Ever since humanity harnessed fire around a million years ago, people have put innovations to good use, such as cooking and keeping warm, along with bad, such as razing an enemy’s crops.
Today, in Bletchley, representatives of key countries, technology organisations, academia and civil society met at an AI Safety Summit at Bletchley Park to discuss the latest challenge posed by human ingenuity: how to ensure that nations and citizens globally can realise its benefits.
Speaking at the U.S. Embassy in London before joining the summit, U.S. Vice President Kamala Harris urged the international community to focus on both individual impacts, such as protecting vulnerable people from exploitation and disinformation, as well as the profound opportunities of AI in tackling global issues such as climate change, confirming American’s intention to ‘lead the way’ on safe, secure and trustworthy AI.
To mark the opening, His Majesty The King delivered a virtual address, hailing the technology’s enormous potential to transform the lives of citizens worldwide and the ‘clear imperative to ensure that this rapidly evolving technology remains safe and secure.’
The 28 countries convened by the UK for the two day summit- including the US, EU and China – agreed on The Bletchley Declaration on AI Safety, the urgent need to understand and collectively manage potential risks through a new global effort to ensure ‘frontier AI’ –where we face the most urgent and dangerous risks – is developed and deployed in a safe, responsible way for the benefit of the global community.
The declaration has been welcomed, though with caveats.
Leaving aside the doomsday scenarios beloved of Hollywood movies, along with the potential impact on the workforce, a number of tangible AI risks have emerged, such as using AI to generate fake news, creating AI that is biased because it is trained on racist, sexist or ableist data, or designing viruses capable of causing a pandemic.
However, as has been the case for any innovation since fire, the conference will focus on regulation, not a ban, because there is huge potential for good in AI, notably in tackling the twin crises of climate change and biodiversity loss, the underlying agenda of a summit organised a few days in the Science Museum, with the Natural History Museum and the Royal Botanic Gardens, Kew.
Biodiversity can bolster efforts to curb climate change because conserved or restored habitats can remove carbon dioxide from the atmosphere, helping to prevent global heating. However, around 1 million plant and animal species are threatened with extinction, many within decades, more than ever before in human history.
One way that AI can help is to analyse the sound of forest ecosystems to identify birds, amphibians and mammals to track the state of biodiversity. In a similar vein, Conservation AI has used machine learning, a form of AI, to analyse footage and images from drones or camera traps to identify wildlife — including critically endangered species – more rapidly than humans, and the same approach has been used to identify poachers too.
At Cornell University in America, researchers have used AI with data from citizen scientists to highlight which bird species occur where, when, with what other species, and under what environmental conditions, so that researchers can identify and preserve landscapes of high conservation value.
And at Flinders University in Australia, researchers have used AI to show how species interact and predict which are most likely to go extinct.
When it comes to climate change, machine learning has been used to weigh up the risk of armed conflict as a result of extreme weather, understand how well European cities are mitigating climate change, and predict the deaths due to heart disease as a result of heatwaves. Machine learning can be used to aid the development of solar power, batteries and smart grids too.
One irony is that AI itself is energy intensive, and powering it could use as much electricity as a small country. With this in mind, DeepMind in the UK is one of the companies using AI to cut energy demand, developing AI that can optimise industrial cooling in data centres and make computing more efficient.
Artificial intelligence has long been a preoccupation of the Science Museum Group, ranging from iconic objects in its computing collection, from the mechanical calculators of Babbage to early supercomputers, to blogs, events, such as one on when robots will outsmart us, mass experiments harnessing AI, coauthored publications on AI and drug design, and exhibitions, such as Science Fiction and Driverless, which examined the prospect of AI controlling autonomous vehicles, and how to fool machine learning.
AI presents us with choices that we have never had before: now we must work out how to minimise the harms and maximise its benefits.