An Early Look at the Issues of AI Pilots in Maritime Shipping
by Nick Tidmarsh
As artificial intelligence (AI) grows as an influence in all industries, with particular media focus upon shipping by road, the less highlighted future of autonomous ships is a tantalising but daunting prospect.
The UK Government dedicated a substantial part of the Technology section of Maritime 2050, the UK’s first ever long-term maritime sector governmental strategy document, to the concept in January 2019. The report recommended that over the next 15 years the government and industry together develop proof of concept demonstrations and then legislate for autonomous vessels to attract international trade, to make autonomous shipping a reality in 15+ years. But as the shipping industry faces this new frontier, various practical and legal issues readily rear their heads.
The practical problems of implementing autonomous shipping
The more chaotic nature of maritime shipping immediately presents complications. Trucks operate within established rules of the road, while the road itself is clearly demarked, sectored and regulated (by lanes, speed signs etc.). Additionally, most other road users possess a practical knowledge of these rules. All this creates a relatively controlled environment that allows extensive testing and refinement of AI on the road.
By contrast, at sea AI pilots would be operating far larger and slower vessels, requiring superior visual sensors to observe and adjust their course early enough. Maritime traffic is less regulated, and less signage exists for AI to recognise. In addition, AI-piloted ships have far more hazardous conditions to navigate, from storms and unsettled seas to moonless nights, which would effectively blind their crucial sensors. In theory, sufficiently advanced sensors and testing should eliminate these problems, but the journey to AI shipping by sea will be delayed behind its road counterparts by the higher technical standards and limited testing environments. Future shipowners will naturally be reluctant to support the technology by volunteering their ships for refining trials, considering the greater costs and ramifications of maritime collisions that may arise in testing.
As referred to in Maritime 2050, the current law is silent on automated vessels, only imagining a crewed ship. So for now, we can only lightly touch upon the hypothetical myriad implications of such a trend.
The obvious issue with an AI-piloted ship operating with a reduced crew (or no crew at all) is that it could not react to certain dangers and legal risks. Piracy, stowaways and rescues at sea are obvious examples. New case law would necessarily emerge on the appropriate actions for AI/remote-piloted ships to take when they encounter the diversity of hazards at sea.
Autonomous vessels could factor into the choice of flags of convenience. Shipowners would no longer need to hire crew, let alone from lower-wage countries, so environmental regulations and tax should become more important considerations.
If a ship is acts without a crew, the sale and handover of such a ship is hard to imagine. Without crew members, transfer of ownership could become removed from the ship, centring around the transfer of access codes to a computer, or sale of the property title to a remote-control facility.
As with automated land vehicles, the potentially liable parties in a maritime collision increases if AI pilots are involved. The creators of the AI system, the manufacturer of the hardware, and the suppliers/installers of both systems would join the potentially liable parties if a case arises from an AI fault.
A vessel employing an AI piloting, or assisting the pilotage of, a vessel would need to record meticulous navigational and surveillance data, just for the AI to function. Such information would only further increase the clarity that tracking data is providing in modern claims.
The risks of fault within an AI system, or of an autonomous ship being unable to effectively face a more old-fashioned maritime danger, encourages a cautious approach to automating ships. Remote navigation may be less risky than full automation for a futurist shipowner.
The prospects of reduced crew costs and the removal of human error in piloting may be enticing. But any shipowner hoping to help pioneer on this exciting challenge would be wise to crew their autonomous vessels at least initially, to minimise any dangers. The most sensible route may even be to hold back initially, and let others risk their property first during the suggested roll-out in 15 years. Any new technology needs its flaws worked out, let alone one operating in an environment as fraught as the sea.
If you have any queries about this article, please get in touch with Nick Tidmarsh.