Decentralised AI is a hot buzzword at this year’s TOKEN2049. We touch on the various panels that delve into the topic.
With artificial intelligence (AI) becoming the latest buzzword in the tech space, it was just a matter of time before AI would be introduced into the Web3 space. Decentralised AI is now the next trend in Web3, and it could potentially accelerate the development of new Web3 technologies and services.
We dive into some of the discussions around Decentralised AI at TOKEN2049.
Decentralised AI: The Power of Permissionless Intelligence
- Emad Mostaque, Founder, Schelling AI
- Sean Ren, Co-Founder and CEO, Sahara AI
- Tarun Chitra, Founder and CEO, Gauntlet
- Alex Skidanov, Co-Founder, NEAR
- Santiago R Santos, Managing Partner, SRS
With the ability to learn and derive actions from large blockchain datasets, artificial intelligence (AI) is fast becoming the next technological buzzword in the Web3 space. All eyes were on the panel of leaders in AI development as they took to the stage to debate the power of decentralised AI at TOKEN2049.
Touching on one of the central themes, control and monetisation of data, Mostaque pointed that the current model involves users unwittingly giving away vast amounts of personal data in exchange for using AI-powered services. The panelists agreed that decentralisation could offer a way for users to regain control over their data, but emphasised the need for clear and viable pathways to achieve this in practice. Blockchain, for instance, could provide a mechanism for users to monetise their data while maintaining privacy, a sharp contrast to how centralised AI providers currently operate.
Ren explored the challenges that AI developers face when it comes to accessing resources such as data and computational power. Prototyping an AI model is often time-consuming process, and decentralised platforms could fast-track access to financial and computational resources, shortening the time it takes to bring ideas from conception to implementation.
AI Cost and Efficiency
Despite the optimism around decentralisation, Chitra and Skidanov were pragmatic about the current state of the technology. While decentralisation offers ideological advantages like privacy and open access, many users prioritise cost and efficiency over these ideals.
AI, particularly at the scale needed for cutting-edge applications, also remains prohibitively expensive. Citing examples like the $3 billion supercomputer used by OpenAI, he argued that, at present, centralised solutions still have a distinct advantage in terms of cost and performance. However, using reinforcement learning models and making different trade-offs between latency and bandwidth could create new opportunities.
Unique challenges of decentralised AI in industries like scientific research and healthcare. Chitra and Mostaque noted that while commercial AI applications receive significant funding and resource allocation, less profitable but equally important areas, such as medical research, struggle to secure the necessary computational power. Mostaque mentioned that decentralising AI development could allow for a more democratic distribution of resources, potentially enabling breakthroughs in fields like cancer research and personalised education.
Ownership and Governance
One particularly compelling aspect of the conversation was the discussion of ownership and governance in decentralised AI. Ren highlighted the potential for new economic models where stakeholders—whether they contribute data, computational resources, or development time—could own a portion of the AI models they help create. This system of shared ownership could enable more collaborative and equitable AI development, particularly in contrast to the current model where large corporations reap the majority of the rewards from AI advancements.
The panelists also considered the darker side of permissionless AI, with Santos and Mostaque debating whether decentralised AI could pose existential risks. While some fear that AI could eventually be used for malicious purposes, the panelists were more concerned about the governance and control of these technologies. Santos argued that open, decentralised systems could be more resilient to certain types of attacks than centralised ones, which present single points of failure. Mostaque echoed this sentiment, suggesting that allowing more actors access to AI tools could create a balance between good and bad actors, reducing the likelihood that any one group would be able to monopolise the technology for harmful purposes.
In summary, the session provided a nuanced exploration of the opportunities and challenges of decentralising AI. While the technology and infrastructure required to fully realise permissionless AI are still developing, the potential benefits—ranging from enhanced privacy and data ownership to increased innovation and resource access—are clear. However, the panelists agreed that a balance must be struck between decentralisation and the need for efficient, scalable AI solutions, particularly as AI continues to advance into increasingly critical areas of society.
AI, Crypto, and the Future: Competing for Mindshare in an Exponential Age
- Dovey Wan, Founder, Primitive Ventures
- Diogo Mónica, General Partner, Haun Ventures
- Soona Amhaz, Managing Partner, Volt Capita
- Eric Wall, Co-Founder, Taproot Wizards
- Larry Cermak, CEO, The Block
In this second discussion on artificial intelligence and blockchain technology, the discussion amongst the panelists revolved around the challenge of incentivising users and traders in the era of open-source AI. Tech giants are racing to advance their algorithmic capabilities without fully democratising access, and open-source models are sometimes opaque on the backend functionalities of their AI learning models.
Within the realm of AI and crypto, Oracles are injecting information on-chain while AI-powered models are being used to resolve discrepancies efficiently. Large language models (LLMs) are now being leveraged to inject data on-chain, further blurring the lines between AI and blockchain. One relatively new area of this intersection are decentralised models and Initial Model Offerings (IMOs), but the panel notes that its knowledge and adoption is still slow on the uptake.
Looking ahead, the intersection of AI and crypto adoption could spur massive investments and perhaps a new kind of peace in the current atmosphere of fractured international trust. Still, the panelists stressed that the extent of its impact, and the rise of artificial general intelligence, leaves the future uncertain.
Read our top panel picks of TOKEN2049 2024
Learn how Web3 and crypto will be like in the next three years
