The Convergence of AI and Crypto
In a discussion hosted by Coinspire, a16z Growth partner David George and a16z crypto partner Chris Dixon shared their insights on how artificial intelligence (AI) and cryptocurrency are poised to revolutionize the internet landscape.
Dixon opened the conversation by highlighting that technological advancements often emerge in pairs or triplets. He drew parallels to the significant trends of mobile internet, social networking, and cloud computing fifteen years ago, emphasizing their interdependence. He then suggested that AI, cryptography, and new devices (like robots and VR) represent the most compelling trends of the present. The discussion emphasized cryptography—particularly its potential to architect new internet models—as a powerful tool for building networks with unique capabilities.
Decentralized AI and Open Source Concerns
Dixon expressed concern regarding the increasing closure within the AI field, noting the shift from open research to proprietary models driven by competitive interests. He expressed hope for open-source projects like Llama, Flux, and Mistral but questioned their long-term viability due to issues like the unavailability of model weights. He noted that many models are open source, but their data pipelines are not, making true reproducibility difficult.
Crypto’s Role in Decentralized AI
According to Dixon, some projects are developing a decentralized internet service architecture tailored to the AI ecosystem. He cited Jensen, which uses a model like Airbnb for distributed computing resources. He also mentioned Story Protocol, a platform for registering intellectual property on the blockchain, which allows creators to set their own terms for content usage, fostering a more democratic and transparent marketplace.
“The blockchain creates a broad democratic resource where small creators can set their own terms,” Dixon explained. He highlighted the inherent composability of cryptography, a key driver for the success of open-source software. Dixon also considered this composability similar to how Wikipedia integrates knowledge. Story Protocol, he added, allows creative content to be combined freely, fostering new universes with shared revenues.
New Economic Models for Creators
George suggested that the transparent and equitable flow of funds is the key to success for the model. Dixon added that this approach incentivizes creators to use AI tools while rewarding them financially, a crucial element as a16z considers supporting new economic models for creative workers in the AI landscape.
They also discussed the concept of ‘crowdsourcing’ data for AI. Cryptography can design new incentive systems for collecting more AI training data. WorldCoin, co-founded by Sam Altman, was highlighted as an example. WorldCoin uses blockchain technology to verify human identity in a world where AI can generate it deceptively. It will also provide simple applications like CAPTCHA to verify that users are human and not bots. Dixon saw many more opportunities for decentralized AI at the infrastructure level for machine-to-machine payments and beyond.
Breaking the Internet’s Contract
Dixon believes that the existing economic model of the Internet is being disrupted, particularly by the rise of AI. The current system, where search engines and social platforms benefit from creators’ content, may deteriorate. AI content generation could threaten this model. It could lead to a scenario where a few AI companies control all content. Dixon pointed out that there is a need to find mechanisms to support innovation and entrepreneurship.
AI, Crypto, and New Hardware
Dixon analogized the combination of AI, crypto, and new hardware to the mobile internet, social networking, and cloud computing. He mentioned AR and VR glasses that rely on AI for enhanced user experiences, self-driving cars that use AI to perform in the physical world, and cryptography for decentralized networks to support these applications.
He presented Helium as an example of decentralized physical infrastructure (DPIN), where a community-owned telecommunications network competes with traditional operators, incentivizing network building by individual users.
Network Effects and the “Cold Start” Problem
Cryptography has the ability to solve the “cold start” challenge, particularly in network effects. Dixon noted that network effects depend on sufficient participation and that crypto can incentivize early adoption by using token economics, as Helium had done. He also added that this concept could be applied in a variety of industries, including data and scientific research.
AI: Icing or Sugar?
George asked about AI being “icing” or “sugar.” If AI is just an add-on (“icing”), existing giants will likely prevail. If AI is a core component (“sugar”), then new companies are likely to emerge. Dixon cited Clayton Christensen’s concepts of “disruptive innovation” and “sustaining innovation,” noting that disruptive innovations don’t fit the business model of existing companies. For example, AI may change data storage and retrieval, making traditional databases useless. This will change database structures forever.
The Low Network Effects of User AI
Looking at consumer AI, Dixon said no existing products have strong network effects. A strong data network effect that would give AI products a strong moat is difficult to achieve. Dixon indicated that many believe that having more data will improve AI models and generate more users—but this has not happened. This leads to intense market competition, ultimately resulting in price wars. This is dangerous for AI companies, he stated. The question is whether the product can establish network effects.
Come for Tools, Stay for the Network
He emphasized the importance of creating an AI creative community, using the example of Adobe Photoshop software, where long-term users valued the strength of the network. He also spoke on the importance of the “Come for the tools, stay for the network” approach. This is a key part of user growth if the AI tool is not to become a commodity.
