David Johnston, entrepreneur and one of the leading contributors to Morpheus AI, envisions a future where AI isn’t just decentralised, but deeply personal.
His vision for decentralised AI through Morpheus is a call to empower individuals in an era dominated by tech giants. By leveraging open-source models and Web3 infrastructure, Morpheus enables users to own their AI, customise their digital experiences, and protect their data.
The concept of brainpower as a measurable unit underscores the transformative potential of AI, offering a concrete way to understand its capacity to amplify human intent. As AI agents take on more complex tasks and reshape economies, Johnston sees a future where intellectual labour is democratised, which will help foster innovation and freedom on an unprecedented scale.
In a recent interview with Deeptech Times, Johnston breaks down the coming AI revolution, the perils of centralisation, and why APAC could be the epicentre of the next wave of innovation.
You mentioned Morpheus is like Bitcoin in that there’s no company or founder. How does that work?
Exactly. No company, no CEO, no foundation. The Morpheus protocol was launched by anonymous developers. It’s open source and over 300 contributors globally build on it. I’m just one of them.
The community is very active across APAC, especially in places like Singapore and Malaysia. Just like Bitcoin didn’t need marketing to grow, Morpheus is expanding organically because people want it, not because someone’s spending VC money on ads.
What does decentralised, permissionless AI mean, and how does it differ from centralised AI systems?
At its core, decentralised AI is about putting the individual at the heart of the technology. I envision each person owning a personal AI, much like a crypto wallet with a private key, where you control your data, prompts and outputs.
This contrasts with centralised AI, where large companies or governments control the models, filter responses and limit customisation. With decentralised AI, like what we’re building at Morpheus, you can set your own system prompts, tailor the AI’s voice or persona, and ensure it aligns with your values and goals. It’s about giving power back to the individual, not a corporation like Google or Amazon.
Centralised systems, such as ChatGPT, often come with filters that shape responses based on what the provider deems appropriate. For example, a few weeks ago, users noticed ChatGPT became overly flattering, avoiding critiques entirely. With an open-source model on Morpheus, you bypass these restrictions, enabling a more authentic interaction with the AI. This level of customisation and control is what makes decentralised AI so powerful.
What are the key advantages of decentralised AI over centralised models?
The biggest advantage is customisation and control. In a centralised system, you’re interacting with a filtered version of the AI, not the raw intelligence.
Decentralised models are interoperable and composable, meaning they can work together and be tailored to individual needs. Think of the early internet: closed systems like AOL which tried to dominate, but the open, permissionless web won because anyone could build on it. I expect the same for AI where open-source agents will surpass closed ones because they’re more flexible and user-driven.
Another advantage is actionability. Centralised AIs, like those from Google, can’t take actions like sending stablecoins or interacting with Web3 wallets due to regulatory and legal constraints.
Morpheus-enabled agents, however, can act on your behalf, whether it’s paying for services, managing tasks or interacting with other agents. This ability to “bake the cake” rather than just provide a recipe is a game-changer. By the end of 2025, I believe decentralised AI will outpace centralised systems because of this freedom and functionality.
What role does regulation play in this transformation? Are governments ready for decentralised AI?
The good news is most early proposals to restrict open-source AI have faded. Places like Singapore, Japan, and even the U.S. (after a brief dark period in 2023–24) have become much more supportive. Developers go where they’re treated best.
Just look at crypto: when the U.S. pushed it out, talent and innovation moved to APAC and Europe. The same will happen with AI. Singapore, for instance, has one of the most forward-looking policies and could become a global hub for decentralised AI.
How will AI agents change daily life and the economic landscape over the next five years?
AI agents are already transforming my daily life. I use half a dozen agents to summarise emails, follow up on tasks, and manage digital interactions. For example, after a conference, I can drop a business card into a digital channel, and my agent handles follow-ups automatically.
In the next five years, I see millions of agents wrapping every API and website, communicating seamlessly in the background. Your personal “super agent” will understand your intent and connect with other agents to accomplish tasks, much like a search engine finds websites.
Economically, AI agents will follow Jevons Paradox: as the cost of intellectual labour drops, we’ll use more of it, not less.
Instead of replacing developers, AI will enable a billion developers by making software creation accessible to everyone. This will remove bottlenecks, allowing anyone to create apps or websites, and leading to an explosion of innovation. Morpheus is already empowering this shift, with over US$10 million in rewards paid to open-source developers through its subnet staking model, making it easier to monetise contributions.
Can you elaborate on the concept of measuring brainpower and its significance for AI’s future?
Measuring brainpower is a way to quantify AI’s intellectual capacity in a standardised, relatable way, much like horsepower measures physical power.
I’ve proposed a unit called ‘brainpower’—defined as 3.3 tokens per second, which is roughly equivalent to 2.5 words per second, the average speaking rate of a human. This gives us a baseline to compare AI output to human cognitive output.
For example, an NVIDIA H100 chip can process 1,200 tokens per second, equating to about 300 brainpower units. With NVIDIA’s Blackwell chips, which are four times faster, we’re looking at 1,200 brainpower units per second per chip.
In 2025, based on the projected production of Blackwell chips, we’ll have access to 50 trillion hours of AI brainpower globally, compared to 7 trillion hours of human brainpower.
That’s a fivefold increase in available cognitive capacity, and it’s growing exponentially. This metric is significant because it allows us to move beyond vague descriptions of AI’s potential and talk about it concretely, like we do with horsepower for cars. It’s not just about raw computation but about understanding how much intellectual work AI can perform relative to humans.
This explosion in brainpower will amplify human potential, not replace it. Just as cars didn’t eliminate the need for human travel but expanded where we could go, AI brainpower will enable us to tackle more complex tasks and achieve greater outcomes.
For instance, where intellectual labour like software development or research was once a bottleneck due to cost and time, AI agents can now perform these tasks instantly, freeing humans to focus on defining goals and priorities.
The blockchain ensures this brainpower is accessible to individuals, not just corporations, through persistent, decentralised AI systems. As brainpower scales to 10x, 100x or even 1,000x human capacity in the coming years, the question isn’t about AI developing free will, it’s about where humans want to ‘drive’ this intellectual power. It’s a tool for amplifying our intent, and Morpheus ensures it’s in the hands of individuals, not centralised gatekeepers.
How close are we to realising the full potential of decentralised AI, especially with tech giants dominating the space?
We’re closer than you might think, potentially by the end of 2025.
Open-source models like Llama and DeepSeek have already caught up to proprietary ones, and Web3-based models can do things centralised systems can’t, like integrating with crypto wallets or performing tasks autonomously.
The ecosystem is growing rapidly. Morpheus, launched just last year, already has over a million users and 100+ projects, including Venice AI. This growth is organic, permissionless and unstoppable, much like Bitcoin or Ethereum. The key is the infrastructure. Blockchain ensures data persistence, so your AI and its memories are yours forever, wherever you go.
