The Next Internet Paradigm: Onchain Self-Writing AI

The future of the internet is not only decentralized; it’s intelligent and alive with onchain AI, enabling a seismic shift in how we create and use technology. While recent consumer and media attention on AI has been largely focused on the potential of large language models (LLMs) and Big Tech’s focus on AI tools, the future of the internet is not centred around monolithic models that help with search and the creation of content. Instead, it will be centred around a real-time, customisable and personalisable internet, enabled and protected by onchain AI; which is critically important.

In an internet age increasingly defined by hacks, data breaches and misinformation, AI on chain is not just an ambition, but a necessity. It ensures far greater levels of security, data integrity, and immutability. No single points of failure, no centrally-controlled cloud, no need for millions spent on cybersecurity apparatus to try and protect the data and models from hackers that are also using AI for nefarious purposes.

In this future state, our shared view of AI will evolve - both in terms of its applications and the infrastructure that supports it.

With AI on chain, models will be able to deploy and update fully running applications in real-time, using simple and accessible prompts. Users will describe the app or service they want to create, and within moments, it will exist - hosted on a decentralized network. This is AI at scale. This is our vision for the future of the decentralized internet.

The next internet will create virtual reality experiences for us, tailored in real-time to be the most enjoyable and compelling, based on our immediate feedback and interactions. The signs of this new paradigm are all around us.

Personalisation at Scale
The most popular consumer applications on the internet are the ones that are the furthest down the pipeline towards a personalised internet. Consider a TikTok or Netflix account, where each profile is a personalised reflection of the signals the user has given the AI algorithm about their preferences for TV shows, movies or video content. As more and more of our time is spent online, which invariably feeds the revenue models that sell data trends as an outcome of our attention, it is no wonder that the pages of social media and streaming sites feel most personalised and end up the most popular.

In our view, the future of these personalised internet use cases will be omni-present.

Imagine TikTok’s algorithm, but instead of recommending videos, it builds personalized applications on-the-fly, tailored to your exact needs, whether it is a CRM for a local business, a research page for a law firm, or an admin tool for a school district. All sounds great, right? But certain challenges remain.

Key blockers for this future state are many. In particular, creating a personalised internet application is currently complex, as well as time and resource intensive. It requires expertise in coding, data management and cyber security. Using current internet infrastructure, anyone building or interacting with an AI-customized application has to give over their data to large corporations who have the means to deliver on these aspirations, but ultimately may have ulterior motives for the data. Meanwhile, developers who have the creativity and necessity to build new and better applications also have no ownership of the models with the status quo.

Most crucially, as enterprises and developers build these personalised models, the risks of hacks or compromises of data become more critical and more prevalent than ever. An illustrative use case is a law firm that wants to drive efficiency, reduce man-hours, and use an AI model to prepare for upcoming court cases. For the law firm and their clients to trust the AI model, it will need to be so much more than just functional. Instead, they will need to be absolutely confident that neither the model nor the data cannot be hacked, compromised, or modified by a malicious actor.

AI Onchain
Onchain AI solves many of these problems. By using AI models that are on the blockchain, people who create custom applications and internet services will - first and foremost - own the applications that they create, as well as the data inside the models.

Onchain AI is sovereign, making it inherently tamperproof, resistant to cyberattacks, and unstoppable. This means creators of new web applications will not need expertise or be required to invest in configuring complex cloud technology to ensure their applications are safe from attack.

The Internet Computer (ICP) protocol is uniquely positioned for this new paradigm, by providing a more secure and easy-to-use infrastructure layer that enables the intuitive creation of secure and unstoppable AI models, as well as reducing the complexity and user cost of maintaining them. ICP is building fully onchain functionality that empowers users to prompt AI with simple instructions, that will then build and update digital services and apps quickly and efficiently. Meanwhile, ICP provides orthogonal persistence which automatically handles data storage and security for AI-driven applications, removing the complexity of AI handling.

Beyond ease of use, ICP offers significant benefits in the security of AI models. ICP is the only blockchain protocol that can run an AI model as a smart contract. This means that all AI applications built on ICP are protected against common cyber threats. ICP’s decentralised node infrastructure means that AI models are unstoppable. They cannot be stopped by centralised cloud infrastructure providers “turning off the taps”, and models cannot be revised by malicious actors because they are running as smart contracts.

Finally, ICP’s unrivalled approach to AI onchain has the potential to vastly increase the proliferation of new and personalised applications, and will enable everyone to be a problem-solver and the guardian of their own creations. This will, in turn, create a seismic shift in online infrastructure. Rather than software engineering jobs being focused on creating cookie-cutter applications and managing the cybersecurity or infrastructure of applications, software engineers will support innovators and users to better personalise their applications, if desired. In short, the democratisation of these tools has the potential to offer new levels of user ownership in defining our shared online infrastructure, creating huge numbers of software engineering jobs globally, focused on supporting new customer applications and services to solve specific issues or interconnecting challenges.

Excitingly, this will unleash human creativity and break down the barriers to entry that are well established and maintained by traditional IT infrastructure. Instead, innovation will come from all corners of the globe, from all levels of digital literacy, and all ends of societal spectrums – creating and contributing to the internet with self-writing sovereign apps and services.

The future is bright when the internet is rebuilt on the blockchain. LLMs are just the entrée to this evolution. It’s time to think bigger and bolder.

About the Author
Article authored by Tracy Trachsler, Head of Institutional Relations, DFINITY Foundation