Artificial intelligence is a no-brainer for people pushing innovation and changing how we live and work. But as AI is taken up and we grow more dependent on it, so too come questions about how centralised the infrastructure on which it operates is – and what risks that carries with it.
Cryptos have shown us the value of decentralisation, and the potential problems with centralising compute and data within a few discrete facilities…. Although it may seem efficient to take that approach, it comes with some significant weaknesses and questions about access and governance. “Decentralisation means no single point of failure, more transparency, and more user control”, and accessibility for everyone. Welcome to the world of AI blockchains – the bedrock for a more robust, fair and sustainable AI industry.
Centralised AI systems are hugely fragile, since the processing that underpins them is all held in big server farms as a kind of single point of failure that could bring hundreds of apps crashing down. Data centres that fuel AI models such as ChatGPT are also a tempting target to hackers because they have access to vast amounts of information.
Centralised servers also lead to more regulatory headaches. If a country-based AI system is only available within one particular country’s jurisdiction, that poses a problem for users in other jurisdictions with different rules around data sovereignty and privacy. Centralisation, of course, also means monopolisation, and we already have ample proof of that in the way OpenAI, Google and Anthropic have been super cagey about how they train their super duper advanced AIs. The risk is that only a few big corporations will eventually become the only gatekeepers of a technology that has become a crucial part of modern life, excluding anyone who doesn’t want to pay whatever price they demand.
Thankfully, decentralised AI is a pre-built answer to the issues. The infrastructure now that the model hosts, but also a large group can have the models running it can be scattered across a massive population of users, removing the centralisation concerns.
Key features of AI blockchains
The intersection of blockchain and AI is very promising because of their combination. The immutability of blockchain can be helpful in terms of providing integrity and trust of the data that drives AI-infused systems, while AI can add automation and intelligence to systems based on blockchain. The synergies are clear. Take, for example supply chain, where blockchain can provide complete transparency and visibility, and AI can anticipate shifts in demand and adjust logistics accordingly. It is also the healthcare case, where blockchain may secure medical records, while AI can diagnose disease by analysing images and sensory information for trends and correlations.
1: Clear data attribution
An important feature of AI blockchains is transparent data attribution, which leverages a proof-of-attribution consensus algorithm to trace and reward the data source used by AI systems, which enhances fairness. It adds transparency behind the data that trained AI, disclosing who supplied it, what impact it had on the AI’s outcomes, what its worth is and how the supplier of the data should be remunerated.
One example of this in practice is OpenLedger, which eventually developed a rewards mechanism to ensure that whenever a model feeds on someone’s data, the person who provided the data would be compensated with digital tokens. This is opposed to centralised AI companies, which collect data without the creator knowing and putting it outside of the value chain.
2: AI ransoms and the monetisation layer
Let’s say someone asks a question to a distributed chatbot, and the bot uses information it finds in a post on Substack or Medium to answer. The system would register that the model utilised this input in determining its output, and it would process the payment of tokens to the creator of the content through the means of smart contracts. This opens the door for a new creator economy of people who make specialised data sets for AI models and host them on blockchains, where their contributions can be entirely attributed and remunerated.
3: Decentralised lifecycle of models
Another important distinction is the fact that the process of developing blockchain-based AI is completely transparent: from an idea proposal, through model learning and finally to its deployment. Foundation It enables the creation of models owned by a community of users and governed through democratic mechanisms where those holding a token vote on new features to be added.
4: Efficient, scalable underlying infrastructure
Decentralised infrastructures offered by their users AI blockchain run on decentralised infrastructures that are decentralised, offered by their users. For example, Render Network has cultivated a network of GPUs, but they are not sitting in a centralised data centre. Instead, people who belong to the network rent out the spare GPU capacity of their laptops and desktops, which is then consolidated and provided to AI applications that require processing power. Developers can obtain the infrastructure they want at a lower price, while those furnishing it can be compensated in tokenised form.
OpenLedger is playing a major role in increasing the efficiency of decentralised infrastructure with OpenLORA. It’s a highly scalable and high-performance framework that can handle hundreds of fine-tuned AI Models in parallel on a single GPU block; therefore can run all those models at the same time, resulting in significantly lower operating costs. This makes advanced AI applications far more accessible and radically reduces the cost for the end user.
Why do AI blockchains matter?
Most AI services today in the world live inside centralised servers in so-called “black boxes” that are all but impossible to see into, so we never actually know what these systems do, or the data they use, or how they work. They are owned by a few powerful companies, and this concentration disrupts AI’s democratizing promise. AI companies such as Google, OpenAI and Microsoft could aggregate massive power over our societies and lives, keeping all the profits they make from these systems, even if they obviate numerous current jobs.
Decentralised AI networks are our chance to stop this monopolisation. There is no question that AI is the most powerful new technology to emerge since the advent of the internet, and it’s simply too important to allow it to fall under the complete control of a relatively small number of companies. Should the predictions come to pass, AI should be incorporated into everybody’s lives, be it in the places they work and travel, and even send tailored experiences to them, and transform health care.
We do that with AI blockchains, and now, when we construct intelligent systems, there’s the possibility of trust-based deployment models that decentralise control so power doesn’t condense, and an incentive for people to participate in developing these killer apps. The possibilities for grassroots innovation become endless, giving anyone the ability to present an idea and work with a community to realise it with decentralised governance ensuring it evolves in a way that serves everyone’s needs, rather than the profit-driven objectives of ca orporation. It will encourage a more diverse ecosystem of AI applications that everyone can use, and reduce its application for oppressive purposes.
We can’t allow AI to be monopolised
As much danger as promise exists in today’s AI landscape. Technologically, it’s advancing so fast in such a short time that there is an actual danger of monopolisation of this, and that is not without danger of misapplication.
AI blockchains are how we might stop that happening, and the only way to build platforms for AI that’s free at the point of use and done in a decentralised way, but with checks on abuse.
Creating the decentralised future of AI needs coordination at every level, from the data that is used to how the models are being trained, and the infrastructure on which they run. Once we have a mechanism of transparent attribution, we can reward this type of cooperative behaviour, so that participants can contribute to the next technology revolution, and no one will be left without a reward.
