Google is working on a project called ‘Astra’ that aims to develop AI agents capable of engaging in open-ended dialogue and assisting with complex tasks by 2024. These AI agents will be trained using a new approach called ‘Constitutional AI’ that aims to instill the agents with a sense of ethics and values. The project is being led by Google’s AI research team and involves collaboration with external experts in fields like ethics and philosophy. The goal is to create AI agents that can engage in open-ended dialogue, assist with complex tasks, and adhere to ethical principles. Key challenges include ensuring the agents are safe, reliable, and aligned with human values. If successful, Project Astra could lead to significant advancements in AI capabilities and human-AI interaction.
Source: https://www.businessinsider.com/google-future-ai-agents-project-astra-2024-5