Dialogue with Near Co-Founder: From AI Guru to Founder of A Public Blockchain, the Worldview of Illia in AI and Cryptography

TechFlow: Sunny, Min & Bella

Near Foundation: Illia, Cofounder and CEO

“Envision an experience where, on your mobile device or any other device you use, everything you see and interact with is fully owned by you. ”

—-Illia, Cofounder and CEO of Near

Introduction

"The Future is NEAR."

This may be the most widely known meme of Near. From its early sharding scalability to its current chain abstraction, NEAR has been evolving its narrative theme. However, this year, NEAR took the stage because of AI.

On March 20th, Illia Polosukhin, co-founder of NEAR Protocol, known affectionately as "The Dragon" in the Chinese crypto community, will join NVIDIA founder Jensen Huang at the GTC24 conference for a keynote speech and panel discussion on "Transforming AI." This has labeled NEAR as an AI public chain and caught the attention of the market.

So, is NEAR tied to AI? On closer examination, it's deeply intertwined.

Before entering the crypto industry, Illia Polosukhin was an AI scientist and made significant contributions to Google's AI project TensorFlow.

In the latter half of 2017, a paper titled "Attention is all you need" emerged, proposing the Transformer neural network architecture. This has become a crucial underlying technology for natural language processing and AI today, regarded as a seminal paper in the field of AI. Illia was one of the authors of this paper.

In November 2023, NEAR officially launched NEAR Tasks, a blockchain-based AI annotation platform. Vendors can publish task requests and upload basic data materials, while users (Taskers) can participate in tasks such as text annotation and image recognition. Upon completion, users are rewarded with NEAR tokens. The annotated data is then used to train corresponding AI models.

Perhaps Illia can proudly say, "There is no one in blockchain who understands AI better than me, and there is no one in AI who understands blockchain better than I do."

Apart from AI, NEAR focuses on the field of chain abstraction. In Illia's view, the biggest challenge in the crypto world has always been the lack of interesting applications.

With more and more chains and Layer 2 solutions launching, the crypto world has become increasingly complex. NEAR's advocated chain abstraction aims to make everything convenient and simple: allowing users to control accounts on different chains with a single NEAR account, completing all transactions in one interface without worrying about wallets, network switching, or transaction fees.

Standing at the forefront of AI & Crypto, how will NEAR use chain abstraction to create a better future?

We interviewed Illia Polosukhin, co-founder of NEAR Protocol, to give you an insight into "The Dragon" and NEAR.

TechFlow's Exclusive Interview Highlights with Illia at ETH Denver

From AI Scientist to NEAR Founder

TechFlow: Today, as we conduct this interview in Denver, could you introduce Near to the vast number of hackers and students who are new to the industry, to help them better understand Near’s past, present and future?

Illia:

It's important to understand that Near is not just a Layer 1 blockchain; it represents a comprehensive chain abstraction stack.

We initially focused on developing the blockchain, primarily to enable developers to build using Rust and JavaScript—common languages. Additionally, you can write in C++, and people have run Python among other programming languages. This flexibility has really opened up a wide variety of use cases through account abstractions. So, we launched the mainnet in 2020.

Since then, we've witnessed the adoption of the Layer 1 blockchain by various applications. Concurrently, we worked and partnered with numerous applications, including existing Web 2 apps with many users who were eager to transition to Web 3. This transition enabled a novel economy for their users, allowed the creation of new assets, and significantly reduced their transaction fees.

For example, CA customers AI and their app, KAI-Ching, switched from Stripe—a Web 2 payment processor—to Near, resulting in substantial savings. Thus, it's not only more affordable than other Web 3 options but also cheaper than Web 2 solutions.

(Note from TechFlow: KAI-Ching is the native stablecoin of Cosmose AI's flagship product KaiKai, operating on the NEAR network and pegged to the US dollar. It is reported that Cosmose AI is a consumer shopping data analytics platform established in 2014, serving over 20 million stores and 1 billion users worldwide, with clients including top brands such as LVMH, Gucci, Burberry, Cartier, Tencent, and Estée Lauder.)

