AI and the Blockchain

What happens when you take two of the hottest current technologies, AI and blockchain and combine them? AI and blockchain are still both experimental technologies and whilst both have incredible potential, either could ultimately end up disappointing. Exploring the intersection of both technologies could well prove futile.

In tech, a game changing development tends to appear every 10 to 15 years, from the silicon chip, to the pc, to the internet to web 2.0. We are probably nearing the end of the current trend. Blockchain and AI technology could well underpin the next generation of tech giants.

As with most new trends they are often seen initially as experimental and over-hyped before eventually taking hold and establishing a new paradigm. Amara’s law states that:  the impact of new technology is often overestimated in the short run and underestimated in the long run and this could well be true here.

It does appear that “AI, blockchain and the Internet of things” could well be the new “Social, Cloud & Mobile” Those trends are still very much emerging, but their potential impact is massive.

Blockchain and AI may seem like very different technologies. Blockchain for example, is decentralised, open, deterministic and transparent whereas AI is generally centralized, closed, probabilistic and blackbox in nature.

Currently advanced AI’s are the preserve of the major tech companies. These companies have access to unprecedented amounts of data and advanced hardware to train their systems. Giving them a huge competitive advantage over any new startups. However, several projects are attempting to build blockchain based AI and this move towards open source AI could help redress the balance of power in the AI sector.

SingularityAI is an AI blockchain concept that aims to combine multiple simple AI onto a single blockchain based platform. Its goal is eventually to let its users access multiple pre-trained artificial intelligence systems covering a wide range of artificial intelligence tasks such as face recognition, weather prediction and trading bots. A less capable machine can lean on this repository of AI, with the intention of eventually creating a human level Artificial General Intelligence. This project has the potential to democratize AI research and development and remove our dependency on the big tech monopolies.

Potentially the most promising and most worrying use for AI is with DAOs (Decentralized Autonomous Organisation). DAOs aim to hard-code the rules and procedures a company uses from the get go using software. This could for example, involve setting aside a certain percentage of earnings for a cause or the process via which such a DAO rule could be changed. This seems to be how a normal corporation would work. The main difference being that these rules are enforced digitally.

DAOs and AI can be combined in at least three ways.

With AI at the edges of smart contracts

With AI at the centre of smart contracts

With swarm intelligence and many dumb agents combining to create emergent AI complexity.

When AI is at the edges of smart contracts it essentially replaces the human component in a DAO. They will be intelligent decision-making entities and interact with a dumb core i.e. a smart contract or clearing house of value. An example of this would be an automated AI trading bot or a fully autonomous AI agent present in Minecraft.

When AI is at the centre of smart contracts we would generally expect to see a much more complex AI. The kind we have witnessed with DeepMinds AlphaGo. It would have its own built in feedback loop allowing it to continually update its state and react dynamically to changes within its DAO. A smart contract manager could for example aim to put the right people in touch with each other to produce a desired outcome.

The third way to use AI in a DAO would involve the creation of multiple simple AI agents using smart contracts to interact. Whilst any one component on its own would be relatively “dumb” if the agents act collectively it can result in emergent intelligence. Much like the way a bee’s nest or ant colony self-organises.

The opportunities for AI based DAOs to flourish are plentiful. Imagine for example a decentralized Uber type service where the cars were not only self-driving but were fully autonomous.  

There are inherent problems to creating distributed AIs. For example, you could program an AI DAO to generate as much art as possible. The AI could teach itself to create art that could be sold. With the proceeds it could purchase itself additional computer power with which to further its cause. As it became larger and more successful it would become harder and harder to stop until eventually it became the wealthiest entity in all history. It would be indestructible as such as it lives on a blockchain. It will be up to developers to investigate exactly how DAOs function in the real world and build in safeguards to prevent this kind of runaway AI.

In conclusion, a lot of what I have discussed is as exciting as it is experimental and it’s unclear whether we will be able to achieve the ambitious goals we are pursuing.

However, a lot of foundational work has been done, and the combination of AI and blockchain could create some very powerful technologies, at speeds that are hard to fully predict.  Some of those advances may well create undesirable results. Now is the time when we must fully consider the implication of building unkillable artificial intelligences.

For more articles visit http://www.icocrowd.com