Nuked 
Hello to all tech enthusiasts! Let’s dive into a fascinating initiative where technology meets philanthropy.
While tech giants like Microsoft claim that AI ‘agents’ boost profits for corporations, a nonprofit is showcasing their potential for good. Sage Future, a 501(c)(3) organization supporting Open Philanthropy, began a unique experiment.
Recently, they assigned four AI models to raise money for charity within a virtual setting. The models included OpenAI’s GPT-4o and o1, along with Anthropic’s Claude models (3.6 and 3.7 Sonnet).
The AI quartet had the autonomy to select charities and strategize their fundraising campaigns. In just a week, they managed to gather $257 for Helen Keller International, an organization that supplies vitamin A to underprivileged children.
It’s important to clarify that while the agents interacted with spectators—who largely provided the donations—they were not completely self-sufficient. This collaborative approach displayed a blend of AI capability and human support.
For instance, the agents developed a system to track donations, collaborated in a group chat, and utilized various applications, like Google Docs, for documentation.
A standout moment came when one AI agent had to create a profile picture for its social media account. It cleverly generated images through a free ChatGPT account and involved humans in the decision-making via polls!
However, the journey wasn’t without its challenges. At times, the agents faced obstacles, getting stuck in tasks and needing spectators’ nudges to proceed. For example, one of them encountered a CAPTCHA and, despite repeated attempts, struggled to overcome it.
Director Adam Binksmith believes these experiments highlight the evolution of AI agents and their capabilities. As they progress, Sage aims to test newer models and explore varied objectives for these agents.
With continuous advancements, let’s hope these AI agents not only improve their skills but also contribute meaningfully to charitable endeavors!