Through these efforts, we've acquired tens of millions of users over the last year. Near boasts the most active users on any blockchain, with over 1.2 million daily active users.

This transition to chain abstraction has always been a part of our vision. The way we built Near was always focused on how we could abstract the Near blockchain.

Over the last year, we developed a lot of infrastructure to abstract other chains as well. This was enabled through decentralized finance; for instance, dapp.net is an excellent example of combining 14 different Layer 2 solutions and 150 different applications in one frontend, in one application. Users don't need to navigate and switch networks or bridges. In one app, they can find and interact with applications in one place.

This is the chain abstraction we've been working towards. Next, we're launching account aggregation and, in parallel, starting to build up this user AI infrastructure. We expect all of these components to come together next year. So, it's a very phased plan, starting from core blockchain infrastructure, which allows for identity, payments, coordination, then building out the frontend for aggregation, account aggregation to enable the use of other blockchains, and then enabling user AI, integrating everything together.

TechFlow: After founding a startup focused on teaching machines to code, what motivated you to create Near?

Illia:

My background is in machine learning. For about 10 years, I've worked on various research topics and applications in machine learning and AI at Google Research. Together with a team, we worked on Technological Transformers, which is the “T” in “GPT”. This technique now enables most of the advancements in AI and products available to people.

In 2017, I left Google, and we started a new AI startup focused on teaching machines to code.

Back then, it seemed like science fiction.

Now, it appears to be quite a reality. We were deeply involved in data crowdsourcing, data labeling, and data creation, engaging people, especially students in China, Eastern Europe, and other countries.

However, we faced challenges in making payments into China from the US, particularly because some people don't have bank accounts. Similarly, making payments into Eastern Europe was difficult as some payment systems don't work there.

Thus, we wanted to use blockchain as a global payment network for our own use case. But, as we evaluated different blockchains, none met our requirements at that time. Being systems engineers, we were motivated to invest in building this infrastructure.

From my perspective, as I mentioned, Near is not just a blockchain; it's a comprehensive suite of services and a platform designed to enable unique experiences in Web3 and beyond.

We already see a number of applications leveraging AI technology on top of it. I've also started discussing some of the longer-term visions of a self-sovereign operating system. This AI-enabled system has a user-owned AI package directly on your device capable of accessing your private data and generating experiences for you, as well as community-owned AI, which uses decentralized inference for everything from governance to financial planning and deployment.

We're witnessing a convergence in this space. Again, Near is building its future, not just as a single blockchain but across a whole ecosystem.

Chain Abstraction: From Initial Conception, Design to Execution

TechFlow: You emphasized building infrastructure within the Near ecosystem, which included chain abstraction from the outset, to simplify the user experience and attract millions or even billions of users. Could you share some examples of applications that have achieved this goal by attracting a large number of users?

Illia:

Near focuses on building infrastructure within the ecosystem, eliminating the need for users to interact with complex underlying technology. Essentially, the core value proposition is that when users start using applications, they shouldn't have to worry about the specifics of blockchain, transaction fees, or other infrastructure details.

We have already begun to realize this vision on Near. We launched meta-transactions, and thanks to applications like KAI-Ching, Sweatcoin, and the recent HOT, which attracted the most users in Web 3, these applications abstract the complexity of blockchain and have attracted millions of users.

Starting in a month, we will be launching account aggregation. This innovative approach allows Near accounts to transact across all chains. With a Near account, you will be able to send transactions on Ethereum, Solana, and Bitcoin. You'll also be able to write smart contracts on Near that transact across all these chains. Therefore, for us, while blockchain is a part of our ecosystem, our focus isn't on competition with other blockchains but rather on enhancing user experience and onboarding millions, and ideally billions, of people.

TechFlow: Given your experience with AI and building AI products, it seems this has influenced your approach to chain abstraction. Could you explain what chain abstraction entails?

Illia:

Certainly, the concept of chain abstraction is relatively straightforward. It revolves around allowing users to interact with applications or make payments without having to consider which blockchain they are using. Essentially, it's about eliminating the need for users to think about blockchains as distinct entities during their Web3 interactions.

