AUG 10, 2022
This year’s Reith Lectures ask: will artificial intelligence save or destroy us?
DEC 02, 2021
What could be more festive than contemplating the extinction of the human race by robot overlords.
he nights draw in, the air turns cold – what could be more festive than contemplating the extinction of the human race by robot overlords? This year’s BBC Reith Lectures are being given by Stuart J Russell, a Brit, born in Portsmouth, who founded the Center for Human-Compatible Artificial Intelligence at Berkeley, and literally wrote the book on artifical intelligence (AI).
Russell begins by giving a brief overview of AI, starting with the first successful learning program. Arthur Samuel’s draughts-playing computer proved on television in 1956 that it had learned to beat its own creator; the first self-driving Mercedes went on the autobahn in 1987. Over the past decade, deep learning systems have evolved to recognise human speech, objects and images, and today, AI is a crucial component of the economy, used in everything from search engines to autonomous delivery planes.
Still, Russell has serious concerns over AI’s rise. He quotes Francis Bacon, writing in 1609: “The mechanical arts may be turned either way, and serve as well for the cure as for the hurt.” The hurt inflicted by AI includes racial and gender bias, disinformation, deep fakes and cyber crime – even deadly weapons.
Russell looks at social media content algorithms as a key example. These algorithms aren’t, Russell explains, “particularly intelligent – and yet they have more power over people’s cognitive intake than any dictator in human history”. Designers believed these algorithms would maximise click-through by showing users more of what they already like. Instead, they learned “to modify the state of [their] environment – in this case the user’s mind – to maximise [their] own reward, by making the user more predictable. A more predictable human can be fed items they are more likely to click on – users with more extreme preferences seemed to be more predictable.” We’ve all witnessed the results.
Natural Language Processing