Sitting in a bar in Singapore, Martin Moodie asked me: “So what is ChatGPT and what can it do for the travel retail industry?”

Here, I attempt to de-mystify generative artificial intelligence (AI) and highlight the key emerging use-cases for the travel retail industry.

For many, the advent of generative AI is already sounding like the beginning of an apocalypse movie – on one side are the simple concerns of job losses and human productivity and, on the other, are far more serious threats to mankind and the planet.

Recently, Elon Musk and a group of AI experts and industry executives released an open letter asking for a six-monthly embargo on development of AI Systems.

ChatGPT is a select branch of data technology called generative AI (Gen AI). It is only about communication in human language or natural language. It’s a sub-section of AI technologies, which refers to machines that can match or excel human intelligence.

Gen AI can understand and communicate in human language. It creates a possibility of humans directly interacting with machines. Gen AI in that sense is not a cure for all problems, but resolves and revolves only around understanding and generating human language.

The current version is special with never-seen-before capabilities which make it mimic a human language with ease. Several things which make a ChatGPT so special are its precious capabilities to:

1. Comprehend and interpret human language (Natural Language Understanding – NLU)
2. Determine the user’s purpose or request in each message (Intent Recognition)
3. Maintain context and flow during a conversation (Dialogue Management)
4. Generate human-like responses – Natural Language Generation (NLG) focuses on creating text that sounds natural and appropriate based on the context and intent
5. Incorporate visual and auditory inputs and outputs to create multimodal conversational experiences (Multimodal Conversational AI).

ChatGPT is an acronym for Chat Generative Pre-Trained Transformer. The Transformer architecture is the backbone of ChatGPT, enabling its powerful language-processing capabilities.

ChatGPT is not alone. OpenAI itself has a family of algorithms for different jobs. Chat GPT has GPT, GPT 3.5 & GPT 4.0. These are upgraded versions with improved capabilities but a similar skill set – ability to understand the human language.

CODEX is a sibling of ChatGPT which focuses on writing computer programs. CODEX can convert human language to a computer program in any language you desire.

DALL.E 2 is a powerful image-generation algorithm. It can see and draw images – you can instruct it to create an image by describing what you want, and it generates it on-demand. It is being used by those in the creative and advertising worlds.

All’s well in Paradise? 

Probably not. There are two big concerns and hurdles emerging which can impact or delay the adoption of these technologies by businesses. These are:

Hallucination

This is where a large language model (LLM) such as OpenAI’s GPT4 or Google BARD makes up false information or facts not based on real data or events.

Two New York attorneys are facing sanctions for citing bogus cases after allegedly using ChatGPT to create a legal motion which never happened. Separately, Google AI Chat Bot gave incorrect information during its launch when it credited the James Webb Telescope for discovering the first exo-planet. The first photograph of an exoplanet was taken in 2004; long before the telescope was commissioned.

This is a big issue. LLM Models, like humans, tend to lie – especially in areas where they have not been trained enough. These hallucinations, however small, become a road blocker for many implementations since a single logical error is enough to derail a much larger solution.