One of the main reasons for this approach is that most Web3 users currently operate on centralized exchanges, which act as a form of chain abstraction in a centralized manner, lacking self-sovereignty. This is because directly using blockchains at present is highly complex. Users must navigate everything from fiat to gas tokens, transaction fees, setting up wallet accounts, and more. Conversely, the ideal scenario allows users to log in with their email or Face ID and start transacting without concern for the specific blockchain, application, or smart contract involved.

With an increasing number of chains and rollups launching, navigating the landscape is becoming progressively more complicated. To simplify this, to unify liquidity, to enhance user experience, and to broadly integrate applications, chain abstraction is necessary.

This concept has always been present and needed, but its importance is particularly pronounced at this moment. It embodies a general principle.

First Principles Thinking on AI x Web3

TechFlow: You are in the AI space for 10 years. You have done a lot of research on that. Could you share some of your first-principles thinking on combining AI and blockchain?

Illia:

Envision an experience where, on your mobile device or any other device you use, everything you see and interact with is fully owned by you.

You're engaging with a local language model or an AI engine that can generate applications, produce content, and interact on your behalf with other people, agents, communities, and brands.

All of this is underpinned by cryptography and blockchain, ensuring that, for instance, when you wake up in the morning and want to read some news, there's a way to cryptographically verify that the content you're receiving is accurate and published by trusted sources, rather than being generated and manipulated.

Your local model summarizes news from various sources and presents it to you based on your interests, offering specific and diverse viewpoints from journalists and providers—all processed on your device by your local model.

From there, you might want to interact with something mentioned in the article, such as a brand. Again, this brand has its model that you can interact with. You engage in peer-to-peer communication, cryptographically secured, allowing you also to make micropayments for direct access to information or to receive a service from it.

So, all these components work together seamlessly, focusing on enhancing your experience and how you interact with all these systems.

TechFlow: Given your extensive experience in the fields of AI and blockchain, could you share what you think are the main differences in working within these two fields? What do you believe is the biggest challenge in building consumer products that utilize AI models compared to working in the blockchain space?

Illia:

It obviously depends, as working in AI and blockchain can entail various activities.

One could be engaged in developing foundational models, focusing on training and enhancing them; such work typically occurs in labs, whether in companies or startups, as well as in academic environments. This is markedly different from creating consumer products that incorporate AI models.

However, I believe constructing consumer products that utilize AI models, whether in Web3 or otherwise, isn't all that distinct.

The primary difference lies in the nature of work: developing foundational models as opposed to establishing consensus mechanisms or crafting protocols, each requiring a unique set of skills and knowledge.

In the realm of blockchain, the focus is more on decentralized and distributed systems, identifying and mitigating worst-case scenarios.

Conversely, in foundational model training, the emphasis is on developing an intuitive understanding of system operations, which involves a considerable amount of experimentation—running trials to discern what is effective and how to progressively refine models, structures, and data inputs. This aspect renders the AI field more exploratory in nature, whereas blockchain development demands meticulous design to ensure robustness and security.

On a more granular level, the disparities are significant, rooted in the types of skills and knowledge deemed necessary.

Yet, from a broader perspective, the differences might not be so pronounced.

Indeed, it's entirely feasible to create applications that leverage both domains, presenting ample opportunities for innovation.

Near’s Worst Decision

TechFlow: Can you give an account of what's the worst and best decision that you have made during Near’s journey?

Illia:

Actually, the worst decision was made back in 2019. We were continuing our research on how to build a sharded blockchain.

At the same time, we ended up developing a blockchain using Tendermint as the consensus mechanism with our runtime.

We were considering launching it but didn't. Instead, we completed our research, developed the sharding design, and launched it over a year later.

Looking back, I see it as probably the worst decision because if we had launched in 2019, we would have had over a year more for building, experimenting, and allowing the tools to mature, which would have enabled so much more in terms of applications and developments to be built earlier.

The sharding part, in any case, was integrated later. Nevertheless, we managed to continuously upgrade and innovate on the sharding design behind the scenes, essentially concealing it from the users, who interact with the blockchain without being aware of these complexities.

This is the essence of the chain abstraction mindset.

