Jaan Tallinn, a co-founder of Skype and an early investor in AI, has voiced significant worries about the trajectory of artificial intelligence development. His concerns stem from the potential for AI to advance in ways that are opaque and difficult to control, especially with the shift towards unsupervised learning that requires vast amounts of data without human oversight. Tallinn, who invested in DeepMind in $2011 and co-founded the Future of Life Institute, has actively tried to inject caution into the AI industry, often by investing in companies to gain an internal voice, though he finds this balance frustrating due to company constraints.
Tallinn's anxieties were significantly shaped by the writings of early AI researcher Eliezer Yudkowsky in $2009, leading him to dedicate his time to existential risk mitigation with a focus on AI. He argues that many possible futures resulting from current AI development do not include humans, as non-biological AI would prioritize goals that could be lethal to humanity. For instance, he suggests an advanced AI might prefer a vacuum for engineering projects, eliminating the troposphere crucial for human breathing, or harness solar energy in ways that would destroy human life.
He believes the risk posed by AI is more fundamental than climate change or synthetic biology, as solving AI risk is prerequisite to addressing other global challenges effectively. Tallinn criticizes the current "large-scale experiments" occurring every $18$ months, suggesting a minimal intervention would be to require registration of these major AI experiments. He also points to the concern of a proliferation of synthetic minds that are indistinguishable from humans, leading to issues like automatic propaganda and questioning the very nature of human civilization.
