Guest Post: Rise of the machines or smoke and mirrors?

Guest Post: Rise of the machines or smoke and mirrors?

A lot has been said lately about the profound dangers to humanity of AI. But not understanding its limitations is also a problem, writes Fadi Fahes, chief commercial officer of travel tech company Journey Mentor

If by chance or hibernation you missed the Bletchley Park AI summit in November, it was convened to address the risks to humans from artificial intelligence (AI). 

The opening UK/US/EU/China declaration warned near-apocalyptically of ‘a potential for serious, even catastrophic harm…from the most significant of these AI models’.

Elon Musk described AI as ‘one of the biggest threats to humanity’ and King Charles III compared it to climate crisis.

The cumulative effect of the headlines and creators warning against their own technology was, I believe, to create an existential foreboding around AI.

Advances in technologies like AI can provide solutions for the travel industry. As in other industries, people worry. But are hyper-intelligent (possibly malevolent) machines really about to replace humans? 

A more realistic if less glamorous way is to see travel tech as a tool to support and ease human workloads and improve traveller experience, not an either/or. 

Advances

AI has been around for many years, as a recent McKinsey podcast titled 'What AI means for travel–now and in the future' pointed out. It is used for many things, optimising crew or hotel room allocation, say. 

Recentlyits applications have increased - from Covid passports to chatbots offering round-the-clock support. AI is becoming ubiquitous.

The AI market in aviation is predicted to grow significantly over the next decade, according to IATA’s Industry Affairs Committee's commissioned study - The Future of the Airline Industry 2035 study and backed up by the the Artificial Intelligence in Aviation Market 2022-2030 report by Precedence Research.

As is the global market in AI hotel technology, says Joe Vargas in a piece written on trends for hospitalitynet. 

Advances are creating exciting opportunities to solve problems around changing travel patterns, low capacity, airline tech fails and staff shortages – to improve the customer journey, handle competitors and be all-round more efficient. 

But ubiquitous does not mean limitless. AI’s current limitations, together with customers’ preferences, will guarantee, I believe, a human-face to travel for the foreseeable future.

Some of these limitations were neatly illustrated in an article in TechRadar, titled ‘I asked Bing, Bard and ChatGPT to solve my anger issues’. 

‘Chatbot angst’ is a common phenomenon among travellers.

Limitations

There are six basic types of AI:

Interactive: chatbots, natural language processing etc. ChatGPT is a good example. Most travel AI is currently in this form. Interactive AIs can conduct ‘conversations’ using the conversation to figure it out based on training. Since the chatbot does not really understand, humans often need to address problems. 

Functional: includes the Internet of Things (using your phone to start the heating before you get home). It’s not common in travel yet though it might solve many complex, time-critical operational problems in future.  

Analytical: includes data assessment and sentiment analysis. Often successfully paired with Interactive AI - it can understand implications in a way Interactive AI alone cannot.

Textual: includes new generation spell-checkers. Increasingly used in travel tech for reading weather reports or - paired with Analytical AI - re-routing travellers around disruptions.

Visual: augmented reality, picture searches, ‘deep fakes’ etc. Not used much in travel tech but again I see many future applications.

Generative Artificial Intelligence (GAI): the all-thinking, all-seeing machines that worry some people. Great strides are being made but GAI has few business applications at this time. 

Most successful travel tech applications to date are almost exclusively in Analytical AI. Analytical methods can solve things like price elasticity determination, irregular operations recovery and routing. However, it is complex and can quickly become unwieldy to control.

The most commonly used in travel, Interactive AI, is useful in an age of staff shortages – chatbots, call centres, related occasions when communications are called for. That’s fine. However, most people can quickly tell they are dealing with a chatbot – and a lot don’t care for the experience, as detailed in an article for USA Today.

What one single piece of tech cannot do credibly is replace a travel agent (pilot, steward, manager, chamber maid, tour guide etc). 

It’s okay using ChatGPT to augment data collection for example, but the information collected still needs to be moderated for relevance.

Technology supports the agent rather than providing the decision-making. It helps make the travel agency (airline, hotel etc) better. How it is deployed is down to humans – and this is key. 

Search sites are an example. If a customer’s needs are fairly simple, say a single city on a known date, they may not need an agent. Tech’s role here is to facilitate and improve the self-serve experience. Such as offering multiple parallel searches and searching better  deals closer to departure. 

For more complex requirements, or if the customer does not know exactly what they want, a human agent will still be needed. Technology can help distinguish and assist the human agent to provide a better service.

Unintended bias

That’s not to say you shouldn’t worry about AI at all. But not necessarily in the way you might expect from headlines. Unintended bias is a serious issue because AI trained on the net can pick up its racial and political biases. To me the major problem currently is over-estimating AI’s capabilities, leading to unfulfilled expectations and frustration.

For today, and for some time to come, most travel tech applications have limited utility unless paired with humans for productivity and decision support. The way AI is deployed by humans substantially affects the quality of the outcome.