Are we living in reality? Is this the past, or the future? And is there a human on the other side of this screen? These questions rear up and twist back on themselves in Ubi Sunt, a genre-breaking imaginative work by Blaise Agüera y Arcas.
The title, borrowed from Latin and Medieval poetics, describes elegiac verses modeled on the formula Ubi sunt qui ante nos fuerunt?, meaning “Where are those who were before us?” Such was the mood of the anonymous early English poets who spun stories of giants and ancient battles amid the tumbled Roman masonry of 8th century Europe. Fragments of our own digital civilization stand like ruined columns throughout Ubi Sunt—transcribed lectures and drone footage, recorded cab rides and text messages. In a parallel present, an engineer is caught in the solipsistic first-person loop of a life in tech during COVID lockdown.
Is this book fiction or nonfiction? Though speculative, its historical material is accurate, and its present tense is drawn from life; some of its AI dialogs, too, are generated by interaction with a real large neural language model. Postmodern in the spirit of W.G. Sebald’s The Rings of Saturn and Benjamin Labatut’s When We Cease to Understand the World, Ubi Sunt is an erudite, compulsively readable, all-terrain joyride across the uncanny valley between yesterday and tomorrow.
Blaise Agüera y Arcas is a Fellow at Google Research who has invented AI and privacy technologies, spoken at TED and many other conferences, done pioneering work in computational humanities, and written widely circulated essays at the intersections of machine intelligence, art, ethics, and social science. He leads an organization focusing on distributed, ethical, and on-device AI at Google’s Seattle office and other locations worldwide. In 2016, he wrote an influential piece exploring the relationship between art and technology, and in 2017 he co-authored another popular essay on physiognomy and bias in AI, as well as a refutation of claims that facial structure reveals sexual orientation. His recent essays, Do Large Language Models Understand Us? and Can machines learn how to behave?, offer a perspective on the current state of the art in language AI and its implications.