DeepMind’s latest annoyance is ‚misaligned AI agents‘

DeepMind’s latest annoyance is ‚misaligned AI agents‘

20. September 2022 0 Von Horst Buchwald

DeepMind’s latest annoyance is ‚misaligned AI agents‘

San Francisco, 9/20/2022

Alphabet’s AI lab DeepMind has distanced itself from a controversial paper co-authored by one of its researchers. It argues that misaligned AI agents are „likely“ to lead to a future existential catastrophe for humanity. The article, published in AI Magazine, was written by DeepMind’s Marcus Hutter and Oxford University researchers Michael Cohen and Michael Osborne.

The research has sparked controversy because it suggests that future AI agents will eliminate threats in the environment and devote all their energy to the pursuit of rewards.

Humanity should not expect to win against a superintelligent system that competes at every turn,“ the researchers said.

Since the study’s publication, DeepMind has asked Vice publication Motherboard to remove its link to the study.

Hutter lists his titles as a senior researcher at Google DeepMind in London and an honorary professor at the Research School of Computer Science at the Australian National University.

DeepMind claims Hutter was a co-author under his position at the university. A spokesperson told Vice, „DeepMind was not involved in this work, and the authors of the article have asked for corrections to reflect that.“