AI is our generation’s moonshot, and we’ve got to get it right. That’s what Stanford University Professor Fei-Fei Li told Joe Biden this month as he became the first president to talk about the impact of AI in his State of the Union address. Our original moonshot required new technology and new ways of thinking to propel us forward. While the unimaginable capabilities of the latest AI models capture most of the attention, the most important story yet to be told in AI is the technology that is emerging to make it safe, accountable and beneficial for everyone.
Blockchain and Web3: The Ideal Hosts for Neutral AI
Blockchain and Web3 technologies are uniquely positioned to address the challenges of modern AI applications. By securely anchoring training data, models, and AI operations, Web3 provides guardrails and transparency that are sorely lacking in closed AI systems. This visibility into data origins and decision-making processes is crucial for preventing bias, ensuring fair content use, and governing the capabilities of increasingly powerful AI models.
The key technologies driving Web3 are a natural fit for neutral AI. Blockchains, decentralized compute, and transparent governance each solve for the risk of AI becoming an unaccountable black box that puts us at the mercy of any one player.
With distributed ledgers, blockchains offer an indisputable level of transparency and provenance. No single player owns a truly decentralized blockchain. Independent parties keep their own copies of every transaction, ensuring that data origins are traceable and immutable. This cryptographic security, coupled with consensus mechanisms, guarantees the integrity of data, making blockchains an ideal platform for securing neutral AI applications.
While blockchains back the data security of neutral API, decentralized Web3 compute networks provide the horsepower. Decentralization creates an open market of servers, GPUs, and storage that are available on demand to train and run AI. The code portability required for decentralization creates strong incentives around open source AI frameworks instead of the proprietary toolkits favored by tech giants. This open market also provides a rational basis for anyone who owns compute resources – be it a cloud provider, a startup, a university, or a public consortium – to power neutral AI by removing the economic benefit of hoarding those resources.
The biggest Web3 innovation that will foster neutral AI, though, is transparent governance. This is how we as the users of AI will get to decide how we keep it safe, honest and fair. Transparent governance, as expressed through smart contracts and other forms of verifiable code, provides clear rules and kill switches that align with our consensus as a society about what we want AI to do for us. The ability to get paid when content you create is used by AI can be automatically enforced across any derivative works. Biases and blindspots can be permanently eradicated by enforcement of coding and training data requirements. And all of these rules can be continuously, publicly audited for compliance.
Verified Compute Makes Web3 Right for The Job
Web3 has the power to make AI trustworthy and neutral, and the technology is moving fast to fill this need. Web3 skeptics previously have pointed to scalability, data privacy, and environmental impact as obstacles to adoption. A new generation of blockchains and a vastly expanded Web3 technology stack are closing these gaps. This sets Web3 up to become a key part of the emerging infrastructure for AI, and further advances in Web3 will require investment orders of magnitude smaller than the astronomic cost of buying GPU chips.
Zooming into the Web3 stack even further, verified compute stands out as the capability that unlocks the potential to create neutral AI. Verified compute allows networks of independently owned computers to operate securely and transparently. Whereas blockchains provide the ledger to record the audit trail of a decentralized network, verified compute makes it safe to send the AI training jobs and model inference requests to servers you don’t control. Verification makes it possible to observe the code, inputs and outputs of an AI compute task and provides immutable proof of correctness.
This ability to auditably run any code anywhere also serves as a platform for transparent governance. The complex code needed to independently track training data across innumerable sources, and enforce AI safety can be reliably run on a verified compute network without any one party holding dictatorial control. This approach ensures that AI applications remain unbiased, transparent, and verifiable, fostering trust and integrity in digital ecosystems.
The Marriage of Web3 and AI
It’s no secret that the race is on to realize the upsides of AI across every industry and scientific discipline. As we wrestle with the bigger question of how to lay the groundwork for these profound changes without sowing seeds of destruction, we must move deliberately toward a level playing field. Web3 allows us to pool resources, collaborate on the development of underlying technology, and fairly compensate for the real human work that makes AI possible. When rationally weighed against a model where only a handful of players invest trillions of dollars to “win it all,” Web3-driven AI is the fastest path to the most gain for everyone.