ChatGPT’s Role in the Supply Chain

The ChatGPT phenomenon

It’s hard to remember a piece of tech that’s made a bigger splash than ChatGPT. It hit 1 million users in five days and 100 million users in just two months post-launch, an unprecedented take-up rate in the world of consumer apps.?

Perhaps this was due to the fact that it is far and away the best AI chatbot ever released to the public (its flexibility, functionality and articulate use of language can veer quite sharply into the uncanny valley).?

Perhaps it’s because it’s one of those rare pieces of technology that gives us an immediate, tangible sense of what the future might feel like. Whatever the reason, its premise and performance have been compelling enough to firmly and quickly embed it in the public imagination as proof of the science fiction-like world to come.

ChatGPT can do all sorts of things, from debugging code, to saving the life of a beloved pet, to even composing music. But many shipping industry leaders are wondering what role it could play in the supply chain, and, indeed, whether deploying it is a good idea at all.?

Widescale automation is the only viable way forward for the shipping industry, but is ChatGPT in its current iteration ready to be of any help?

A closer look at ChatGPT

ChatGPT was first launched in November, 2022. In technical terms, it’s a Generative Pre-trained Transformer built on a group of Large Language Models (LLMs). The first iteration of ChatGPT was built on a family of LLMs its developer, OpenAI, collectively called GPT-3. GPT-3 was trained on 570GB of internet text data, including articles, e-books, websites and social media sources.?

GPT-4 has just launched, and the upgrade allows ChatGPT to handle multi-modal inputs, meaning it’s able to understand (if that’s what it’s really doing) both visuals and text. The upgraded ChatGPT-4 is bigger, faster and much better at multitasking, handling about 8000 words at a time.?

In essence, ChatGPT works by predicting which word, sentence and paragraph should come next, based on context provided by the data it’s been trained on. It’s startlingly good at this, producing eerily human-like responses, and generally useful answers.?

How can ChatGPT enhance supply chain efficiency?

So, how could we apply ChatGPT to the supply chain? There are a few areas where its potential applications are easily imagined.

ChatGPT’s human-like conversational and responsive abilities make it a natural fit for customer-facing applications, such as customer service chatbots and query management. Companies like Meta, Shopify and Canva are already integrating ChatGPT into their customer service chatbots.?

ChatGPT could be used to analyze large sets of supply chain data to identify patterns and forecast trends.?

Correctly trained, ChatGPT could also help generate customs documentation, contract drafts, sales proposals and any number of other shipping-related documents in an instant, cutting down hours of manual processing time.?

What are the limitations?

To my mind, there are three major limitations to ChatGPT as regards its potential application to the supply chain.?

The first is the fact that, like many algorithms of this kind, it interpolates. When faced with a gap in its information, it estimates what should fill it by examining the context on either side of the gap. This means it is (too) often inaccurate, and the fidelity of the data it's trained on has a direct impact on the quality of the information it provides.

Which brings me to my next reservation: there’s no way to check its work. ChatGPT does not supply sources, citation trails or any other means of checking the accuracy of the answers it's giving you.?

In many instances, it renders incorrect, even nonsensical, answers in a way that appears completely legitimate, a phenomenon known as “hallucination” in the world of AI computer science. ChatGPT presents itself as authoritative, and often, there’s no way to gauge its reliability. In an industry that relies heavily on accuracy and precision, this is a major liability.

Lastly, there are some significant concerns around data security. Until its developers can make clear how supply chain data will be siloed and protected, integrating ChatGPT into daily operations strikes me as high-risk.?

Are we there yet?

While there are many areas of the supply chain where ChatGPT’s potential application seems obvious, it's probably a good few years away from being of any real use to the shipping and logistics industry.?

Realistically, there are already a host of AI-powered platforms which perform many of ChatGPT’s promised functions (automated document generation, data analysis etc.) but with 100% accuracy and bullet-proof levels of data security.?

Widescale automation throughout the supply chain is inevitable. ChatGPT is certainly a thrilling piece of tech, but whether or not it’ll have a role in the supply chain of the future is still up for debate.?