AI breaks Web3 ─ Lack of GPU drives Web3 to the mainstream | CoinDesk JAPAN | CoinDesk Japan

A shortage of GPUs will drive the mainstream adoption of Web3 infrastructure.

AI (artificial intelligence) training requires a high-performance GPU (graphic processing unit) such as the NVIDIA A100 or H100. But GPUs are only needed for a short period of time, and GPUs are very expensive and in limited supply. This makes GPUs out of reach for many startups.

In addition, many worry about the damage deepfake images will cause to people and the economy, and have serious concerns about the safety of AI, such as the fear of evil AGIs (general purpose AI) like the Terminator that seek to destroy society. is holding AGI (Artificial General Intelligence) means the realization of an autonomous system called “Singularity” that surpasses human intelligence.

Access and security with DePIN

Decentralized Physical Infrastructure Networks (DePINs), specifically physical networks of computing and storage, offer a solution to these concerns, pushing AI and Web3 into the mainstream for two reasons. do. access and safety.

Despite being nearly a decade old, Web3’s use is limited to a small number of crypto advocates. This fact puzzles those who expected Web3 to spread faster than the World Wide Web. And while the original web protocol had to start from scratch, the decentralized Web3 protocol could take advantage of the existing web.

But mainstream users are not yet participating in decentralized social networks or using protocols like decentralized file storage. Because there is no need to be so strong.

Despite the widespread perception that Web2 platforms are heavily monitored and manipulated, for many, those services work “just fine” and strong network effects keep users stuck. ing.

AI Breaks Web3

The rise of AI will trigger a big break for Web3 infrastructure. The Web3 protocol offers a solution to the GPU shortage and data-related challenges faced by AI startups. DePIN will transform AI development and, in the process, bring the Web3 protocol to the first mainstream users.

AI startup founders have the technical ability and motivation to overcome the hurdles that currently prevent mainstream users from using them, such as having to purchase tokens through a poor user experience (UX). ing. Founders need high-quality, expensive, and hard-to-find GPUs to succeed.

Protocols such as Akash, also known as “Airbnb for GPUs,” provide a decentralized marketplace for GPU owners to rent out their GPUs. Such protocols democratize AI innovation by making high-cost resources accessible to smaller players, allowing hardware owners to receive unearned income.

AI startup founders came to the Web3 protocol to get access to the GPUs they needed, but those protocols enabled new capabilities and the founders kept going, and Web3 It will effectively improve competitiveness.

File storage protocols like Arweave disrupt data oligopoly by offering a one-time payment model for persistent data storage usage. Using persistent storage for AI training data makes machine learning open and verifiable, increasing trust in AI models.

Build trust with open source

Currently, AI products from major companies such as OpenAI’s ChatGPT and Google’s Bard are leading the way, but open source versions of AI may have a competitive edge.

AI can tell the difference between two round items like an apple and a baseball because it has been taught thousands of times by human input. We all use boxes containing bicycles, traffic lights, bridges, etc. when filling out a CAPTCHA (Completely Automated Public Turing test to tell Computers and Humans Apart) By choosing , you are unknowingly creating training data for your AI model.

If such training data is controlled exclusively by a company, can we trust it?

AI use cases are unique in that open, published training data is not only more reliable, but also of higher quality. Open models can exchange information and accelerate learning, creating a virtuous circle that makes the next version even better.

Additionally, open-source frameworks like zkML (Zero-Knowledge Machine Learning), a research area pursued by Sam Altman’s Worldcoin project, allow AI models to be trained on specific datasets. Provide proof by encryption technology to prove that, and increase the reliability of the output.

While some fear that AI will lead to a dystopian authoritarian future, DePIN provides a defense against both trusted code and the physical infrastructure network that runs it.

gain a competitive edge

If startup founders develop apps on infrastructure managed by internet giants like Amazon Web Services (AWS), they will be paying the majority of the revenue to their competitors, making it harder to compete. Not only that, but it also exposes you to counterparty risk.

If Web3 apps succeed, they may be seen as competitors by Amazon and Google, upon whom they depend, and cut off from the infrastructure they need. DePIN provides founders with a more reliable, attack-resistant and much cheaper infrastructure.

DePIN’s benefits for AI startups extend beyond access to training data and hardware.

For example, deepfakes such as the “Pentagon Explosion” photo, which temporarily caused a $500 billion market reaction (approximately 70 trillion yen, equivalent to 140 yen to the dollar), have become increasingly common and the credibility of online content has increased. The Web3 protocol provides a public-key cryptography solution to the loss of

While the benefits of cryptographic signatures seem obvious to us in the crypto space, AI pessimists still don’t understand the solution. Public-key cryptography provides a unique, unforgeable signature that can be used to prove provenance and assess the authenticity of content.

A positive result of insufficient GPU

GPU shortages may eventually motivate startup founders to use DePIN protocols, and the capabilities unlocked by these protocols will allow apps developed on such protocols to beat Web2’s existing apps. support that.

AI apps developed using decentralized infrastructure will have optimized job routing to reduce latency, faster and cheaper commerce through micropayments, all kinds of smart contract functionality, deepfakes and spam feeds. It offers benefits such as automatic filtering of data to exclude it from

DePIN provides AI startups with better access and features at a lower cost, providing fertile ground for development and competition. Users of these emerging AI applications will become users of the Web3 protocol, often without realizing it.

This behind-the-scenes integration provides a seamless experience that introduces users to the benefits of the decentralized web and helps overcome traditional hurdles to Web3 adoption. AI and Web3 are not competing, they need each other.

Amy James and Devon James: Co-Executive Director of the Web3 Working Group.

｜Translation and editing: Akiko Yamaguchi, Takayuki Masuda
| Image: FritzchensFritz/Creative Commons (edited by CoinDesk)
｜Original: Web3 Can Actually Compete in the Computer Chip Race