ChatGPT marks a paradigm shift in artificial intelligence. It is generative AI that works with natural language, can answer almost any question, and solve various tasks. ChatGPT interacts with its users in real time. You ask the AI where the best beach vacation spots outside Europe are, and the relevant information magically appears on your screen.
In relation to the hospitality industry, this could mean that in the future, guest inquiries might be predominantly answered by artificial intelligence instead of your favorite colleague Sabine. This sounds like poor Sabine might soon lose her job. But don’t switch off yet! If you’re wondering whether ChatGPT can replace the customer service team in a hotel, take a look at what ChatGPT’s developers have to say:
“ChatGPT sometimes writes plausible-sounding but incorrect or nonsensical answers.”
So, no. If you want to provide your guests with an accurate service experience as before, you cannot have ChatGPT directly answer guest inquiries. At least not yet. However, there is a way to combine ChatGPT with a system that prevents the AI from giving incorrect information about the hotel.
But let’s start from the beginning. What is ChatGPT, what challenges does this technology currently present, and how can the hospitality industry utilize this new development stage of generative artificial intelligence in the near future?
Oh, and by the way: This is not one of those articles where I reveal at the end that ChatGPT wrote the entire text for me. I wrote it myself, and any mistakes are the responsibility of my human brain and the human body that houses it.
The chatbot from OpenAI took the internet by storm when it was released to the public in December 2022. It had the uncanny ability to answer all sorts of questions. It could do human-like things such as writing largely coherent essays, generating functional computer code, planning trips, and pondering the meaning of life. The AI remembers what it has written and makes careful corrections upon request. It even accepts the most arbitrary prompts and weaves stories that seamlessly tie together disparate threads. ChatGPT can tell jokes and explain why they are funny.
It is based on an algorithm for text generation and dialog-based interaction. The algorithm was fed with immense amounts of text from the internet, books, and other sources, and then underwent additional training to answer questions. OpenAI has not released full details but has shared some information in a blog post. They state that the team fed the language model with human-written answers as training data and then used a form of simulated reward and punishment known as reinforcement learning to encourage the model to provide better answers to sample questions. ChatGPT is just one of many tools that were released last year utilizing generative AI, sparking a veritable hype. Generative AI is "the next big thing."
Increasing computer power in this century has led to significant developments in artificial intelligence, which have entered our daily lives in various ways. For the first time, humans are living alongside artificial intelligences. They are in our smartphones, understand our language, assess our health for insurance, assemble our cars, or trade our money. Most artificial intelligences are specialists, excelling in one specific task.
But the development of artificial intelligence is advancing faster than ever before. Experts speak of a paradigm shift, the next evolutionary stage of AI, which we reached in 2022. AI has become generative; it can generate, or create, text, computer code, art, music, video, and a lot of controversies. The technology behind these tools is neural networks, which identify patterns in vast amounts of data and make predictions from them. Language- and text-based applications like ChatGPT are large language models (LLMs) that analyze and generate text. These models are becoming larger and more complex. Your grandpa has long since dropped out at this point, but you stay engaged! The model behind GPT-4, the latest model from OpenAI, has 100 trillion parameters. 100 trillion. I can’t tell you how many football fields that would be, sorry.
What’s new is that the technology is now available to everyone. The tools are now publicly accessible, sometimes for free, and part of everyday life. Reportedly, ChatGPT will soon be integrated into Microsoft products like Word and Outlook. This could help you, for example, answer emails.
Generative AI may soon radically change web search and how potential guests find hotels online. The world’s largest search engines are starting a race to use generative AI algorithms. Online search may soon become dialog-based and interactive. It may become less about clicking links and exploring websites and more about leaning back and asking a chatbot.
Contrary to Mark Zuckerberg’s belief, the next big thing isn’t the Metaverse (like, if anyone still remembers it), but the new wave of AI content machines. In the 1980s, we experienced a flood of products that moved tasks from paper to the PC. In the 1990s, moving these desktop products to the internet could quickly make a fortune. A decade later, the shift to mobile devices followed. In the 2020s, the focus will shift to developing generative AI.
Interacting with ChatGPT is remarkably smooth and natural. It feels like a sentient sci-fi computer—one of the good ones, hopefully. However, there are still some problems with generative models like ChatGPT. The talkative AI always sounds very confident, even when the answers are completely wrong. ChatGPT can’t distinguish between facts and made-up information. It’s like a multiple-choice test in high school: no idea, but just answered anyway. It might work, but it doesn’t have to.
Due to how ChatGPT works, finding statistical patterns in text rather than associating words with their meanings, it can fabricate facts and figures, misunderstand questions, and exhibit biases found in its training data.
For these reasons, it is currently not possible to use ChatGPT directly as a hotel chatbot answering questions about booking, the hotel, or travel planning. In customer service, it is essential that responses are accurate and consistent. Large language models, however, are a foundational technology on which chatbots and voice assistants can be built.
DialogShift is developing a new generation of hotel chatbots that combine two capabilities: the natural interaction, generative, and holistic nature of large language models and the ability to provide consistent and hotel-specific answers. Multiple language models from OpenAI are used in combination for this purpose. Among them is GPT-4, so the chatbot reliably communicates information and matches its tone of voice to the hotel and brand. The chatbot naturally writes the answers itself.
This is likely the future that has long been said to be just around the corner. 2023 is expected to be the year that reveals which meaningful applications based on these algorithms are possible.
And what about human writers, designers, or hotel employees? Do they need to worry that artificial intelligence will soon take over their jobs?
As with any technological revolution, some jobs will disappear, others will become more interesting, or new ones will emerge. Certainly, the collaboration between humans and AI will become a given. Creatives will become more like strategists and curators—supported by AI. Hotel employees will become caregivers and companions, addressing the wishes and concerns of guests on-site.
The technology will enhance the abilities of employees. It will give them more time for the human component of service, which is a cornerstone of hospitality. However, seamless digital service is also a cornerstone. The goal is to elegantly combine both.
