The Cosmos Institute is a nonprofit organization whose founding members include Antropic co-founder Jack Clark and former Department of Defense engineer Brendan McCord, and has announced venture programs and research initiatives that it describes as “nurturing a new generation of philosophical thinkers and entrepreneurs to pioneer uncharted territory in the age of AI.”
Cosmos chairman McCord said in a blog post that the institute will establish an AI lab at Oxford called the “Human-Centred AI Lab,” or “HAI Lab.” The lab will be led by Oxford philosopher Philip Collaras, another of the institute's founding fellows, and will aim to “translate philosophical principles of human flourishing into open-source software and AI systems.”
What exactly does “translating the philosophical principles of human flourishing” mean? It's not clear. But the gist is that it encourages the creation of AI technologies that respect human dignity without harmful disruptions (such as mass job losses due to automation). It's a tall and vague bar to be sure, but one that HAI Labs intends to strive for.
The Cosmos Institute is also launching a fellowship called the Cosmos Fellowship, which will initially consist of four fellows who will work “for one semester to one year” at HAI Labs or other partner institutions, collaborating with a Cosmos mentor and working on an independent project.[explore the] “It's the intersection of AI expertise and deep philosophical insight,” McCord said.
The institute's plans will have an investment component, in the form of a venture arm called Cosmos Ventures. Led by former DeepMind product chief Jason Chao and former Stripe corporate strategy chiefs Alex Komoroske, Darren Chu and Zoe Weinberg, Cosmos Ventures will back “exciting new prototypes, essays and creative projects that explore fundamental questions about the philosophy of technology,” McCord said.
McCord frames Cosmos Ventures as a “low-cost” venture, modeled after fellow Cosmos Ventures Tyler Cohen's Emergent Ventures: Investments range from $1,000 to $10,000 per project, and new or existing projects must produce a “major deliverable” within three months.
McCord said funding has already been raised for the first group.
“We're interested in exploring the impact of AI and machine learning in areas such as economic planning, scientific discovery, human behavior, and the role of philosophy in the tech industry,” McCord continued. “We welcome people from all backgrounds, both as individuals and teams. Cosmos works with academics and entrepreneurs across the fields of humanities, technology, and science.”
Cosmos Labs is not the first lab to try to promote a more ethical, humane vision of AI. OpenAI, founded with the mission of bringing the benefits of advanced AI to all humanity, disbanded its entire safety team. Clark's Antropic, which once positioned itself as a more secure, ethical vendor, has resisted AI regulation in recent months and actively collected data without permission.
Perhaps the Cosmos Institute would do better by embracing what McCord calls “accelerationism” – the idea that the future is not predetermined, and therefore humanity is responsible for what it brings into being.
“We must build AI that encourages exploration over complacency and active engagement over passive reliance, particularly in education, healthcare, the public sphere, and other critical domains,” McCord wrote. “We support AI that decentralizes power, enables bottom-up solutions, and empowers individuals and communities to co-create a richer, more diverse society.”
These are honorable positions, and I hope that as journalists we never allow the temptation to abandon them to be overwhelmed by money, influence or power.