AI, like everything else, revolves around stories and myths. Companies often depict AI as an arms race to justify their rapid market entry, creating a fear of the need to rush advanced AI development, even if it risks human extinction.
However, Katja Grace, the lead researcher at AI Impacts, challenges this myth. She argues that AI is not an arms race but a unique situation. In an article for Time, she points out that rushing advanced AI could be a losing move, unlike the classic arms race where a party could theoretically win.
Grace emphasizes that taking time to address AI safety issues would benefit everyone, as the improvements made by one lab or nation could be adopted by others. She compares the AI situation to a group standing on thin ice, cautioning that sprinting ahead could lead to disaster for everyone.
In contrast to a race, Grace suggests that the best individual action for AI may be to move slowly and cautiously. She urges against letting the world plunge into destruction due to a reckless race, emphasizing the need for coordinated escape strategies.
She offers a counternarrative, likening AI to the tragedy of the commons, highlighting the need for collective action to solve the coordination problem.
Grace’s data-driven insights have influenced the debate around AI risks, as seen in a survey showing that a significant percentage of machine learning researchers believe in the possibility of extremely adverse effects from AI.
Since 2022, Grace has been advocating for a more cautious approach to AI progress, challenging the myth of technological progress being inevitable. She highlights that societies have previously chosen not to pursue certain technologies or placed extensive restrictions on them.
The idea of slowing down AI progress was once taboo but is now being seriously considered, thanks to voices like Grace’s.