For those just joining us, every day during this challenge, I’m going to try and do something different or better, using currently available Artificial Intelligence tools.
Today’s challenge is going to take on chatbots. These digital assistants are pretty commonplace now and while they are designed to make things easier, faster and more efficient, the converse is usually true. How often have you engaged with a Chatbot and left the conversation frustrated by the lack of understanding or a useful response? Hopefully though, generative AI can fix the “intelligence” problem that plagues chatbots today. Plus we’ll take a stab at creating a chatbot or two, using tools available today.
Common Uses Of Chatbots
Chatbots have come a long way since their humble beginnings as basic text-based interfaces. Today, they serve a multitude of purposes and industries, including:
- Customer Support: Businesses use chatbots to assist customers, answer FAQs, and resolve common issues, cutting down on response times and human resources.
- E-commerce: Chatbots help customers find products, provide personalized recommendations, and facilitate smooth transactions.
- Virtual Assistants: Think Siri, Google Assistant, or Alexa. These chatbots help manage tasks, answer questions, and provide information on-demand.
- Social Media: Chatbots engage with users, provide news updates, and even conduct polls or surveys.
- Healthcare: Chatbots can perform preliminary diagnoses, schedule appointments, and provide advice on minor health concerns.
- Education: Chatbots aid in learning by answering queries, providing feedback, and offering personalized tutoring.
Transforming Chatbots with Generative AI
Generative AI, like OpenAI’s GPT-4, is a game-changer for chatbots. It allows them to generate human-like text based on context, making them more intelligent and versatile. Here’s how generative AI can revolutionize chatbots:
- Enhanced Conversational Abilities: Generative AI models understand context better, enabling chatbots to provide more accurate and relevant responses.
- Personalization: Chatbots can tailor their responses to individual users, improving overall user experience.
- Better Decision-making: Generative AI helps chatbots analyze user input and make informed decisions or recommendations.
- Multilingual Support: Advanced language models empower chatbots to communicate effectively in multiple languages.
Creating A Chatbot
A few considerations before we get cracking with our own chatbot:
- Define the purpose: Determine your chatbot’s primary function and target audience.
- Choose a platform: Select a chatbot development platform that suits your needs.
- Design the conversation: Create a conversational flow, including greeting, user prompts, and possible responses.
- Train your chatbot: Leverage generative AI models and integrate them into your chatbot to improve its conversational capabilities.
- Test and refine: Test your chatbot, gather user feedback, and fine-tune its performance.
For our little experiment today, I’m going to use small parts of publicly available datasets to create two use cases (and hence bots). One will use select content from this Star Trek dataset to allow Q&A on one of the series (The Next Generation). The other will use select content from this Hotel Review dataset to query the bot about what guests think about a specific hotel (somewhat similar to our previous test using ChatGPT directly).
Simple Services To Create Your Own Chatbot
Here are a few different tools I found:
Chat Bot Kit:
One is ChatBotKit, which allows you to work with custom datasets and skillsets.
…and templates (Customer Support vs Q&S vs a more generic chabot)
Another is Berri AI, which offers a versatile Chatbot builder.
The service allows you to get started with a simple upload…
Yet another is Tidio, which allows Chatbot builds for more traditional use cases…
…for example, it offers various chatbot templates – below is an example for a Restaurant website FAQbot:
Today’s Experiment – Our Test Chatbots
For my tests, I settled on Berri.ai after trying all three tools. Based on what I had in mind, this service seemed to handle the requirements best. The backend interface below looks a little complicated but is actually quite simple. I simply uploaded sample data in text format and hit “Build Instance” – then opened up the Shareable Web App to test the bots.
Star Trek – The Next Generation Q&A
Here’s what the little experiment generated. The response times are a little slow, but they seem pretty accurate…I also love the fact that references are provided based on the content input into the bot.
Hotel Review Q&A
Here’s the bot that was fed less than 500 rows of hotel review data. Unlike my previous test that fed text directly into ChatGPT, this chatbot was less versatile and couldn’t really handle questions about the contents of the overall file. However, when asked pointed questions about a specific hotel in the file, it was able to produce accurate answers.
Well, there you have it. It’s great to see some really easy tools out there that allow you to create chatbots pretty quickly and intuitively…and with integration with GPT models, these bots are a lot better at handling context and providing useful answers. The services also allow for a range of chatbot integrations, from standard HTML on your website to standalone web apps and also plugins for popular services like WordPress, Shopify, etc.
Of course, creating an effective chatbot isn’t just about the technology. You also need to be transparent about your chatbot’s limitations, respect legal and privacy regulations and make it truly valuable in order to remove friction from the user experience.
hey very nice blog! I just followed you, if you want have a look at my blog and follow. If you want to collaborate contact me, you’ll find my contact in the “contact” section of my website. cheers!