If you run a travel and tourism business, and believe AI is here to stay, OpenAI, Google DeepMind, LLaMA by Meta, Anthropic, DeepSeek and Mistral AI are six names you should be familiar with.
AI’s Cambrian moment is now. But despite the technology’s rapid diversification and proliferation of deep learning techniques and technologies, it is also caught in a situation where there is a fine line between vision and hallucination.
Sharing his perspective, OAG COO Filip Filipov’s presentation during the recent Aviation Festival Asia in Singapore was titled Generative AI: There’s a Fine Line Between Vision and Hallucination, a statement he said that was generated by AI.
Common and well-known hallucinations in AI occur in chatbot systems, particularly those driven by large language models, when for example an AI-powered chatbot is asked a specific factual question, where it may confidently provide a completely wrong answer.
While the strengths of AI lie in data analysis, is domain agnostic and is powerful in language understanding, its weaknesses apart from hallucination, are being resource intense as well as ethical and privacy challenges.
Examples of generative AI aviation use cases have included airlines improving customer satisfaction, in airport operations and soon-to-be released capabilities that forecast demand and market trends with precision, empowering real-time decision-making and to price it right.
Among OTAs, Filipov shared that Expedia’s Dream It, Plan It programme has generated 29 million virtual conversations.
If OpenAI on Tripadvisor can search for the highest-rated tour of Rome and provide the details once a suitable option is found, then “winter is coming” and travel and tourism businesses need to start investing.
According to Filipov, the typical range of technology spend as part of overall revenue, based on SITA data, seems to be between four and six per cent annually,
“That number is growing roughly between 10 and 15 per cent annually, as we as an industry are accelerating the use of technology and part of our capital expenditure tied to planes and infrastructure moves to operating expenses of cloud, AI and others.
“My advice is not to focus on the overall numbers and assignment in the budget. AI is still in the early stages and we as an industry should focus on implementing the technology, rather than building it on our own; chances are Google, Meta, OpenAI and Microsoft would be better at building the AI models.
“Respectively, they will invest US$65 billion (Google), US$60 billion (Meta), US$$100 billion (OpenAI), and US$$80 billion (Microsoft) into AI infrastructure, which are numbers none of the travel players can afford to invest in a single year.
“I’d say that at least 10 to 15 per cent of the technology and IT spend of industry players should go into understanding AI, training, and licensing, but that will depend on the needs. For an airline, probably smaller number. For an OTA, probably a higher number.”
Filipov continued: “I personally believe now is the defining wave of technology for years to come. It is as big as the Internet, mobile (devices), social (platforms) and cloud (computing services).
“Many companies missed the waves, as they didn’t believe the waves will matter and now we are fully into this kind of technology. So, the biggest risk is not investing in AI.
“In terms of overestimating the potential is the quote from Bill Gates and that is ‘we overestimate what could be done in a year, and underestimate what could be done in 10.’
“We are already a few years into the cycle and the rate of development is nothing like we have seen – the first public version of ChatGPT was mostly summary of text, now just a couple of years later we have agents that can do tasks for you.
“I’d advise companies to understand one major premise – the technology will get better over time, it won’t be fit for purpose for a while, but cases will be amazing. Customer support chatbots are a mature user workflow, payments maybe not.
“So it is not a cure all, but it is pretty powerful when applied to specific problems.”
