Artificial Intelligence has a history that goes back a long way, the first major works were by John McCarthy, Allen Newell and Herbert Simon and date back to the 1950s. The first experiments on Blockchains (chains of blocks) occurred instead in 1991 thanks to the work of researchers Stuart Haber and W. Scott Stornetta, even if the real revolution took place only in 2008 with Satoshi Nakamoto, a pseudonym which nobody has yet been able to ascribe to an individual or a team, with the presentation of the paper “Bitcoin: A Peer-to-Peer Electronic Cash System”.
Despite this important age difference, these are two enabling technological ecosystems with interesting complementarities and which are inevitably destined to collaborate closely in the near future.
According to this report by PWC, by 2030 Artificial Intelligence will produce an increase in global turnover of 15.7 trillion dollars, generating a 14% increase in the world’s GDP.
Blockchain technologies will also play their part, according to this analysis by Gartner in the same year they will have an impact on the global economy equal to a further 3.1 trillion dollars, generating an increase of another 3% of the World GDP.
Given the volumes obtained by taking into consideration these technological ecosystems separately, it is easy to sense that a relationship between the two worlds, as part of the perfect integrating spirit of innovative enabling technologies, can lead to new and interesting scenarios.
Here are a few fundamentals about these technologies
When we talk about Blockchain, as described in detail in Engineering’s white paper: Unchaining business through the Blockchain, we refer to a family of technologies that allow the creation of distributed ledgers, the so-called DLT (Distributed Ledger Technology), which are defined as public or private based on who can read their content and are defined as permissioned or permissionless based on who can access them in writing. These ledgers are made up of a chain of transaction blocks and these blocks are encrypted, irrevocable, and shared by all the participants in an ecosystem. These are technologies to which, with some limitations, we tend to delegate trust with respect to the transactions which are stored therein. Many public Blockchains also have the characteristic of being able to “notarize” information, that is, to validate its existence from the moment this information was entered into the system and to guarantee its immutability over time. However, it should be clarified that “notarizing” is a very different thing from “certifying”, as there is no guarantee that the information entered in a public Blockchain is true, the only certain thing is that that information exists from the moment in which it was entered in the Blockchain and that from that moment onwards it can no longer be changed.
Public Blockchains have, in their most used implementations, two fundamental characteristics: they are completely decentralized, in the sense that there is no node in the network that is more important than another, and they are transparent, in the sense that the transactions are all public.
Artificial Intelligence, on the other hand, refers to the ability of some systems to complete tasks which would normally require human intelligence, such as the recognition of images and texts, the ability to make decisions based on accumulated experience and not only on the basis of rules written in an algorithm, strategic capacity, multi-context reasoning. Most of these applications are based on adaptive algorithms, that is, able to take into account experience, which use particular architectures called “neural networks”, actual computational models that try to replicate, in a very simplified way, the functioning of a biological neural network.
Neural networks, in their classic implementations, also have two distinctive features in total contrast to those of Blockchains: they are centralized, in the sense that the knowledge base is unique and not distributed, and they are black-boxes. This latter aspect is particularly interesting; in practice when a neural network provides its assessment, for example on the content of an image, there is almost never a way to understand what are the elements which contributed to the intermediate choices which then determined the final decision.
A few examples of potential future integrations
Blockchain as a knowledge base for AI algorithms
Many distributed ledgers have the characteristic, except for particular optimizations, of replicating the entire contents of the ledger on all the nodes of the network. Think for example of the Bitcoin Blockchain, the nodes contain all the Bitcoin transactions which have been carried out from the beginning, that is, since the first node was turned on and the first exchange was recorded. Blockchains can therefore represent extraordinary knowledge bases which, fed to specific neural networks, can be used profitably by dedicated Artificial Intelligence algorithms.
For example, think of Artificial Intelligence algorithms that need to use immutable data for regulatory reasons, such as those concerning the smoke pollution values of a factory, the speed reached by heavy transport vehicles, the contents of the regulations published in the Official Gazette.
Adaptive Smart Contracts
Smart contracts are automatisms that allow applying specific actions when particular conditions occur in transactions published in a Blockchain. The current implementation is mostly based on traditional algorithms, while in the future we could expect that some of these implementations may reach higher levels of complexity, right up to actual adaptive algorithms, typical of classical Artificial Intelligence methods. In these cases, it will be natural to use the entire knowledge base contained in the Blockchain.
The fact that the content in many bBlockchains is public could also pave the way for new initiatives based on the use as a service of adaptive algorithms implemented through smart contracts, creating proper marketplaces available to third parties.
Consider, for example, algorithms which automatically calculate the cost of an electricity charge, activate the economic transaction and calculate when and where it will be advisable to stop for the next charge based on the user’s driving experience and the selected route. These algorithms could also be granted to third parties who must perform similar calculations, acting on different data.
Distributed computing as a service
Blockchains, especially those based on Proof Of Work, have the characteristic of needing great computing power to encrypt the blocks containing the transactions. Likewise, many Artificial Intelligence algorithms are, under some conditions, hungering for performance due to the complexity of neural networks. This can trigger a real sharing of computational power which can be transferred from architectures which are at that moment emptier to architectures under pressure, with undoubted advantages of an infrastructural, energy and economic type.
Finally, there are hypotheses of integration, between Blockchain and Artificial Intelligence, able to mitigate black-box phenomena and the introduction of bias within neural networks. What is certain is that the combination of these concepts, on a technical and operational level, can open up new scenarios of enormous interest in the industrial sector.