You're pledging to donate if the project hits its minimum goal and gets approved. If not, your funds will be returned.
Cover the CHF 2,262 (~USD $2,888) Article Processing Charge for a paper accepted at Frontiers in Artificial Intelligence on foundational questions in AI alignment evaluation.
The goal is to get the CNS (Cognitive Near-Singularity) paper published at Frontiers in Artificial Intelligence, where it has been accepted pending APC payment. Publication unlocks citeability, peer-reviewed status, and distribution to the journal's AI research readership — prerequisites for institutional engagement with the broader research program on AI alignment evaluation infrastructure.
The paper develops a possibility theorem about conditions under which AI systems approach evaluation limits, along with a witness-based protocol and a corpus case study. Preprint: https://doi.org/10.5281/zenodo.19154217
Achievement is straightforward: Frontiers processes the APC, the paper enters production, and it appears in the journal within their standard publication timeline (typically 4-8 weeks after payment).
100% of requested funds go directly to Frontiers as the Article Processing Charge (CHF 2,262, approximately USD $2,888 at current exchange rates). No portion covers overhead, salary, or other expenses. If the grant is made in USD and exchange rates shift unfavorably before payment, any shortfall will be covered from personal funds; any surplus will be returned.
Single-author project. I'm an independent researcher affiliated with the Caribbean Center for Collective Intelligence (St. John's, Antigua and Barbuda) and the Nobeah Foundation (Nairobi, Kenya), working without institutional funding.
Research track record relevant to this project includes a coordinated set of working papers on AI alignment foundations, including: a Zenodo preprint on dependency-theoretic accounts of reliable AI alignment, and a companion paper currently under peer review at AI & Society applying a formal impossibility result to Anthropic's Project Glasswing initiative (preprint: https://doi.org/10.5281/zenodo.19487386). The CNS paper being funded here is the foundational result on which the Glasswing application builds.
The main failure mode is straightforward: the APC is not covered and the paper does not publish at Frontiers. In that case I would either (a) attempt to secure the APC from another source on a longer timeline, during which the acceptance may lapse under Frontiers' payment deadlines, or (b) withdraw from Frontiers and resubmit to a lower-APC or no-APC venue, which costs 6-12 months of review time and reduces the paper's reach given Frontiers' indexing and readership.
A secondary failure mode is that the paper publishes but generates limited uptake in the AI safety community. This is a real possibility I've factored in. The paper's value does not depend on immediate uptake; peer-reviewed publication is a prerequisite that enables downstream distribution work (direct outreach to named researchers at frontier labs, journalism pitches, citeability in subsequent work) that I will conduct regardless of initial reception. Publication is necessary, not sufficient, for that distribution work to function.
The ask is bounded: the funds either produce the publication outcome or they don't, within a short and verifiable timeline. There is no risk of scope creep or extended uncertainty.
Zero. This research program has been entirely self-funded over approximately a decade of sustained work. Prior fundraising attempts (primarily for larger-scale deployment projects) have not produced results, which I attribute largely to category mismatch between the asks and the evaluation frameworks of institutional funders; that is a separable strategic question from the bounded APC request here.