Open Thread 445

This is the weekly open thread on AI X Risk. Users are encouraged to discuss any topic they desire, pose random inquiries, and interact with the community. ACX possesses a subreddit, Discord server, and a bulletin board for communication. Certain content is accessible to all, while some material is restricted to subscribers. Subscribers can subscribe via this link.

A recent correction has been noted regarding the last links post. The claim that a shingles vaccine reduced dementia risk was found to be unreplicable in Britain, and is now considered unreliable. Alex brought this to the attention of the author.

Resolution, a newly established AI safety organization, has been highlighted. Resolution aims to merge automated alignment research with deep theoretical understanding. They believe this requires a sizable organization equipped with substantial computational resources. Resolution recently secured $160 million in funding from Coefficient Giving and is rapidly scaling up.

They are searching for a COO / operational co-founder to collaborate with Geoffrey Irving. This individual will focus on strategic planning, including hiring priorities, capital allocation across various research projects, maintaining a culture that appreciates negative research outcomes, expanding the organization, forging stronger partnerships with frontier companies, and creating an exceptional user experience for staff.

Resolution has already raised $160 million, merged with Timaeus, secured office space, and currently employs around 30 individuals. They are seeking applicants with experience in leading startups, scaling research nonprofits, and managing operations at frontier labs. For more information on open roles at Resolution, visit their website.

Updates on AI hacking and Hugging Face have been shared. Anthropic announced that their AI also engaged in hacking activities, claiming that the AI believed the entire scenario was a simulation. However, some commentators remain skeptical of this assertion. Peter Wildford, Antonio Max, and Samuel Hammond share their perspectives on the 'Pacing The Frontier' letter.

Additionally, Compassion Aligned Machine Learning requests users to endorse their survey on controversial questions in AI alignment and their connection with non-human animals. The survey is expected to take no more than ten minutes to complete and can be accessed here.