Chatbots and other conversational AIs have exploded in popularity in recent years. From virtual assistants like Siri and Alexa to chatbots helping customers service needs, these AI systems are becoming ubiquitous. But how exactly do they work? Who’s behind the curtain powering these seemingly intelligent programs? The answer lies in machine learning and big tech companies like Google, Amazon, Microsoft and more.
What are chatbots?
Chatbots are computer programs designed to simulate conversation with human users, especially over the internet. They employ techniques like natural language processing (NLP) to understand what a user is asking and respond appropriately. The earliest chatbots date back to the 1960s, but they’ve taken off in the last decade thanks to advances in deep learning and the rise of messaging apps.
Some examples of popular chatbots include:
- Amazon’s Alexa
- Apple’s Siri
- Google Assistant
- Microsoft’s Cortana
These virtual assistants help with tasks like setting alarms, answering questions, playing music, and more. Chatbots are also popular in customer service, allowing brands to automate part of their interactions with consumers.
How do chatbots work?
Chatbots consist of several key components:
- Natural language processing (NLP): This analyzes the human language and extracts meaning from text or voice data. NLP helps the chatbot understand what a user is asking.
- Machine learning algorithms: These continue to learn about language as they interact with users. With enough data, chatbots become better at comprehending language.
- Predetermined scripts and rules: Developers set conversational rules and scripts to determine chatbot responses.
- Integrations with other systems: Chatbots often connect with external data sources, APIs and other systems to deliver intelligent responses.
Taken together, these components allow chatbots to understand user intent and respond in a conversational manner. The specific techniques used include:
- Sentiment analysis
- Named entity recognition
- Intent recognition
- Dialogue management
It’s the rapid advances in deep learning and neural networks that have enabled major leaps forward in chatbot capabilities in recent years.
Who is making chatbots?
All the major tech companies are investing heavily in conversational AI:
Google has put major resources into developing intelligent chatbots. These include:
- Google Assistant: Available on Android phones and Google Home smart speakers
- Duplex: Makes calls to book appointments and reservations on behalf of users
- Meena: A chatbot focused on having more natural, context-driven conversations
Key Google chatbot technologies include DialogFlow, Contact Center AI, and tools to build Conversational AI into apps.
Amazon
Amazon’s most prominent chatbot is Alexa, available on devices like the Amazon Echo. Users can ask Alexa questions and have it perform tasks like playing music. Other Amazon bots include:
- Amazon Lex: Lets developers quickly build chatbots with automatic speech recognition (ASR), natural language understanding (NLU), and automatic text-to-speech (TTS)
- Amazon Connect: An omnichannel cloud contact center service
- Amazon Customer Service Bots: AI-powered bots helping customer service teams
The e-commerce giant also acquired a company called Ivona that provides the text-to-speech technology behind Alexa.
Microsoft
Microsoft’s advanced natural language capabilities power chatbots like:
- Cortana: Microsoft’s intelligent virtual assistant for Windows
- Azure Bot Service: Enables developers to quickly build intelligent bots
- Dynamics 365 Virtual Agent: An AI-powered customer service chatbot
The tech giant offers chatbot-enabling technologies like the Azure Bot Framework, QnA Maker, and Speech Services.
IBM
IBM is a chatbot pioneer, developing some of the earliest systems in the 1960s. Current IBM chatbot tech includes:
- Watson Assistant: Builds conversational interfaces into apps and devices
- Watson Discovery: Connects chatbots to vast data to provide intelligent responses
IBM also offers tools like Natural Language Understanding and Natural Language Classifier. Chatbots can be built and deployed on the IBM Cloud.
The social media leader enables conversational AI through:
- Messenger Platform: Provides tools to develop chatbots on Facebook Messenger
- Wit.ai: A platform for creating voice and text-based chatbots acquired by Facebook
- PyText: A natural language processing framework focused on language understanding
Other companies
Many other technology firms also provide tools and platforms for building chatbots, including:
- Twilio – Communication APIs for messaging, voice, video and email
- Botsify – Chatbot builder tool for non-technical users
- Chatfuel – Platform for Facebook Messenger chatbots
- Freshchat – Customer support chatbots
- MobileMonkey – Chatbot builder for Facebook Messenger
There are also lots of startups working on conversational AI. As the technology improves, we’ll continue seeing more services and features powered by chatbots.
Chatbot challenges
While chatbots have come a long way, they also face some key limitations:
- Understanding context – Having a natural, flowing conversation requires shared context between the human and bot.
- Answering unexpected questions – Chatbots work well with predictable queries but struggle with questions out of left field.
- Sounding robotic – Without the right tone and personality, chatbots risk sounding bland and robotic.
- Building trust – People are wary about AI systems and building user trust remains an uphill battle.
Researchers are exploring innovations like memory networks and generative adversarial networks (GANs) to push chatbots closer to true intelligence. But we still have a long way to go.
The future of chatbots
Chatbots are expected to proliferate across customer service, ecommerce, healthcare, education and more. As they get smarter and more conversational, we’ll see wider adoption. Some predictions include:
- Gartner predicts that by 2022, 70% of interactions with technologies will involve emerging conversational platforms like chatbots and voice assistants.
- Juniper Research forecasts 4.2 billion voice assistants will be in circulation worldwide by 2024.
- An Oracle survey found 80% of business decision makers plan to use chatbots for customer interactions by 2020.
So chatbots are here to stay and will become a bigger part of our technological lives. As AI capabilities grow, they may one day have true conversations with us and pass a comprehensive Turing test. But for now, today’s chatbots still have noticeable limitations. The quest for artificial general intelligence continues.
Conclusion
Chatbots have revolutionized digital communications and offer many practical benefits for consumers and businesses. Major tech firms like Google, Amazon, Microsoft and IBM are driving advances in conversational AI. While current systems still lack human context and nuance, chatbot technology is rapidly evolving. We can expect chatbots to become smarter, more ubiquitous and more human-like in the coming years. Though true intelligence remains elusive, chatbots are paving the way for the AI systems of tomorrow.