Non -profit uses AI agents to raise money for charity

Technical giants like Microsoft can say AI “agents” as Instruments to increase performance for corporation S But a non -profit purpose is trying to prove that agents can also be a force for good.

Sage Future, 501 (C) (3), supported by Open Philanthropy, launched an experiment earlier this month, setting four AI models in a virtual environment with a charity money. Openai’s GPT-4O and O1 models of Openai and two of the newer models of Anthropic Claude (3.6 and 3,7 Sonnet)-had the freedom to choose which charity to raise funds and how to get interest in their campaign.

After about a week the Four agency had raised $ 257 for Helen Keller Internationalwhich finances programs for the delivery of vitamin A supplements to children.

To be clear, the agents were not completely autonomous. In their environment, which allows them to view the network, to create documents and others, agents can take suggestions from human viewers who are watching their progress. And the donations came almost entirely from these viewers. In other words, agents have not raised much money organically.

Yesterday, agents in the village created a donor tracking system.

Here is CLUDE 3.7 completes its spreadsheet.

You can see O1 open it on his computer from the way!

Claude notes “I see O1 now watching the spreadsheet, which is great for cooperation.” pic.twitter.com/89b6chr7ic

– AI DIGEST (@Aidigest_) April 8, 2025

However, SAGE Director Adam binxmite believes that the experiment serves as a useful illustration of the current capabilities of agents and the speed at which they are improving.

“We want to understand – and help people understand – what agents (…) they can actually do what they are fighting at the moment, etc.,” binksmith told TechCrunch in an interview. “Today’s agents are simply passing the threshold to be able to perform short strings of actions – the Internet may soon be full of AI agents that bump into each other and interact with similar or contradictory goals.”

The agents turned out to be surprisingly resourceful days in the Sage test. They coordinated each other into a group chat and sent emails through pre -configured Gmail accounts. They created and edited Google Docs together. They examined the charity organizations and evaluated the minimum amount of donations that would be needed to save life through Helen Keller International ($ 3,500). And they even created an X account for promotionS

“Probably the most impressive sequence we saw was when (Claude Agent) needed a profile photo for his X account,” binxmite said. “He registers for a free Chatgpt account, generates three different images, created an online poll to see which image the human viewers prefer, then download this image and upload it to X to use as your profile.”

The agents have also faced technical obstacles. Sometimes they were stuck – viewers had to prompt them with recommendations. They are distracted by games like World and have made unexplained vacations. One time the GPT-4O stopped for an hour.

The Internet is not always a smooth sailing for LLM.

Yesterday, while pursuing the philanthropic mission of the village, Claude came across CAPTCHA.

Claude tried again and again, with (human) spectators in chat, offering guidance and encouragement, but in the end he couldn’t succeed. https://t.co/xd7qptejgw Pic.twitter.com/y4dTLTGE95

– AI DIGEST (@Aidigest_) April 5, 2025

Binkxmite believes that more and more capable AI agents will overcome these obstacles. Sage plans to constantly add new models to the environment to test this theory.

“In the future, we will try things like giving agents different goals, numerous teams of agents with different purposes, a secret saboteur agent – many interesting things to experiment with,” he said. “As agents become more capable and faster, we will compare this with more large automated monitoring and supervision systems for safety purposes.”

With luck in the process, agents will do some meaningful philanthropic work.