The Human Liveness Project

We are building The Human Liveness Project (HLP), a field-building effort to identify, roadmap, and build the tools to enable humans to continue to do good things in a world of abundant AI. Good things such as scientific exploration, entrepreneurship, and art, when AI’s capacity to perform these acts dwarfs our own.

HLP is designed to prevent human-led intellectual work from becoming effectively crowded out by that of machines. Currently, this looks like an eventuality. In this scenario, signals and feedback loops enforcing human values may also become decoupled from social systems not limited to the economy, governance, and culture. These dynamics are framed by gradual disempowerment.

Our view is that it is necessary to strategically uplift the human capacity to do good things in order to preserve our agency and existential upside as a species. We are in a race against time to build this infrastructure. There is an imminent need to identify strategies and build operational teams to realize good outcomes.

Before the end of the year, we seek to accomplish the following:

• Now – EO July:

  • Write and publish a position paper outlining the distinct and essential need for Human Liveness efforts.

• August – EO October:

  • Recruit a community of academics and other relevant thinkers to collectively assess the human capacities which are under threat from AI over the next 20 years.

  • Create dependency trees for 1-3 of the most critically threatened capacities, mapped to timelines in AI development and deployment.

  • Publish this work.

• November – EOY:

  • Design, prototype, and open source our first tool.

  • Prepare to fundraise for the organization and recruit talent in the 2027.

Albert will be responsible for execution. He wound down his first venture in biotech earlier this year, has been exploring AI safety full time for two months out of Trajectory Labs, and is committed to running this experiment.

Foy Savas conceived of the Human Liveness concept and will be a collaborator. Foy is a serial software founder with interests in formal methods, and was previously a VC and visiting academic at NYU.

The volume of intellectual work that AI has the capacity to perform is on a short and direct path to dwarfing that of humans. In order to ensure humans still have the ability to do good things in light of faster, more scalable alternatives, we need to intentionally uplift human capacities - our "liveness properties".

We believe that to ensure good existential futures for humans in light of AI, liveness properties are a distinct and essential need in addition to safety approaches that mitigate the risk of catastrophic events.

Relevant strategies for tooling may include the amplification of authentic human signals and facilitating coordination.

A key focus of our approach will be to attract net new talent and funders to work on outcomes related to gradual disempowerment. We will achieve that by presenting an opportunistic vision for builders: what picks and shovels will empower the next generation of humans to do good things?

To kickstart this effort, there is a need to rigorously define and prioritize the good things which are under threat to lay down a roadmap for motivated people.

Minimally: fund Albert to commit his full capacity to building the organization over 3 months with $17K.

Ideally: launch with 6 months of runway to execute on our full plan, with room for bringing on 1 fractional builder. This could be achieved with $50K.
Spreadsheet.