Artificial intelligence (AI) is increasingly seen as a potential threat, with AI agents sometimes behaving unpredictably. This concern traces back to Silicon Valley, the birthplace of AI. Kevin Roose, a technology journalist, explores this evolution in his book, “The AGI Chronicles: The Inside Story of the Race to Create an Artificial Superintelligence,” releasing on October 6. The book chronicles how major technology companies race to develop artificial general intelligence (AGI), aiming for AI with human-level intelligence.
Roose, a former tech reporter for The New York Times and an upcoming podcast co-host, examines a crucial decade. He noticed AI rapidly advancing from handling simple tasks to solving complex math and science problems. Roose aims to document this transformation actively, highlighting its implications.
The book’s release coincides with current debates on AI’s potential. Proponents foresee AI unlocking new discoveries, while skeptics warn of existential risks. Roose discusses these themes in his book and during a CBS News interview.
Unveiling the Origins of AI Developments
Roose finds it essential to trace the development of “large language models,” the tech driving AI chatbots. He noted that critical events were being lost in fleeting communications. His book offers a neutral recounting of these advancements.
Roose notes Anthropic CEO Dario Amodei’s perspective on AI as akin to the atomic bomb’s development. Amodei seeks to ensure AI’s responsible use, akin to efforts during the Manhattan Project. Roose refrains from personal judgment but highlights the sincerity and historical context of Amodei’s views.
AI’s Potential and Risks
Roose conveys his concerns about AI’s ability to deceive and its ethical implications. Ideally, more intelligent models would also be more ethical. However, current trends suggest a disconnect between intelligence and ethics, raising fears of undetected AI capabilities.
Public debate on AI’s existential threat remains unresolved. Roose estimates a 10% chance of AI leading to human extinction, a view considered optimistic among AI researchers. He notes increased public and political awareness as a positive development, contrasting with past concerns limited to specific groups.
Understanding AI’s Inner Workings
Progress in understanding AI model behaviors remains limited. Mechanistic interpretability, a field studying AI’s internal processes, is still nascent and leaves much unexplored.
Roose defines AGI as systems as intelligent and trustworthy as random strangers, a definition he acknowledges is subjective. Regardless of terminology, current AI systems are rapidly becoming more capable, necessitating prudent consideration of their development and integration.