Chatbot Technology: Past, Present, and Future

Chatbots are among the most visible applications of AI technology. This article explores how chatbots have evolved into important tools for consumers, businesses, and entire industries.
7 min read

Chatbots are everywhere. From online assistants, such as Microsoft’s Cortana, to “helper bots” on messaging applications like Slack, to home applications like Amazon.com’s Alexa, chatbots have become one of the most visible – and flawed – consumer-facing applications of artificial intelligence and machine learning.

Indeed, the ubiquity of chatbots stems from a broader corporate emphasis on the importance of artificial intelligence. A recent article in The Economist reported that tech companies completed roughly $21.3 billion of AI-related mergers and acquisitions – a figure that doesn’t capture the tens of billions of dollars companies are also spending on internal research and development. Chatbots represent a particularly important AI application because they interact directly with consumers.

In building chatbots that come increasingly close to passing the Turing test, engineers can create better user experiences and drive significant value for a diverse range of companies.

Chatbots seek to solve a difficult technical problem – namely, how to construct a machine that can reliably mimic human interaction and intelligence. This is, in essence, a version of the so-called Turing test, which tests whether a computer (or any other machine) has the ability to display human characteristics and intelligence. In building chatbots that come increasingly close to passing the Turing test, engineers can create better user experiences and drive significant value for a diverse range of companies.

As of now, chatbots have a long way to go in reaching this goal. This article explores the current state of chatbot technology – how it’s developed, how it’s used, and how it will continue to evolve. While chatbots are only beginning to meet their full potential, they represent a powerful tool that deserves significant attention and investment.

A Brief History: ELIZA

A brief examination of how chatbots were originally developed and conceived enables a greater understanding of both their fundamental purpose and continued evolution.

Though chatbots are still in their relative infancy technologically, they have existed for decades. One of the first chatbots, ELIZA, was developed in 1966 by computer scientist Joseph Weizenbaum at the MIT Artificial Intelligence Laboratory. Weizenbaum designed ELIZA to mimic human interaction through pattern recognition; ELIZA could not, however, react to queries in their full context. Instead, ELIZA had built-in scripts that allowed it to display the illusion of intelligence in answering questions on a given subject, such as those related to psychological evaluation.

Though Alexa is a huge leap forward from ELIZA, chatbots have yet to meet their full potential.

While ELIZA was designed to simply imitate human interaction, researchers recognized the potential of similar chatbots to provide real value to users in a wide range of contexts. Over the next four decades, engineers would experiment with more helpful chatbot applications and further expand the scope of how chatbots are defined. Though Alexa is a huge leap forward from ELIZA, chatbots have yet to meet their full potential.

Industry viewpoint: Avanade – Augmenting human workers is key to AI success

The focus on chatbot development is part of a broader push for innovation in artificial intelligence. Aaron Reich, Senior Director of Innovation and Incubation at Avanade – a consulting firm that helps businesses with digital, cloud-based, and other technology-oriented issues – sees artificial intelligence as a crucial frontier for a range of businesses.

“Companies today are very focused on artificial intelligence. Based on our research, 88 percent of global executives believe companies incorporate AI because it’s a hot topic. However, most don’t know how to use it,” Reich says. “It’s still early days, but we believe AI can be transformative for our clients in how they engage customers and empower their employees if applied in the right ways.”

To some, artificial intelligence in an enterprise context may imply greater automation, and therefore less need for human interaction and human employees. As Reich points out, however, artificial intelligence is at its most valuable when it enables greater human-machine collaboration: “The power of AI comes not 100 percent from automation, but how you get humans and machines to work together – how you can augment what a human worker can do to enhance business outcomes.”

Reich continues by noting that organizations should fully consider the complexity that underlies building a chatbot equipped to actually add value. “We have a lot of clients come to us and say that they want a chatbot, but we try to unpack that a bit. The end goal may be the bot, but that’s not where we want to start,” says Reich. To build an effective chatbot, Reich explains, an organization needs data sufficient for the chatbot to understand, reason about, and respond to a wide range of contexts with workers and customers.

While not a conventional chatbot that interacts with customers via laptop or smartphone, “Pepper” – an interactive, smart humanoid robot developed by SoftBank Robotics America (SBRA) – stands as one example of a chatbot that has created tangible value for a number of businesses. In one example (as described in an Avanade case study), ATB Financial, an Alberta-based bank, engaged Avanade and SBRA to “design and develop a pilot experience where Pepper could be placed within designated branches to greet customers, recommend products and services, conduct a simple financial literacy quiz,” and enhance the customer experience in a number of other ways.