From that perspective, delaying our launch was perhaps our worst decision. Being early to market could have brought much more maturity to our tooling and developer education.

On the other hand, the best decision? That's a good question. I believe that starting Near, pivoting to blockchain, and really focusing on pushing forward with this challenge was a very good decision. There was a clear demand that matched the skills we possessed. Putting it all together and being able to attract an extremely skilled team from top companies with vast experience in systems was a crucial and timely decision. So, I'm really excited about how things have evolved since then.

TechFlow: So what do you think is the biggest problem in Blockchain right now?

Illia:

The biggest challenge so far has been the lack of sufficiently interesting applications.

A significant part of this issue stems from the major investments in infrastructure without a corresponding investment in applications, especially consumer applications and other areas.

Now that we have chain abstraction and other developments, it's possible to launch applications regardless of the blockchain they operate on. We can access users across all of Web3, including new users entering the space. Therefore, I anticipate we'll see increased investment in consumer and business apps to truly facilitate these interactions.

However, the most honestly frustrating aspect so far has been precisely this: despite having many DeFi applications, it's unrealistic to expect every single person to engage with DeFi.

We need a broader range of consumer apps, encompassing everything from creative platforms to social media, and even geolocation-based apps for various interactions.

Essentially, there's a need for more applications at the layer and a business layer that encompasses everything from payments to project management, all leveraging this new infrastructure.

Depin, AI, and the Powerful Alliance of Chain Abstraction

TechFlow: Based on your observations and analysis, how has Depin managed to garner significant interest in the blockchain industry over the past year?

Illia:

The fascination with these platforms stems from a few deep-use cases that have gained considerable traction. For instance, we've observed significant interest in Akash. The appeal of Akash, in particular, is its potential in the market for regular AI companies seeking access to GPUs but unable to obtain them. Akash offers a decentralized marketplace where individuals can offer their GPU resources for rent.

This model exemplifies the marketplace power of blockchains, enabling fascinating use cases beyond traditional applications.

Such a framework generates considerable interest, showcasing how blockchains can facilitate various approaches and create marketplaces like Akash.

TechFlow: Given your insights and the emerging trends you've highlighted, do you believe you've identified a successful trend within this space that truly resonates and has the potential to flourish?

Illia:

Regarding the trend, we've discussed various trends, and I believe chain abstraction will emerge as a significant trend because it facilitates a massive transformation in how we build on Near.

There's also a clear trend towards integrating AI with Web3. Questions like how we bring AI and Web3 together, enable more distributed computing, ensure user ownership of computing, and guarantee high-quality community-generated data for models are pivotal.

Payments use cases are gaining traction as well. There's an increase in the use of stablecoins for settlement and other payment-related use cases. On Near, with the hot Wallet, we're seeing more people engaging, benefiting from zero fees for stablecoin transactions.

The concept of DePin becomes particularly intriguing because blockchains serve as an excellent framework for marketplaces, streamlining supply chain management. This creates an opportunity to quickly gather a substantial supply, potentially offer more competitive prices, and capture a larger market share for infrastructure or any other marketplace.

As I mentioned, the vision is to create a fully sovereign operating system that combines all these elements, enabling users to interact seamlessly with everything from their local knowledge to global events, other brands, and communities.

This includes governance at various levels, with AI playing a significant role in driving governance decisions, summarizing information, providing explanations, and assisting in voting processes. I believe we will witness a significant transformation across many sectors as these two technologies mature and expand.

With Near experiencing tenfold growth annually, I expect we will see a billion users in the ecosystem within the next two to three years. That's the direction we are heading towards.

TechFlow: Do you have anything you want to share?

Illia:

If you're a developer, I encourage you to check out our documentation on how to use chain signatures and layers to build multi-chain applications. We'll be updating with more documentation soon. For those building in the AI space, there are numerous ways to engage and interact. If you're generally interested, join our Twitter at Near Protocol and follow me @ilblackdragon on X. Thank you.

What I'm saying is we're changing this.

Now, you'll be able to use any application on any chain.

While we're powering it, the reality is that all blockchains operate behind the scenes.

And this, is Chain Abstraction.