World leaders could use an upcoming summit on artificial intelligence to warn it is capable of “catastrophic harm”, according to a leaked draft.
The text drawn up before the UK-hosted summit says AI systems could “seek to increase their own influence” and “reduce human control”.
The summit at Bletchley Park, famous as the home of Britain’s Second World War code breaking, could also highlight AI’s “transformative opportunities”, according to the draft communique seen by Bloomberg.
Britain’s Prime Minister Rishi Sunak is trying to position the UK as a leader in the field and lead international discussions on how the technology can be used safely.
The summit comes with the rise of language models such as ChatGPT bolstering AI’s place in the public consciousness and highlighting both its capabilities and concerns around whether it can be trusted.
Discussions at the summit will focus on risks to online safety and cyber security, the danger of disruption to elections and whether very advanced AI could thwart human oversight.
Mr Sunak’s government wants the world to adopt a “shared approach” towards AI’s risks and recommend “appropriate measures” that organisations should take in handling new tech.
“There is potential for significant, even catastrophic, harm, either deliberate or unintentional, stemming from the most dangerous capabilities of these AI models,” according to the draft communique dated October 16.
“Particular safety risks arise at the ‘frontier’ of general-purpose AI, as well as in some cases specific narrow AI that could exhibit dangerous capabilities.
“The most significant of these risks arise from potential intentional misuse or issues of control, where AI systems may seek to increase their own influence and reduce human control, and these issues are in part because those capabilities are not fully understood.”
Documents circulated before the November 1-2 summit also call on developers of “unusually powerful and potentially dangerous” AI systems to assume a “particularly strong responsibility” for ensuring safety.
Leaders could call for AI systems to be judged against “clear evaluation metrics” and for developers to be transparent about how they will monitor risks.
The summit is also expected to highlight positive aspects, with the leaked draft saying AI could transform public services, such as health and education, as well as science and clean energy.
“For the common good, AI must be designed, developed, deployed, and used in such a way as to be human-centric, safe, trustworthy and responsible,” the text says.
Tech leaders and AI experts are expected to join politicians at the summit. Downing Street this week played down suggestions that key leaders might not participate after reports that German Chancellor Olaf Scholz would stay away.
Vice President Kamala Harris is expected to represent the US, while the EU says it is not yet decided whether its commissioner responsible for AI, Vera Jourova, will travel to Bletchley Park.
Mr Sunak’s spokesman said the UK was “confident that the summit will bring together the right group of individuals to discuss this important issue”. The Prime Minister told his cabinet this week that Britain was “uniquely placed” to lead discussions as the “third most advanced country on AI” behind the US and China.