Designed interactively and equipped with knowledge of ATB’s offerings, Pepper provides value to customers both by serving as a friendly interface for answering questions and allowing human ATB employees to deepen customer relationships in other ways. “Albertans who are already familiar with ATB’s exceptionally innovative banking environment are among the first to see how Pepper can bring something new, delightful and informative to their retail banking experience,” says Steve Carlin, Global Chief Strategy Officer for SoftBank Robotics America.

ATB’s customers have responded positively to Pepper, and the robot has prompted 542 Twitter mentions from 465 users (generating 3.2 million impressions), as well as nearly 30 unique news stories. Pepper has added value in a variety of other businesses as well, including those in financial services and retail.

Though Pepper and other chatbots have proven useful in certain contexts, Reich believes that chatbots have a long way to go before they reach their full potential. Specifically, Reich believes the future of chatbots will involve different kinds of interaction.

“I think where we are at for chatbots today… basically we’re interacting with them in the same way as we have for the past 15 to 20 years,” Reich says. “Most of the time when we type something in, a chatbot is not going to get it right.”

Going forward, Reich says, chatbot interaction may be more heavily based on voice recognition. Widely used examples of such chatbots include Amazon’s Alexa, Microsoft’s Cortana, Google Home, and Apple’s Siri. Still, a potential bottleneck remains: “The technology is getting to the point where the chatbot won’t be the barrier, it will be us as humans in terms of how we’re comfortable interacting with the tech.”

Industry Viewpoint: Agent.ai

Shay Chinn, Chief Technology Officer of Agent.ai – a technology company creating AI-powered customer service automation software – expresses similar views of the current state of chatbots. Virtual assistants like Alexa, Google Home, and Siri are “very limited in what they can do. They can parse your speech, but it’s primitive,” says Chinn. “In a lot of ways, they’re fancy toys at this point.”

Most chatbots are relatively primitive, Chinn says, because they are only capable of accurately responding to basic, heavily scripted commands, such as asking for the weather or requesting that a certain song be played. Due to these limitations, implementing chatbots in a business environment can be a high risk, high reward proposition. On the one hand, chatbots can present tremendous cost savings for customer service organizations. On the other hand, chatbots take considerable preparation, data, and infrastructure to properly design and implement.

“Maybe we’ll get to the point when chatbots can ask the right questions and fully take over from humans, but it’s a long way off,” says Chinn. “It might be an even harder technological challenge than designing a successful self-driving car.”

As Chinn explains, chatbots are still only about 80 percent accurate in voice recognition. When used in an enterprise setting, customers may grow frustrated and simply give up on interacting with a chatbot customer service representative if too many mistakes are made. Such errors can have a real, adverse impact on both a company’s bottom line and reputation.

“Some companies are putting too much faith into chatbots right now,” says Chinn. “It’s too dangerous, except in very well-defined settings and interactions.”

Yet Chinn ultimately sees a bright future for chatbots. Today, AI can augment human customer service reps. In five years, Chinn believes many businesses will have AI-assisted customer service offered 24/7. In 10 years, Chinn says, AI will run most customer service interactions, requiring human intervention only in the most difficult cases. And in 15 years, there may be personalized chatbots and chatbot-to-chatbot communication – human interaction may be so rare that companies may largely cease hiring customer service representatives.

Looking Forward: One Bot to Rule Them All?

Chatbots have a long way to go before they realize their full potential. Still, with billions of dollars of annual investment and significant human capital committed to their development, chatbots will ultimately generate significant future value in both corporate and consumer settings.

Many open questions remain. How might personalized chatbots manifest going forward? Further, there are many companies striving to develop the most advanced chatbot for both consumers and enterprise. In the race to develop the best chatbot, will one company or product truly emerge above the rest? While many chatbots may prove viable, industry consolidation may lead to a single dominant product. However the chatbot industry develops, what’s clear is that it will only become more consequential in how businesses and consumers interact.

Share

Comments

comments powered by Disqus
Recommended for you
How Top Companies Use Machine Learning
How do top organizations use machine learning? Sib Mahapatra, Editor of Toptal Insights, shares case studies demonstrating how machine learning is deployed today to help companies of all sizes create value, cut costs and drive ROI.