Project Ocean presents a thoughtful vision for the future of AI by emphasizing responsible innovation, human-centered design, and long-term governance. Rather than focusing solely on technical capability, the project recognizes that trust, privacy, transparency, and continuity are essential to building AI that people can rely on. I believe initiatives that encourage collaboration across…

Thanks for flagging this. I checked both the ARENA website and our technical trainer to better understand the course’s compute requirements. About 30% of the notebooks include GPU-dependent work, around 20% are effectively GPU-bound, and some newer alignment exercises do not run reliably on free Colab. Reliable compute matters most during the hands-on portions of the course, where participants…

Hi Jude,

I think this is a really interesting idea - ie, making a comparison on the social media effectiveness of prominent AI safety organisations, compared to other science communications and advocacy accounts. AI safety is a relatively young field compared to other science communicators, and I'm sure there would be some lessons that we can learn.

If you look at 'prominent' AI safety…

I'm very sorry, I love your idea, but I downvoted this project. I will outline some of my reasons below in no particular order and would be more than happy to change my vote if some of these concerns are addressed.

In my opinion "Solving" in the title greatly overstates the contributions of this project. I would have put it as "Studying".

In general, solo projects can work well if your top…

I downvoted because three people have already endorsed and this project seems to be overrated probeslop. Are you.playing it too safe? It wasn't clear to me from reading the project details whether there would be some upside risk of a strong original counterfactual contribution. Six months is a long commitment and Rajarshi has a great background, If you want to stick with this project, I'd…

Neutral to slightly negative, sorry. Downvoted based on this sentence alone:

our focus is building public support and political will for AI safety, not promoting specific policies.

My model is that public pressure matters less in the near term than influencing lab employees or other elites.

In the traction section when you say you conducted 50 interviews, this doesn't give me a lot of…

Hi Justin,

Where are you at so far with your existing work? I've seen the website for Persicaria, and your network of volunteers. Have you already drafted the outreach packages? Do you already have model school curriculum etc? Have you already held some outreach events with your different target groups - with schools, communities etc?

If you've already done some of these things, it would…

Hi Ankur,

I like the emphasis on moving from AI-safety research and toy implementations toward tools that builders can actually make and use — and the accountability and epistemics themes seem especially suitable for this.

Have you considered sourcing one hackathon challenge from a small, real unmet need in another active project in this funding round?

For example, a project could…

Hi Valencia,

I really appreciate the focus on lowering the technical barrier for independent researchers while keeping evaluation work systematic and reproducible.

What is the smallest bounded evaluation object that an outside project could bring to EvalLabs today — for example one claim, one system behavior, or one small dataset — and what exact artifacts would EvalLabs return?

I am…

Hi Manu,

Your description of the attention loop in decentralized fields really stood out to me — especially the point that a central recommender may not escape the loop, but become its fastest engine.

In the live, LLM-assisted map, what mechanism will preserve disagreement, missing evidence, uncertain classifications, and genuinely competing interpretations?

I am curious how the map can…

Hi Robi,

Your fixed-recipe, data-swap design seems like a strong way to isolate data-related gains. The boundary I am curious about is that deduplication, filtering, curation, enrichment, and synthetic-data generation are themselves algorithmic processes.

What attribution rule and artifact chain will allow an independent reviewer to verify that a measured gain should be assigned to the…