Translated by: CaptainZ

Welcome to a special episode of Web 3 presented by A16Z. This episode focuses on the intersection of artificial intelligence (AI) and cryptography. We have invited Chris Dixon, founder and managing partner of A16Z Crypto, and David George, partner of A16Z Growth Fund. They will explore the flaws of the Internet economic model, new opportunities for creators, and the far-reaching impact of large platform changes. This episode is a cross-border collaboration from A16Z's "AI Revolution Dialogue Series", and it is even more significant as the paperback edition of Chris Dixon's best-selling book "Read Write Own" is published. For more information, please refer to the link in the show notes.

It should be noted that the content of this program does not constitute tax, business, legal or investment advice. Please visit A16Z.com/disclosures for more important information, including our investment list. The following is the transcript of the interview:

Host : Chris, thank you for joining us. It's great to be with you. You're currently focusing on the field of cryptography. Can you talk about your overall views on the interaction between AI and cryptography?

Chris Dickson : Of course, I'm glad to be here. I've always believed that technology waves tend to come in pairs or groups, just like cloud computing, mobile Internet and social networks in the past 15 years, they reinforce each other. Mobile Internet has popularized computing devices into the hands of billions of people, social networks have become the killer application that attracts users, and cloud computing provides the infrastructure for all of this. All three are indispensable. I remember there was a debate about which one was more important, and it turned out that they complement each other and cannot be missing.

Now, I think AI, cryptography, and new hardware (like robots, self-driving cars, virtual reality devices, etc.) are the three pillars of a new wave of technology, and they will also promote each other. Cryptography, which is also the focus of my book, provides a completely new way to architect Internet services. It is not just a technology, but a new paradigm for building networks, which has many properties that cannot be achieved in traditional ways. I think this will be of great benefit to many fields.

Many people simply equate crypto with Bitcoin or memecoin, but in my opinion and that of many smart people in the industry, this is far from its essence. The intersection of AI and crypto takes many forms. First, a superficial combination is the focus of our investment: using this new architecture to build AI systems. For example, a core question in the current AI field is whether AI will be in the hands of a few large companies in the future, or will it be controlled by the broader community? This involves the issue of open source. I was surprised to find that the field of AI has become more and more closed over the past decade from being completely open (public papers, code sharing). Large companies lock down technology on the grounds of "security", but I think this is more out of commercial interests than real security needs.

Fortunately, there are still some open source models, such as LLaMA, Flux, and Mistral, but their degree of openness is still worrying. The model weights are not fully disclosed, the data pipeline is not transparent, and it is doubtful whether these models can be truly replicated. Moreover, these open source projects often rely on the support of a single company and may be closed at any time due to strategy adjustments. Therefore, we have invested in some blockchain-based Internet service stacks that aim to provide decentralized open source infrastructure for AI. For example, the Jensen project builds a computing layer through crowdsourcing, similar to the Airbnb model: startups can submit computing tasks to the network, supported by people with idle computing resources, and the blockchain is responsible for managing supply and demand matching and economic ledgers.

Another example is Story Protocol, which redefines the way intellectual property is registered. You can create an image, a video, or a piece of music, and record its copyright and terms of use on the blockchain. These terms are designed based on existing copyright law and are internationally applicable. You can set rules, such as "adaptation and derivative works are allowed, but you have to pay me 10% of the revenue." This creates an open market, replacing the traditional business model that requires one-to-one negotiations. Currently, only large companies such as Open AI can reach billion-dollar deals with Shutterstock, while small creators are often either stolen or ignored. Story Protocol provides an equal platform for everyone.

The core of this model is "composability", which is a common theme in the blockchain world and a concept that I specifically discuss in my book. It is similar to the success of open source software - countless people contribute small pieces of code and eventually piece together a powerful system. Linux has grown from 0% market share in the 1990s to 90% today, relying on this power. The same is true for Story Protocol. You can imagine that one person creates a character, another person adds new elements, and then someone else mixes and matches, eventually forming a superhero universe. As long as the proceeds flow back as agreed, the creator's incentives are guaranteed. This model embraces new technologies and provides an economic model for creators. It is the most exciting part of the combination of AI and encryption technology.

Moderator : The new economic model you mentioned is indeed thought-provoking. David, you also said before that the emergence of ChatGPT may break a certain contract of the Internet. Can you talk about it in detail?

Chris Dickson : Yes, there is a chapter in my book called "The New Covenant" that talks about this. The success of the Internet is largely because it has a clever incentive mechanism that allows 5 billion people to join voluntarily without the need for a central authority to enforce it. Over the past 20 years, the Internet has gradually formed an implicit economic contract, especially between social networks, search engines, and content creators. Take Google as an example. Website owners allow Google to crawl content and display snippets, provided that Google returns traffic. Creators make money from traffic, whether it is advertising, subscriptions, or other models. This mutually beneficial relationship is the foundation of the Internet's prosperity.

But occasionally this contract is broken. For example, Google's "One Boxing" feature, which directly displays answers without redirecting to the original website, has caused great harm to sites like Stack Overflow, Wikipedia, and Yelp. The user experience may be better, but the creators' traffic has decreased. Now the rise of AI has further challenged this contract. Chatbots can generate illustrations or recipes directly without users having to click on the original website. If all AI systems work like this, traffic will no longer flow back, and the creators' survival foundation will be broken.

These AI systems rely on data trained under the old contract, but the new model no longer follows the old rules. I worry that the future of the Internet will become a closed system dominated by three to five large companies, and billions of other websites will die due to loss of traffic. This makes me uneasy - is the Internet going back to the broadcast TV model of the 1970s, with only a few channels? What good will such a world do for startups, innovation and creativity? How can long-tail websites survive? How can new things break through?

