'No' to ban on killer robots
Lethal Autonomous Weapon Systems (LAWS) should be regulated rather than banned says Wadham Fellow Tom Simpson in a new Policy Memo for the Blavatnik School of Government.
Putting humanity into AI
Date Published: 28.09.2020
Can AI help solve the world’s problems and what are the risks and benefits?
In a lively discussion with Wadham alumni from across the globe, Wadham's Fellow in Psychology, Professor Chris Summerfield, stressed the need to put humans inside the loop when it comes to creating AI.
Setting the global context of political division and crises such as climate change and covid, Chris highlighted the extent to which we are connected worldwide, with social, political and economic practices mediated by AI controlled technology.
There has been an explosion in the number of commercial entities wanting to build AI – from IBM and Apple to Facebook and Deep Mind - and substantial progress has been made in recent years he said. We have systems that can play complex games, AI can speak in the languages that we use and can generate html from text written in English.
Describing Artificial General Intelligence (AGI) as “an artificial agent that can match or exceed all typical human cognitive abilities in a roughly human-like environment” he asked “Where are the humans in this process?”
“AI is framed by the tools of machine learning using big data… we need to put humans inside the training loop so machines can learn from them,” he added, citing the book Human Intelligence by Wadham Honorary Fellow Stuart Russell (Physics, 1979).
Chris believes that most companies do not have a vision of what AGI would really look like. “Some think of it as a tool, others, as an oracle to inform us, but AGI is more akin to an institution that mediates interaction with each other. Artificial Narrow Intelligence is a system that can perform tasks set for it by humans. AGI is a system that can choose which tasks to perform in order to maximise the benefits for humans.
A range of searching questions followed the presentation with alumni asking about the potential of AI to be used for harm; whether emotions could be built into AI; whether human genetic code can be replicated in AI; whether an all-embracing system of AI is possible when we have such incompatible values and across cultures; regulation; and ways in which AI can help in medicine and economics.
Wadham College Warden Ken Macdonald QC thanked the global audience, particularly the many Hong Kong based alumni who joined the event which was timed to enable their participation at the end of the working day.
