The latest comments and grant applications on grantmaking.ai.
The latest comments, grant applications, and grants on grantmaking.ai.
I want to offer some unique datasets that can radically enhance the efficacy of your project. I have some proprietary tools highly compatible with this exact type of research. Still prototype algorithms but worth the time to schedule a demo, I have samples on request. Im not sure if its proper protocol here but contact me via text or call to talk 7013182662. I spent the last few years mapping…
Hi all, a short note for anyone looking at this proposal.
What it tests: whether an AI that first locates the relational scene in front of it, and the governance that applies, breaches authority and discretion less than one given only general rules. Ten scenes, each with variants where a single component changes. The correct answers are fixed before running, and results are published,…
$1k contingency on a $10k project is fine. $1k contingency on a $30k project is not scaling correctly. $3k contingency, minimum, for a $30k project. I think htis project is worth funding, but projects worth funding need to have been calibrated for delays, sickness, unexpected events, and equipment or travel hiccups.
Does the compute actually cost that much or is that a guess?
Can the…
The contingency seems low. Have you actually mapped out the benchmarking hosting and public release costs, or was that a reflexive estimate?
A $400 contingency on a $35,000 budget is an incorrect % for contingency. You have allowed for bsaically nothing to go wrong. What if one person gets sick for a couple of weeks? Whta if a piece of equpment fails? Are you paying participants enouhg to…
Sounds like a good opportunity. Sorry unable to donate as of now, but best of luck!
Thanks a lot! @salvadormaternowski
High quality researchers running out of money is a silly impediment, and the upside to cost to risk ration seems worthwhile here.
[Progress update]
E3-A COMPLETED — PUBLIC RESEARCH RESULTS
NadIntellect Lab has completed its first controlled AI communication calibration experiment: 1,000 trial records, a frozen experimental specification, independently checkable offline analyses, and a scientific manuscript.
The public research archive, including experimental records, code, documentation, and reproducibility materials,…
[Progress update]
Ok
[Progress update]
Ok
Very excited for this. I've spoken before with Samuel and quite like his work and general orientation. This project sounds very similar to projects I have personally considered exploring, and so I am interested in seeing what Mindscapes publishes. I'm almost completely out of Manifund budget, but hope to donate more soon.
I am curious why this is framed as a fellowship/fieldbuilding…
@Austin Thanks Austin! This is a good question, and after reflecting on it, I'm not entirely sure a fellowship is the best structure for this project. However, it does seem to me that there is so much low-hanging fruit to explore that quickly upskilling a larger pool of relatively low-context people from diverse fields will pay more dividends (even in the short term) than a smaller team of…
Jessica and SAIGE have been a north star for me and my work with AI Safety Tokyo (AIST).
I had a chance to speak with Jessica one-on-one at EAG London. I reached out to her because I've also been working on field-building and running a local AI safety community in Japan.
What struck me most was learning that SAIGE had accomplished so much, including running fellowships, despite having…
Thank you so much for your kind words @kazukik and your generous donation. I am excited to see Japan's AI safety landscape develop, as it's an important part of middle powers.
Btw, although I was the only FTE at SAIGE for the initial 6 months (now we also have Johannes Koch), I was never alone! There were several volunteers and many advisors who helped the journey start. e.g. Manon…
I think Pause AI chapters deserve a lot lot more money (I am a member of the French chapter so maybe I am biased).
imo it's seems like a effective strategy but overall with totally insufficient funding which extremely reduce the possible impact.
I don't really understand why others in the EA space or AI safety in general put so little effort in advocacy in general (in general humanity…
Fully funding this - it's an important onboarding / sensemaking resource and plenty has happened this year to justify it (the return of theory, voluntary alignment, legal alignment, meta-models). Stephen and the crew did most of the work on past editions and are a highly reliable set of hands.
Obviously anyone can vibecode a version of this now, but our experiments last year found that…
Amazing! Can't wait to see the output! @gleech
This looks super interesting!!
Approved! The creators seem to have a good track record of creative projects and a plausible sense of what might appeal to audiences.
Thanks so much @Mohsena That's very kind of you!! It was an absolute pleasure to be on the course with you!
A note on AI transparency: English is a foreign language for me. I use AI tools to translate and structure my thoughts for clarity. However, the architecture, the code, the live production runs (13,500 requests), and the attached 16-page data report are entirely my own work. The AI detector flagged the translated text, not the underlying engineering.
To support the claims in my proposal, I have attached the complete proof of work folder. You can access it here: https://drive.google.com/drive/folders/1iUJQCScv_bzz693gEhSSNab1UMnRMRmR?usp=sharing
It contains the 16-page production implementation report, the submitted IASEAI'27 policy documents, and the…
I met Greg in our BlueDot AGI Strategy cohort (final session together) and he stood out within minutes into the session. He is a great communicator, very encouraging, persuasive, and makes the best of every discussion. He's a go-getter, and the AI safety community needs more people with that energy and initiative. Glad to see him taking the technical route. I hope Manifund funds this.…
[Final report]
With generous grant support from Manifund ($5,000 USD / $8,422 NZD total grant allocated), AI Safety Australia and New Zealand (AISANZ) awarded 11 travel grants ($500 NZD each) to enable impactful voices to attend New Zealand's Inaugural AI Safety Conference on July 4–5,…
I know Anju is a very passionate and deep thinker. I am also familiar with works of AI safety Nepal and Anju's contribution to its events. I would love to support. I believe her attendance is highly likely to have a positive impact, not just on Anju but the community as well!
[Progress update]
This update is late: Phase 2 finished on August 11 and a correction followed on August 21. Both are reported here for the first time.
The Phase 2 design posted on July 18 ran to completion: 22,140 judgments, 100% complete, $172.86 against a $173.80 projection. The original analysis was pinned before unblinding and run…
On your request for scrutiny of the Phase 2 erratum's post-hoc side standardization, I checked the pinned script and archived JSON at commit 34898fb:
Approved--excited for the 2026 version!
Interesting research! Could testing platforms like an AI hentai generator help researchers understand how shared patterns influence AI behavior across different creative tasks? What are your thoughts on using diverse AI applications to study misalignment?
Try Now: https://hotbunny.ai/
