Integrating AI and Blockchain, and a Free Business Idea

Since we founded LearnerShape, our primary technological focus has been on using artificial intelligence to improve education, and we are making excellent progress in that mission. In the last several months,we have also begun to pay significant attention to blockchain technology, not least because of our exciting universal skills authentication project with Cardano.

This has recently got me thinking about incremental and disruptive innovation from the combination of these two technologies:

• How can we use AI and blockchain together to improve existing and new applications?
• Are there new things that we can do by integrating AI and blockchain that we couldn’t do without both technologies?

Below I unpack these questions, and then propose a (free!) business idea emerging from the second one.

Using AI and Blockchain Together

By “using AI and blockchain together”, I mean that AI and blockchain can be separate components that work together to improve an overall application.

LearnerShape is already planning to combine our AI-based technology for flexible, taxonomy-agnostic handling of skills with our blockchain-based universal skills authentication technology being developed for Cardano. This will support a powerful set of applications for flexible authentication of any set of skills. For example, these applications will allow a company or organization to describe required skills using its own language, and then reliably compare those requirements to a variety of similar skills possessed by individuals and authenticated by third parties. This is essentially the typical hiring process, greatly improved by AI elements that allow automatic comparison of skills and blockchain elements that support trusted skills authentication.

But AI and blockchain are not inseparable in this process. Our AI methods are useful without blockchain-based authentication – indeed, AI on its own is at the core of our current business. And blockchain-based authentication can be very useful without AI elements – that’s what we are building in our first project with Cardano.

This ability to combine AI and blockchain or use them separately fits neatly with our open-source infrastructure approach to building flexible learning applications. Our AI and blockchain functions will be available through separate open-source APIs that will allow us and third parties to combine them flexibly to meet specific needs of different learning environments.

Integrating AI and Blockchain

By “integrating AI and blockchain”, I mean something different, which is perhaps best expressed as a question. Are there things that we can do with AI and blockchain together that are genuinely new, and that would not be possible without both technologies? This would be like what Apple and Steve Jobs did by combining wireless communications, touch screens and an app store to produce the iPhone.

My initial answer to the above question for AI and blockchain was “nothing important that I see yet”. That is, I saw uses for AI and blockchain separately, and ways to combine them usefully (e.g. as above “using AI and blockchain together”), but the two technologies seemed different enough to make genuine integration difficult.

Yet it seems nearly certain that the market will find clever ways to integrate AI and blockchain. The rapid development of both technologies over the past decade or so has brought many surprises – things like:

• for AI, the victory of Deepmind’s AlphaGo over the world’s best Go player, Ian Goodfellow’s generative adversarial networks producing applications like This Person Does Not Exist, and the stunning cross sectoral-utility of large pre-trained language models like BERT and GPT-3 (which I like to call “foundation models”, using a Stanford University neologism)
• for blockchain, a market capitalization of around $2 trillion (and apparently rising) for cryptocurrencies, the current explosion of non-fungible tokens, and the proliferation of ‘Web3’ ideas for decentralization of the internet (I’m currently reading Shermin Voshgmir’s Token Economy about Web3, and highly recommend it).

Such innovation, applied to integration of AI and blockchain, will produce results.

And then this week I encountered a first idea for integrating AI and blockchain, suggested by the Head of Ideas and Creative of my family company Lily Innovation.

A Free Business Idea – Proof of Explainability

What my colleague suggested is to use AI and blockchain together to help deal with the challenges of explainability of AI models. We call this “proof of explainability”.

The idea as we have elaborated it so far is this:

• An organization running an AI model can store data associated with each sensitive decision or inference using the model in some kind of flexible data structure, perhaps using JSON. This structure could contain a wide variety of data, e.g. inputs, outputs, model weights, etc.
• At the time of decision or inference, a hash of the data structure relating to the particular decision/inference could be stored on a public blockchain (or alternatively it would be possible to store only the root of a Merkle hash tree for a batch of decisions).
• If the decision were ever questioned or otherwise examined (e.g. by the subject of the decision or a government regulator), it would be possible to produce (from the stored data structure), prove (from the blockchain hash) and analyze the model parameters that were the basis for the decision.

With the large interest in AI explainability, including legal initiatives like the EU proposal for an AI Regulation, we think this idea has real business potential. But it isn’t related to what LearnerShape does, so we have decided to stick to our knitting and share the idea with the world. Feel free to take it and do with it what you will. Or better, get in touch to talk about this or other ideas for integrating AI and blockchain – I love talking and thinking about this stuff!

Maury Shenk, Co-Founder & CEO, LearnerShape