Making waves: ChatGPT brims with potential for maritime industry, say speakers at NUS MAritime Digital Efficiency Workshop
With the help of large language models, a type of artificial intelligence (AI) that powers the ChatGPT bot, shipowners could be able to report maritime incidents such as accidents and collisions more quickly.
At present, such models, or LLM, can process text and images after being trained on huge amounts of data.
“We can give some keywords to ChatGPT, such as the weather, the incident severity, and the incident location, and then ask it to write an incident report from that,” explained Dr Chen Tianyi, Research Fellow at NUS’ Department of Civil and Environmental Engineering who is researching future applications of LLM in the maritime field.
Dr Chen was sharing his insights at the 2nd MAritime Digital Efficiency (MADE) Workshop held on 14 September 2023 by the NUS Centre for Maritime Studies. The event, attended by close to 80 academics, students and members of the maritime industry, featured talks on topics such as maritime security and how AI could improve efficiency in the sector.
Elaborating on how it would work, Dr Chen said input derived from security camera footage, audio from pilot cabins, and textual records of the incident could also be fed to an LLM, which might then summarise the content, suggest causes of the incident, or automatically generate incident reports for shipowners.
While it will take time before this “multi-modal” application of LLMs in maritime incident reports becomes a norm for the industry, other panellists at the workshop offered their take on the potential applications of AI within the sector.
“AI is a tool that will accelerate your productivity,” said Mr Laurence Liew, Director of AI Innovation at AI Singapore, a research institute located at NUS.
For instance, AI can analyse documents for staff and help them decide which to prioritise. It can also spot dental defects on X-rays and optimise work schedules, said Mr Liew, pointing out that these are real-life projects AI Singapore has been working on with industry partners.
“How can it be used in maritime? The maritime industry has a lot of scheduling (needs), whether it is (for) workers on ships or at the ports,” he noted.
Despite the many promising possibilities, there are still some obstacles that stand in the way.
Getting the right data
A big challenge of using LLMs in the maritime industry is data – how accessible the data is, how much “noise” or irrelevant information it contains, and the risk of a data breach. Training the LLM is also time-consuming, and computational resources are limited, Dr Chen added.
Similar views were echoed throughout the event. “The first thing you need to do is ensure you have the right data…Reliable, scalable, and secure – that is the Holy Grail,” reiterated Mr Chakib Abi Saab, Chief Technology & Innovation Officer at Lloyd's Register, during a panel discussion.
He noted that Lloyd’s Register, which provides classification and compliance services for the maritime industry, has been able to respond to clients more efficiently due to AI. However, humans must still verify the information to make sure it is accurate.
Mr Chakib also cited his company’s portal where clients send them technical questions about different aspects of the ships they are building. Through AI, his company is able to get the question from the client, summarise it so that their employees can read it more easily, and suggest the answers – changing how manpower is allocated for different roles.
“Now we need much fewer people to answer those questions, because generative AI is doing most of the work. And we need more people on the data analysis part to ensure that (there is) really good, accurate data,” he noted, emphasising that the human touch is still critical.
“If we let AI reply (directly) to them, if the reply is incorrect and an accident happens because of that, then we would be liable.”
Similarly, while he thinks AI will one day play an “essential role” in autonomous shipping, where vessels do not have crew on them, accountability remains a key issue.
“One of the biggest challenges is regulation. What if one of these ferries hits a fisherman? Who is responsible? Is it the company that built the collision avoidance system – that didn’t work? Is it the owner of the ship?” he stated.
Stuck in silos
Concerns about data privacy and security pose another challenge. The LLM can’t be used effectively if you haven’t fed it the necessary data, but many industry players are reluctant to share theirs.
Responding to a question about security issues from moderator Yao Chen, Research Assistant Professor at NUS Computing, Mr Chakib said this involves two things.
“One is ensuring that the type of data you are (sharing) does not disclose information that is confidential. The other is the actual cybersecurity of the data,” he noted, adding: “There is not one technology today that is unbreakable. What we can do as technology professionals is to ensure that you have the best possible practices and the best possible technologies protecting your information.”
Besides “sanitising” data – scrubbing out sensitive details from it – there are other things that might make industry players more willing to share their data. These could include benefits such as intelligence and tax reductions, suggested Mr Chakib.
What the industry needs, he added, is an “orchestration” of data. “In the oceans economy – ports, ships, and so on – there’s a lot of digitalisation, but it is happening (separately). There is no one platform that is horizontal and connects all of them,” he said.
The issue of data privacy struck a chord with workshop participant, Mr Mun Yong Jian, a digital specialist from the maritime industry.
“One of my key takeaways is that proprietary data remains the main roadblock to the development of AI – which requires data,” he said. “How do we break down the silos?”
That is a question the industry will continue to mull over as it seeks to make greater use of ChatGPT and other forms of AI to enhance their work processes.