“Here Be Dragons”: Navigating the uncharted waters of AI, with lessons from the ocean

25 July 2023, 0930 EDT

You feel the gentle touch of the sea breeze on your face. Seagulls squawk overhead whilst waves crash against the shore. You glance at the book by your side, but its pages have lost their battle against the summer heat. You go for a swim. The feel of the cold salt water against your body brings back childhood memories of family trips to the seaside.

When you return to your beach blanket, you catch sight of a hulking shadow cutting across the horizon: a merchant ship carrying goods that may well eventually make their way to your own doorstep. The colossal creature continues its journey across the sea; you briefly consider tracking its journey on your mobile phone before deciding to give your book another try.

If you had lived a middle-class European life only five centuries ago, the ocean would almost certainly not have filled you with such contentment. It would, most likely, have been a source of danger, mystery, fear, and opportunity. The ocean was the untamed, ungoverned dwelling place of macabre and exotic creatures. One journeyed to sea with no promise of return, no expectation of avoiding disease, starvation, piracy, or shipwreck: as the ancient cartographers wrote at the edge of their maps, “Hic Sunt Dracones.”

Only two frontiers now hold such a potent brew of exhilaration, fear, and uncertainty in the popular imagination: outer space and technological innovation. In 2023, the dragon hunters have their sights firmly set on one domain in particular: artificial intelligence (AI). 

Whether or not there really be dragons, the political and legal history of the ocean can shed light on the challenges posed by AI.

Mutually Assured (Self) Destruction?

Much of the recent focus on AI as our great “unknown unknown” is driven by ChatGPT. Numerous tweets describe ChatGPT “prompts” that “will make you superhuman.” Newspapers routinely run stories about how ChatGPT, and other large language models (LLMs), pose a political danger to humanity because of its unprecedent capacity to propagate extremist messages presented as true facts. Some even predict that ChatGPT could extinguish humanity by generating a new species whose intelligence is greater than ours (within 2 years!)

The political discourse on AI scarcely mentions the sea. Commentators are more likely to draw analogies between AI and nuclear weapons. We find this analogy in policy discourse. The UK, and Japan named their recent agreement to cooperate on industrial science, innovation, and technology the “Hiroshima Accord.” Some propose an IAEA-like institution in London to monitor the use of digital technology. Scientists have started to conduct war games involving scenarios in which AI assumes control over nuclear codes. Some even argue that an American-led international agreement, similar to the Geneva Convention, is necessary to control the apocalyptic potential of AI, which, unlike humans, lacks a moral compass

But what happens when we stop looking at the catastrophic examples of nuclear weapons and turn, instead, to the ocean to find alternative answers for the challenges posed by AI?    

Welcome to the ocean of data 

Infrastructure provides the clearest connection between AI and the sea: a significant percentage of international digital telecommunication—and hence of the information that undergirds the global economy—flows through a series of underwater cables. But the principles governing our world’s oceans may provide guidance for how to manage our “ocean of data”. These principles tap into the problem of excess and overwhelming abundancy, pervading contingency as well as the ambivalent gains of interdependency.  

The ever-changing nature of the ocean, always exceeding itself and our understanding, is similar to the continuous stream of data that we emit every day: From credit records to the flickering up-and-down movement of our thumbs against the screens of our phones, we are constantly updating the narratives of millions on our social media timelines. ChatGPT is perhaps the most advanced vessel ever produced to navigate the ocean of data that we ourselves have created. Just like the depletion of fish-stocks can jeopardise marine biodiversity, AI can be a potential danger to the inhabitants of this new digital ecosystem of data: humans.  

What went well…

The questions that Grotius and others raised nearly five hundred years ago concerning the “freedom” of the oceans apply to the regulation of AI: How “freely” should data circulate on its own? How should governments influence the processing and distribution of data? What role should private companies play in the “sea” of data and analysis?

The most recent version for these answers are in the United Nations Convention on the Law of the Sea (UNCLOS). UNCLOS, along with a series of international agreements, regulate the security of ships and ports; supress unlawful actions against navigation; prevent the pollution of marine environment; ensure the protection of deep-seabed-related activities and perform a series of other regulatory activities. Ocean governance requires that States and non-state organisations coexist to guarantee the so called “good order at sea”. The result is a complex web of interdependence where merchant ships operating under flags of convenience, carrying billions of dollars of international goods, blend populations, nationalities, sovereignties, and global companies. 

However, these series of regulations only give us a rose-tinted glimpse of how complex managing unknown unknowns can be. Other dragons came back to haunt us, yet to be slain. 

The bite of the dragon 

The Atlantic was once the centre of a very different kind of transnational trade: the horrific ‘Middle Passage” that transported enslaved Africans—if they survived the journey—to the Americas. Between 1501 and 1875 over 10 million women, men, and children were victims of the Transatlantic slave trade

After centuries of profiting from chattel slavery, UK policy turned toward its eradication. Parliament passed an act abolishing the slave trade in 1807, and one prohibiting it throughout the Empire in 1833; Britain’s West African Squadron seized thousands of slave ships. But the Royal Navy, despite all its might at the time, proved unable to successfully police the entirety of the vast South Atlantic. The slave trade at sea endured; British banks even aided and abetted slavery in Brazil, where it remained legal until 1888,

Even after its abolition, the Brazilian elite upheld a “whitening ideology” that perpetuated many of the injustices of slavery. Naturalists, historians, legal theorists, journalists, and physicians sought to create a European idea of race that not only excluded black people from access to equal rights and privileges, but cast them as biologically inferior. Brazil’s legislative authorities implemented bans against numerous Afro-Brazilian cultural traditions. It designed policies to increase European immigration and encourage racial whitening. Descendants of slaves in Brazil still suffer lasting socio-economic disparities, such as poorer health, lower educational attainment, higher incarceration rates, and continued racial discrimination. 

Navigating Unforeseen Challenges

Long before UNCLOS, attempts to govern the sea generated complex, and unpredictable socio-political effects, such as the racial injustice that slave trade produced, with which we still live today. The Transatlantic slave trade and its impact in Brazil, is an example of the link between the management of an “oceanic unknown unknown” and the systemic racism we currently see in international politics. 

Such a contingent political trajectory offers a more comprehensive and balanced approach to AI than nuclear weapons and its fear-infused model of great power rivalry. Transforming the mass of water that surrounds the planet from an imaginary ecosystem of monsters to a complex set of regulations that keeps the stability of the world economy and allows us to enjoy nice holidays on the beach was never an inevitable process. 

Similarly, adapting the uses and regulations of AI with ever-increasing LLMs such as ChatGPT, will also bring setbacks and unforeseen socio-political problems. But this time, we can draw from hundreds of years of history on ocean governance which might prove the best maps to avoid dragons.