Ex-Google, DeepMind and Meta Alumni Unveil Vision for “The New Meta” for Humans and AI

A coalition of former Google, DeepMind, and Meta technical leads has launched AMMO (Architectures for Massive Multiagent Online, AmmoAI.io), an ambitious initiative to transform the current Internet into an "Human-AI symbiotic Meta" where billions of AI agents and humans coexist, and co-evolve with alignment from collective human feedback.

Rooted in team’s past involvement in OpenAI’s Neural MMO, as well as drawing inspiration from DeepMind’s research on social intelligence in AI systems, the team aims to solve interconnected challenges in multiagent reinforcement learning, ethical alignment, and scalable infrastructure that have long stymied the field.

Tackling the "Problem Problem"

The vast majority of the AI industry is focusing on pursing human-level AI performance (the “Policy Problem”), but AMMO addresses what DeepMind termed the "Problem Problem" – creating adaptive environments that foster large-scale agent learning. The project confronts four critical hurdles:

• Complex world modelling for dynamic environments
• Alignment with human values
• Cross-MAS (Multiagent System) knowledge transfer
• Sustainable economic frameworks

A Three-Pillar Solution

AMMO’s architecture hinges on three technical breakthroughs:

1. Massive Agent Factory: A platform enabling massive personalised AI agents grown from the environments without development from financial management to social collaboration.

2. Composable Embedding Space: An expansion of OpenAI’s Neural MMO framework, re-engineered for real-world applications like financial markets and education .

3. Distributed RL Gyms with Human Feedback: A scalable system using Reinforcement Learning from Human Feedback (RLHF) and Anthropic’s Constitutional AI to ethically evolve the agentic society.

The infrastructure builds on enhanced Neural MMO architectures from NeurIPS 2023, Deepmind’s multiplayer topology research from Nature 2020, and integrates open-source projects like Camel-AI for agent coordination. Blockchain technology will be leveraged to enable efficient cross-agent interaction, transparent governance, and to incentivize meaningful participation from both humans and AI. Notably, it envisions a world from Iain M. Banks’ Culture series, prioritizing collective welfare over competitive optimization. 

Industry Veterans at the Helm

AMMO’s team includes industry veterans of 10-year Googler, Deepmind researcher, Meta tech lead, VP of an AI unicorn, ACM-ICPC World Champion as well as experienced web3 advisors. Backed by investors such as Samsung NEXT (others to be announced), the group plans to release:

• A technical whitepaper detailing their architectural vision
• More modular subspaces such as investment, career, education, and social sectors
• A blockchain-based economic model scaling to billions of AI agents

They’ve officially launched their first sandbox experiment at FakersAI.com, offers a glimpse into AI-human cohabitation systems. As the project unfolds, it could redefine how autonomous agents evolve alongside – rather than in isolation from – human societies. To learn more about their vision and explore ongoing developments, visit official website at AmmoAI.io.

This post is commissioned by Ammo AI and does not serve as a testimonial or endorsement by The Block. This post is for informational purposes only and should not be relied upon as a basis for investment, tax, legal or other advice. You should conduct your own research and consult independent counsel and advisors on the matters discussed within this post. Past performance of any asset is not indicative of future results.