Bangdao Chen, DPhil
Rich Sutton, PhD
Reinforcement Learning Professor, University of Alberta
Bill Roscoe, DPhil
Ramesh Ramadoss, PhD
John deVadoss
Ben Schiller
Haseeb Qureshi
Jake Brukhman
Jasper De Maere
Lex Sokolin
Matt Stephenson
Speakers Agenda (Oct 31st, 2024)
2:00-2:10
PM
Opening Remarks
Bangdao Chen, DPhil
Speaker
Co-Founder, University College Oxford Blockchain Research Centre
Ramesh Ramadoss, PhD
Speaker
Chair, IEEE Blockchain Technical Community
Richard Sutton explored the profound implications of super-intelligent AI at the event. His talk, “Intelligence, Cooperation, and Human Flourishing,” emphasized the importance of decentralized cooperation among AI agents. Sutton stressed that while humans are naturally cooperative, our societies often struggle with corruption and conflict. He advocated for AI systems designed to promote collaboration leading to humanity flourishing. “Reinforcement learning, rather than large language models, holds the key to advancing AI,” Sutton argued. He envisioned a future where AI agents improve economies and society but warned against centralized control, which could hinder innovation and cooperation.
Bill Roscoe delivered a compelling presentation on the need for trust in digital systems. His talk, “Blockchain and Verification – Trusting AI,” emphasized blockchain’s role as a “deus ex machina” for building decentralized, trustworthy systems. Roscoe discussed the challenges of creating secure voting machines and reliable AI systems, arguing that trust must be built through robust verification methods and cryptographic assurances. “The world is becoming increasingly partisan, and trust is eroding,” Roscoe said. “Blockchain provides a way to build integrity into our digital civilization, but only if we educate society about its importance.”
Ramesh Ramadoss followed with an exploration of how blockchain and AI can be integrated to enable autonomy for “Smart Agents”. He traced blockchain’s evolution from Bitcoin to current blockchain architectures, emphasizing their current challenges and potential to serve as the economic network for the agentic internet . He emphasized that on-chain learning agents could play a significant role in the new agentic internet, including human-to-agent transactions, agent-to-agent transactions, and multi-agent economy.
John deVadoss, co-founder of the InterWork Alliance, presented “Securing AI – 7 Things to Note,” highlighting the vulnerabilities of generative AI models. He underscored the risks of tainted data, malicious prompt engineering, and embedding inversions. deVadoss emphasized that without verifiable security protocols, AI could be manipulated in ways that threaten both individuals and organizations. “Generative AI’s guardrails are flawed and easy to bypass,” deVadoss warned. He advocated for a proactive approach to AI security, using advanced encryption and blockchain verification to protect sensitive data and ensure the integrity of AI systems.
Bangdao Chen delivered a forward-thinking presentation titled “Securing Digital Civilization – ID and Wallet for AI Agents.” He introduced the concept of a decentralized and verifiable identity system, bound by time and built on blockchain, to ensure that AI agents operate securely and in alignment with their creators’ intentions and identities. Chen emphasized the need for a new autonomous regulatory framework that balances autonomy with accountability in the digital world. “Digital civilization can only thrive if we establish systems that are both secure and transparent,” Chen explained. He illustrated how blockchain-based wallets and Human-Interactive Security Protocols could safeguard payments while facilitating seamless AI-to-AI and AI-to-human interactions.