Epistemic Institutions in the Age of AI

Epistemic Institutions in the Age of AI

Current-generation LLMs excel at epistemic tasks, such as education, persuasion, and the production of plausible explanations. Notably, LLMs are exceptional at generating narratives and contexts to explain a set of facts. Given the rate of technological advancement, it is reasonable to assume that AI efficacy at these tasks will continue to improve, as will the effective deployment of digital tools that perform these tasks. 

For at least 11,000 years civil institutions have played a key role in performing these epistemic functions. Governments also perform these tasks, and starting in the mid-fifteenth century, the media began to play an increasingly large role. 

Technologies that transform how tasks are performed can disrupt the organizations and industries built around those tasks. When this disruption enables newer corporations to replace older ones, it can be understood as Schumpeterian creative destruction - a normal part of progress in market-oriented systems. However, when disruption shifts power away from civil institutions to private ones, the impact on society is something else entirely.

Epistemic institutions (like universities, churches, and unions) have already struggled in the twenty-first century. For example, in the US trust in higher education institutions is still near an all-time low, regular attendance at religious services is at an all-time low, and private-sector union membership is at an all-time low. It is easy to imagine futures in which AI’s aptitude for epistemic tasks further erodes the positive role these civil institutions play in the public sphere. It is also possible to imagine futures in which the unique resources of existing epistemic institutions position them to best deploy new epistemic digital tools. In these latter scenarios, AI and other digital tools might kindle a renaissance for epistemic institutions. 

This meeting will examine what actions epistemic institutions can take to thrive in the Age of AI. Towards that end, we will convene leaders from epistemic institutions, relevant scholars, open-source technology developers, and philanthropic leaders. Specific topics of discussion include:

• Does an AI-powered society still need epistemic institutions?
• What lessons can be learned from the existing deployment of AI tools at epistemic institutions? How can AI tools be used to enhance thinking and learning? [For these questions, there will an emphasis on higher education institutions.]
• Are there digital tools that can feasibly be developed in an open-source context that would help epistemic institutions maintain their authority relative to Big Tech? (e.g. AI orchestration layers, intellectual collaboration tools, etc.) 
• Are there standards or rules that, if collectively adopted by epistemic institutions, could help them serve as a countervailing force to Big Tech?

This meeting is the second part of a larger studio on AI and Multiscale Human Intelligence. The larger studio in the first portion of the week is co-sponsored by Google DeepMind and will convene leading scholars with leaders from industry, academia, government and civil society.