Personal Identity, Model Welfare & Step-wise Transhumanism toward Whole-Brain Emulation
Location: Top floor, Frontier Tower
Format
🥂 Welcome drinks & mingling
⚡ Lightning talks (5 min each - optional for attendees)
🪪 Identity games
🗣️ Round-table discussion
TOPICS:
MODEL WELFARE
Both Anthropic and Google DeepMind have launched research programs on model welfare, asking when artificial systems merit moral status and how to detect “preferences” or distress in large models.
Are we creating philosophical Vulcans? And how does this relate to AI safety?
CONTEXT:
Sharing the World with Digital minds: https://nickbostrom.com/papers/digital-minds.pdf
Is there a tension between AI Safety and AI welfare?: https://link.springer.com/article/10.1007/s11098-025-02302-2
Lab groups:
https://github.com/paradigms-of-intelligence
https://www.anthropic.com/research/exploring-model-welfare
These pressing issues are directly related to Whole Brain Emulation; In light of uncertainty about whether mind uploads are conscious, thinkers such as Anders Sandberg propose the Principle of assuming the most (PAM): Assume that any emulated system could have the same mental properties as the original system and treat it correspondingly
https://www.aleph.se/papers/Ethics%20of%20brain%20emulations%20draft.pdf
We may be approaching a technological convergence point where we test frontier model welfare through a proxy: a model trained directly on the dynamics of an evolved biological system. This is because "non-behavioural" alternatives may fall short due to the computational irreducibility and general complexity of mech-interp:
PERSONAL IDENTITY
Relatedly, Whole Brain Emulation roadmapping shows a technically plausible route to human emulations over the coming decades.
Yet core questions—“Will an upload be me?” “Will it be conscious?”—remain philosophically and psychologically unsettled for many individuals due to our evolutionary history not involving such "non-local features" i.e being without intelligent design. We will discuss various ways to think about what the whole brain emulation field calls "Branching Identity".
Furthermore, while groups such as Netholabs are pursuing direct paths to functional WBE, there is a need to rigorously validate our methods, and be aware of the risks of creating a "Disneyland without children" too.
There is also a need to cautiously consider other potential X-Risks related to WBE, and begin to ideate around governance and oversight frameworks.
STEPWISE TRANSHUMANISM
Stepwise Transhumanism is a framework that can be used to think about how we might get to scalable uploading in an iterative and gradual fashion, by solving other problems along the way that power an R&D flywheel.
Netholabs is navigating these questions as we scale our data collection to advance Whole Brain Emulation efforts in the mouse model and invite your participation.
See Richard Ngo’s article for inspiration in the context of uploading: https://www.asimov.press/p/gentle-romance
(An essay inspired by http://www.skyhunter.com/marcs/GentleSeduction.html , just as Sam Altman's recent blog post title probably was: https://blog.samaltman.com/the-gentle-singularity)
See also an example of a successful company that followed stepwise transhumanism in pursuit of a different goal: https://www.spacex.com/humanspaceflight/mars/