respond in contextual situations. The AI-based chatbots deliver a more human-like experience. They can integrate with other solutions such as Robotic Process Automation (RPA). They can make recommendations, book appointments, and answer questions.

The future of conversational AI chatbots depends on how well the implementation can capitalize on the following:

  • Understanding
  • Memory
  • Sentiment
  • Personality
  • Persistence
  • Tangents

Each capability adds another dimension to AI-based interaction.

Understanding.

Interpreting a user’s request correctly is the baseline for AI. The technology is also capable of identifying and merging added information to deliver a more comprehensive answer. For example, an employee wants information on an existing product. The chatbot also knows that a recent update was released. The response includes both pieces of information. Delivering a more complete answer saves employees time and ensures they have the most current information.

Memory.

Chatbots remember. They retain pertinent information to use in a conversation or to help during future interactions. For example, a customer uses a company’s online helpdesk frequently. After several uses, the chatbot remembers that the customer always clicked on the FAQs before looking at any other information. The next time the customer asks for help, the chatbot places the FAQs at the top of the search results. Customers are happy that their preferred format is at the top of the list, maybe without realizing that FAQs are preferred.

Sentiment.

Humans use certain words when they’re happy and other words when they’re not. AI can learn to recognize the differences and assess an end user’s mood. Suppose a customer has been interacting with a chatbot about a problem for some time. The chatbot detects a change in sentiment by the length of the response and the words used. An AI chatbot refers the customer to a human because it believes the customer is becoming frustrated, ensuring the customer has a positive experience.

Personality.

AI chatbots with personality can improve engagement. Some companies create characters such as owls or robots to convey a “personality.” Language can also convey a personality or style. The closer the interaction is to human conversation, the more the user becomes engaged. Consumers who are frequent users of Siri or Alexa perceive them as having an identity or personality.

Persistence.

Chatbots use past interactions to continue conversations as users move from one device to another. People do not have to repeat themselves if they move from their phone to a laptop. If it’s one thing consumers dislike, it’s having to repeat themselves every time they start a new interaction. They expect to be remembered.

Tangents.

Humans are known to jump from topic to topic. A group is discussing vacation plans when someone asks if anyone has tried the new Thai restaurant in town. Suddenly, the group goes on a tangent about Thai restaurants without resolving vacation plans. AI chatbots can be trained to adapt to these human tangents. When a customer asks about shipping while deciding which product to buy, conversational AI chatbots are not confounded by the change. They can answer questions while directing customers to the product they need.

Limitless future.

With advances in technology happening daily, who knows what the next step in the evolution of chatbots will be. Maybe, people will report symptoms to a nurse chatbot. Or, students can have metaphysical discussions with a philosophical one.

If you’re interested in learning more about the evolution of chatbots, click the button below.

Gabriel B. is available to chat
Est. wait time: 2 min
Hi, I’m the Capacity Bot 👋
I can help you find information quickly. I'll be here if you have a question.
Sign up for our newsletter
Get the latest Capacity news and product releases directly in your inbox.
Recently Added to Capacity
Web Concierge 2.0 Beta Release .a092