Ashgro

Orthogonal - is a research team with the goal of developing a strong understanding of ai safety theory in order to be able to align increasingly capable ai systems. To this end, we: - ran an extended fellowship program on outer alignment with 3 fellows and 3 mentors, with the goals of helping people upskill in this kind of research and identifying promising candidates to join our research program long-term - ran an in-person research retreat for with 8 people lasting around two weeks (some participants stayed longer), in order to gather information on the effectiveness of in-person collaboration and test the fit of potential team members - output 11 technical research posts, such as "formalizing the qaci alignment formal-goal" as well as 10 strategy posts such as "some biases and selection effects in ai risk discourse", aiming to share our insights towards aligned ai and improve the strategic discourse. - collaborated with researchers from major ai safety organizations, to clarify our understanding of technical topics and share our insights.

Aiss - ai safety researchers have unique mental health and performance impacts due to the nature of their work. Regular mental healthcare and performance specialists may not be able to provide adequate treatment without context on the field. This program provided subsidised access to professional health treatment with an individual who was versed in the particulars of ai safety research. The program was considered successfully impactful by the participants but was not renewed due to exhausting its funding.

Stampy - the stampy's ai safety info project is building an onboarding and education system for ai existential safety, so that people interested in the field or becoming more involved in efforts to address ai safety concerns can find the resources they need. In this time, stampy created multiple resources with references to materials as well as processes by which interested newcomers can be exposed to ai safety research ideas and agendas in a structured way.

Other programs which include fund for ai safety, apart research, alter, robert miles youtube, timaeus and kairos.