PriceWaterhouseCooper estimates that A.I. will add $15.7 trillion to the world economy by 2030. IBM predicts that blockchain technology’s contribution will have reached $3.1 trillion by then.


Artificial intelligence (AI) is the simulation of human intelligence processes by machines, especially computer systems”

“A blockchain is a decentralized, distributed and public digital ledger that is used to record transactions across many computers so that any involved record cannot be altered retroactively, without the alteration of all subsequent blocks.”

Clearly, both technologies have the potential to be extremely valuable on their own. But teven more intriguing is their convergence.

Today, A.I. is a limited technology in its infancy. The hope is that that blockchains can speed up its development, making Aasimov-style robots a reality and helping usher in Industry 4.0.

In this article, we’ll look at why today’s A.I.s are limited; how blockchain technology can help improve them; what the far-reaching implications are. We’ll start with...

3 Problems that Limit Modern A.Is

Today, artificial intelligence suffers from several straightforward but hard-to-solve problems.

1. The black box problem

Modern A.I.s rely on multiple sophisticated algorithms working in tandem to create “intelligence”. These systems are so complicated that humans cannot readily understand nor edit them. If an A.I. learns to make mistakes, there’s little we can do but roll it back. This is the black box problem.

2. Insufficient/unreliable data

Artificial intelligence uses data to predict future results. This creates opportunities for intentional or accidental bias. For example, one A.I. that analyzed hundreds of thousands of articles “learned” to view the words victim and woman as synonyms despite getting plenty of data. This is problem #2.


3. Overfitting

This problem overlaps with the previous one. When machines get too much data, they can create oversimplified, erroneous predictive models. In mathematical terms, this means A.I.s have a tendency to overlook outlying, off-the-curve data points.

All told, the current generation of intelligent machines is limited to making simple, straightforward predictions.

One example is "Siri" or "Cortana" predicting what you said based on millions of previous recordings. This is a cool, useful feature - but not true intelligence.

The hope is that in the near future, A.I.s advance to a point where they can do 2 things.

First, display their inner workings in a way that allows for easy human understanding and editing. Second, gain the ability to self-educate, eliminating problems that stem from overfitting and poor or unreliable data.

Once we get to this point - often called A.I 2.0 - it will be a (fairly) short leap to Aasimov’s robots: intelligent machines that can not only learn effectively and independently, but create new A.Is and algorithms that go beyond current capabilities.

The question is, how can blockchain technology help solve the above problems in the coming years? Will blockchains help us get to A.I 2.0 and beyond, and if so, how?

How Blockchain Technology can Help Us Improve Machine Intelligence

A blockchain is an immutable ledger shared by multiple parties via a distributed network. Put more simply, a blockchain is a way for people and machines who don’t know each other to share records of all past events using a decentralized database.

At first glance, blockchain technology can seem completely divorced from A.I. tech. One works to make accurate predictions by centralizing and analyzing vast amounts of data; the other focuses on decentralization.

But consider this. Vitalik Buterin, the brain behind Ethereum, envisioned a “World Computer” when creating his blockchain.

His ambition was to build a Turing-complete language that could simultaneously run on millions or tens of millions of machines simultaneously.

When you look at blockchains from this angle, their applications for A.I become readily apparent. We may not see Vitalik’s dream come to fruition in the near future - but blockchains can certainly help solve some of A.I’s problems. Here’s how.

Blockchains and the Black Box Problem

Blockchains don’t necessarily reveal the inner workings of a complex A.I. They do, however, give A.I systems immutable and transparent records. These can help humans and machines see how each specific A.I. decision was made, helping operators pinpoint instances where they need to get involved - and highlighting problematic decision trees that need to be fixed.

Blockchains and Data Reliability

Blockchains help improve data accuracy in two ways.

First, by giving other blockchain agents a clear audit trail that helps A.I bots know who to trust and distrust. Second, by feeding machines more data than they could find in a centralized system.

In both cases, data quality and quantity improve, reducing the likelihood of mistakes and speeding up machine learning. Taken to an extreme, blockchains can help power next-gen A.I applications like warm robotics with vast numbers of agents acting in concert in real-time. (This is something Asimov envisioned in his writing.)

Blockchains and Overfitting

Overfitting is one of the biggest problems in machine learning. Blockchains can help solve it by disincentivizing instances of overfitting as well as the decision trees that lead to them.

For example, imagine that an A.I makes a future prediction based on live data. If the prediction is right, we can use staking mechanisms to drive similar decisions. If the prediction is wrong (due to overfitting), we can downgrade the A.I that failed to make correct predictions. Eventually, we will end up with a more nuanced predictive model that helps us beat overfitting.


In 1964, Isaac Asimov predicted that “robots will neither be common nor very good in 2014”. This was true 4 years ago, and it still holds today - both for A.Is and physical robots.

It’s unrealistic to expect that blockchains are going to come in and solve all of A.I’s current problems. Having said that, immutable, decentralized ledgers can certainly help with overfitting, data errors and the black box problem. We may not see a (beneficent) SkyNet or an Asimov-style world where robots do most of the work. But even from where we stand today, it’s clear that blockchains powered by thousands or millions of machines certainly raise the ceiling for where A.I tech can go in the next decade.

How did you like this article?

If you feel we missed anything, or would like to add a tip, question or factoid, please use the form below to do so now.

powered by Typeform