The buzzwords of 2018 are Artificial Intelligence and Blockchain, two very transformative technologies.
Today AI surrounds us in any digital activity we do, be it replying to emails or shopping online. Algorithms can simulate almost any human behavior and decision-making.
Artificial Intelligence is a collection of technologies (such as machine learning, artificial neural networks, and deep learning) that let computers act independently and efficiently. AI aims to train machines to learn and apply the knowledge from large streams of data to make them mimic human intelligence – for example, cars could self-drive and computers can sense your mood by analyzing your emails.
AI depends on having huge amounts of data to learn from. The more data an AI system consumes, the more ‘intelligent” the self-learning algorithm becomes.
Blockchain technology has not pervaded us all yet in a visible way but is being adopted fast enough to be a part of almost every system in the next few years.
Blockchain is a growing list of records or blocks linked using cryptography. In a blockchain-based database, the operations on a block are contingent upon the agreement of all parties rather than a single point (a distributed ledger). It transforms inherently centralized systems with its use of decentralized database architecture. A blockchain is resistant to the modification of data, which is a key factor when you consider its integration with AI.
Today, as many businesses can see how AI can be implemented in their domain; the same cannot be said for blockchain. However, when the two are combined, possible solutions look interesting. Both AI and Blockchain are based on data sharing. Their integration can benefit in three areas:
- Data Sharing for training
- Security of shared data
- Building trust in AI decisions
More Data to Share
Blockchain stores data in an encrypted, distributed ledger format. This makes data safe and secure. Data can be read and updated only by those with permission. This is an important factor for data that is very personal and sensitive.
AI gets better as it consumes more and more data. It needs as many data sets as possible to learn better. This data can be highly personal and sensitive leading to privacy concerns. In such situations, Blockchain can provide the data to an AI system securely. The combination improves AI and keeps the underlying data secure.
Blockchain can bring about an era of open, free and secure data. While the Facebooks and Googles of the world have large amounts for their AI, for other companies, obtaining such large datasets is out of reach. Blockchain can enable access to many such datasets securely and make it accessible outside the organization. Since there is no single storage that can be hacked, data sharing is highly secure and safe.
On the other hand, a cryptonetwork for data can create a “data marketplace” that provides access to AI training data to anyone on the network that needs it. A marketplace can bring together entities offering their data with AI creators who consume it. The key thing here is that the “exchange” data is on a blockchain. There is no intermediary and the shared data is secure since data is shared via smart contracts.
Verifying AI Decisions
AI is being deployed in many business critical decision-making. AI decision-making offers advantages in several areas, but its outcomes need to be trusted by all. Also, sometimes, the decisions made by an AI may not be apparent to humans.
The decisions made by such systems have significant outcomes, for example, they could handle insurance claims. If the underlying decision-making processes are opaque, the combination of blockchain and AI can make these systems more human understandable.
If an AI system makes a decision, and it is recorded in a blockchain along with all the decision making logic, it will be shared with all the parties on the blockchain, and always be available for scrutiny and review. Decision-making failures can be analyzed easily this way.
If the decision-making logic or terms of the contract are not in a central server but spread on the blockchain, anyone can verify it, thus adding credibility to non-human-made decisions.
Blockchain can offer smart contracts that can be used in AI systems. A smart contract executes automatically when certain conditions are met on the blockchain. Such contracts can be used to trigger actions in domains like healthcare or insurance.
These decisions will also need human review at times of dispute. This is where putting the decision information on a blockchain can add credibility. For example, in the financial domain, AI is being used to check fraudulent transactions automatically. This task is simplified if decisions are recorded on a blockchain because the record stays secure between the time the information is saved and the decision is reviewed.
An Advanced Integration: AI Marketplace
Some very advanced combinations of AI and blockchain can be found in platforms like SingularityNet, which is a decentralized marketplace for AI services. Buyers and sellers of AI services use a set of standard AI software and hardware service APIs with the help of smart contracts.
They use blockchain as a tool for managing network transactions: a blockchain-based framework manages interactions between AI Agents and external customers.
While it looks like the Blockchain technology has the capabilities to enhance AI, Blockchain protocols are not optimized for efficiency. This means that AI decision-making can slow down when integrated with blockchain systems. Of course, these are the issues which can be resolved once the technologies show a promise.