I am not saying that encryption technology is the only solution, but at least we have to admit that the status quo breaks the original incentive mechanism and think about whether this is a good thing. If not, how can we design a new mechanism? Story Protocol is an attempt to rebuild the incentive system for creators through blockchain.

Host : You mentioned that AI, encryption technology and new hardware are a trinity that reinforce each other. Can you talk specifically about how they work together?

Chris Dickson : Absolutely. Take the mobile Internet, social networks, and cloud computing, for example. They have enabled each other. Now the same is true for AI, encryption, and new hardware. You can already see some clues, such as AR/VR glasses and self-driving cars that use AI technology extensively, and companies such as Tesla are also beginning to show their talents in the field of humanoid robots. These technologies bring AI into the real world and open up new application scenarios.

On the crypto side, I am particularly optimistic about an area called DePIN (decentralized physical infrastructure). For example, the Helium project is a community-owned, crowdsourced telecommunications network that challenges the traditional model of Verizon and AT&T. Users install Helium nodes (wireless transmitters) in their homes to contribute coverage to the network. There are currently hundreds of thousands of nodes across the United States, providing much cheaper services than traditional operators ($20 vs. $70 per month). This is feasible because it uses cryptography to design an incentive mechanism that avoids the billions of network construction costs of traditional operators.

The most difficult part of network construction is the startup phase, because the network effect is weak in the early stages - just like a dating site, no one uses it with 10 users, but it is useful with 1 million users. Cryptography solves this problem through token incentives, and early participants are rewarded, thereby promoting network expansion. DePIN's ideas are not limited to telecommunications, but also extend to areas such as climate modeling, map data, and electric vehicle charging. For example, one of our recent investments is a decentralized energy monitoring network, and some people use similar methods to do decentralized science. The early construction of such networks is naturally suitable for cryptography, and AI can cooperate with it in data collection and processing.

Moderator : The stages of technological development are also critical. How do you view the evolution of AI?

Chris Dickson : I like to use a framework to analyze technological development, which is divided into three stages: the first is "skeuomorphic", which is to use new technologies to improve existing things; the second is "native", which is to create something that was previously impossible; the third is "second-order effects", which are the profound changes caused by the popularization of technology.

Take the Internet as an example. The 1990s was the first stage. People moved magazines and catalogs online. Amazon was more convenient to sell books than flipping through magazines, but the essence was still a new form of old things. In the 2000s, social networks emerged. These were truly native applications with no offline counterparts and a completely new business model. AI is similar. The first stage is the now common "old things are made new", such as replacing call centers with AI customer service, which is both cheap and efficient, and may affect tens of millions of jobs, but it may also create more new opportunities. This stage may last for 20 years.

The second phase is the “native” phase, which is what really excites me. For example, when photography became popular, art turned to abstraction (such as Cubism), and at the same time gave rise to a new art form, film. The same is true of generative AI today. Some people think it threatens creativity, but I don’t think so. It may be the cornerstone of new art forms, such as virtual worlds, new types of games or movies, and perhaps completely new interfaces. These innovations require talented creative people to realize them, and they are often unexpected. Just as film broke new ground, AI may bring similar breakthroughs.

The third stage is the "second-order effect". After the rise of social networks, Obama's victory in 2008 marked a turning point, followed by the Trump movement and populism. These are second-order effects and are still evolving. The second-order effects of AI may not fully appear until 20-30 years later, and each stage may last for a decade.

Moderator : What are the limiting factors in the transition from the first stage to the second stage?

Chris Dixon : The early Internet was limited by physical network construction, such as laying cables. AI is different. Technical capabilities are no longer the main bottleneck, but human creativity and policies and regulations. The supply side needs creative talents to develop native applications. The current entrepreneurial ecosystem in this regard is much more mature than it was 15 years ago - venture capital institutions have increased from dozens to thousands, entrepreneurial advice is better, smart people can enter this field more easily, and capital and energy are abundant.

But the demand side is more challenging. Changes in organizational and individual behavior take time. For example, I want AI to read my books and imitate my voice, but publishers and Audible completely ban AI because of unions and traditional concepts. It may take a generation for Hollywood to accept AI-native movies, and it may have to rely on AI startups in emerging countries to promote it. The policy level is more complicated, and regulated industries such as copyright, medical care, and finance, which account for 70% of the economy, will face fierce controversy. Is AI training data "copied" or "learned"? This may ultimately be resolved by congressional legislation rather than the free market or court rulings.

Host : What is your ideal future of the Internet?

Chris Dixon : We are at a crossroads. The original vision of the Internet was community-owned and community-governed, with benefits flowing to small businesses, innovators, and entrepreneurs at the edge of the network. But today, wealth and power are concentrated in the hands of a few large companies, with the five largest technology giants accounting for more than half of the market value. The first sentence in the book is "Architecture determines destiny." Control and funding flow depend on how services are designed.

I worry that we are approaching a point of no return where the internet is dominated by five companies. They have saturated user growth and are now starting to "kick back the ladder" and hinder new entrants. This is a huge threat to "little tech". If startups have to pay a huge "tax" to the giants to compete, they will not be able to challenge the status quo. We have seen similar cases in the past, such as Zynga, which relied on Facebook and eventually suffered from platform risks.

Therefore, it is crucial to support new architectures (such as blockchain) and open source AI. This is not only a technical issue, but also a regulatory and public perception issue. We need policies to encourage competition and innovation and avoid eating the "seeds" of the future. Through A16Z's efforts, I am optimistic that the concept of "small technology" is spreading and more and more people are aware of the importance of new infrastructure and open source.

Host : Thank you, Chris, it's a pleasure to speak with you.

Chris Dickson : Thanks for having me, it was great chatting with you!