Artificial intelligence and blockchain: double hype or double benefit

Blockchain and artificial intelligence (AI) are on the lips of leaders of both small and large IT companies. They are regarded as revolutionary technologies that will transform various industries. Both have undeniable advantages, yet each comes with a number of implementation challenges.

Efforts to combine these two ingredients may seem like another marketing ploy. Yet there are serious grounds for considering a combination of the two technologies to address pressing tasks.

Today, AI tools are centralized and concentrated in the hands of large companies, such as Google, Amazon and Microsoft. To obtain a ready-made AI-based business solution, the end user must place full trust in the central authority.

Companies face a choice: to trust or to develop bespoke solutions. There is a third option — using blockchain to create and disseminate AI models. Here is how it could work.

Decentralisation for data and know-how exchange

One of the primary values blockchain can offer AI is the distribution of machine learning (ML) models among all network participants without intermediaries.

Facial recognition technologies provide a good example. If one device knows what a person looks like and uploads the data to the chain, other connected devices will also be able to recognise that person. And when another device uploads its own facial-recognition data to the distributed network, other users can use them in their machine-learning models.

Using blockchain allows decentralisation of data away from a single company. With such an approach, AI algorithm and software developers can train faster, share data securely, and collaborate effectively.

Beyond data, blockchain enables developers to exchange ML models. For example, small online retailers could use a single recommendation system for their platforms, upload data about their customers to train the algorithm, and compete with large online marketplaces like Amazon, which develop such systems on their own using vast amounts of data about their customers.

Solving the “black box” problem

Blockchain can also help illuminate the logic behind neural networks and address the so-called “black box” problem.

Today developers do not know what happens inside a neural network or how it arrives at a given conclusion. This complicates troubleshooting when something goes wrong. Developers simply do not understand which element of the algorithm they should investigate to fix it.

With blockchain, developers can record every operation, every action of the neural network, giving them the ability to revert to any moment of its execution and see what went wrong and promptly fix the fault.

Addressing the “black box” problem will enable researchers and developers to create a robust and efficient infrastructure, for example, для автономного транспорта.

Artificial general intelligence on the blockchain

Accumulating diverse models and data in the network could advance the development of artificial general intelligence — the concept whereby machine intelligence could surpass human intelligence.

If machine-learning algorithms are combined, training data aggregated, and new features regularly added to the blockchain, developers could build a complex system capable of performing not a single specific function but a range of diverse tasks and approaching human intelligence.

This is what SingularityNET is pursuing, launching the world’s first decentralised marketplace for AI models. Its solutions underpinned the humanoid robot Sophia, which for five years has travelled the world, given interviews and spoken at venues of international organisations and forums.

And, despite the fairly justified criticism of the bot in the scientific community, the SingularityNET software platform gives Sophia the ability to sustain the image of a rational and learning humanoid.

Conclusion

Artificial intelligence and blockchain have different histories, prerequisites for emergence, and evolutionary paths, yet much binds them. Both technologies have become responses to the challenges of the modern world.

The centralisation of data and ML models has led to neither business nor the public being able to fully trust AI algorithms. Their opacity hinders the spread of intelligent systems and raises a number of ethical questions.

Blockchain can restore and strengthen trust in AI. Moreover, the use of distributed networks could take the technology to a completely different level, bringing developers and researchers closer to understanding how algorithms reach decisions.

Trust, transparency and decentralisation herald a new era in the creation of reliable and secure applications capable of improving everyday life.

Subscribe to ForkLog news on Telegram: ForkLog Feed — the full news feed, ForkLog — the most important news and polls.