What are the links between blockchain and artificial intelligence?
Few years ago, after the San Francisco tech expo, I described blockchain as a promising technology but I said that it would have a hard time finding applications outside of the financial world.
Fortunately, I’ve grown up since then!
So has the blockchain world.
It has grown so much that it can now help us revolutionize artificial intelligence, and that’s the subject of our article!
About the vocabulary
As with all the cutting-edge technologies, it is sometimes difficult to fully understand the lexicon. To clarify this aspect, the first section will allow me to define the different terms used.
You can skip this part if you already familiar with the world of web3.
What is web3?
More than a new technology, web3 is a real revolution, a paradigm shift in the world of web.
Web1, web2
To better understand web3.0, it is necessary to go back in time and resume the different phases of the web.
Initially, the Internet was an online encyclopedia that allowed to consult information but without interactive content. The content was limited to texts or very light images. The connection was incredibly slow, downloading music would have taken you a whole day.
I am too young to have known web1, but I know very well web2, its advantages, its limits and its dangers.
Web2 is the web of cookies. In addition to reading and consuming content, we can interact with it. Websites can collect our usage data to send us targeted ads, we can pay online and interact with other users.
The improvement of internet connection infrastructures, notably thanks to optical fiber, 3G and 4G, has made it possible to offer more diversified content, with videos, sounds, images with a better quality and even 3D content.
The web2 is the age of social networks, we no longer just consult the internet we’re part of it.
The huge increase in the number of users has led to an explosion in the amount of data generated by web2.0, so much that users’ data has become the basis of many tech giants’ business models.
Web2.0 is the symbol of centralization, a handful of companies control the data, and the digital world. The data of more than 80% of the population is stored in the data centers of Google, Amazon, Twitter, …
Internet is broken
Gavin Wood, Co-founder of Eth
All these reasons push us to revolt. As in the 1776’s America, the time to bring down the kings has arrived…
… just joking 😆
… but let’s still find an alternative, just to be sure 👀
Web 3.0
Web3.0 is based on the model of peer to peer, or decentralization, on which the blockchain is based.
Rather than data being centralized in the data centers of large companies, it is stored in a decentralized and distributed among all network participants.
What is an NFT?
An NFT, or Non Fungible Token, is a digital asset that can be an image, a sound, a video, etc. It is attached to at least one network participant.
It is attached to at least one owner via a digital identity (wallet). Each NFT is associated with metadata that contains its basic information, such as the identities of the owner and creator.
These metadata are registered in a blockchain, which allows to secure and make more transparent the NFT transactions.
Even if the basic use of NFT is purely artistic, it has a lot of applications in different industries. LVMH, for example, associates each luxury handbag with an NFT that the owner of the bag can transfer to the buyer if he sell it to him. This allows to fight against counterfeiting.
What is metaverse?
Web3 and NFT will both contribute to the conception of metaverse.
Metavers, put under the spotlight by Zuckerberg, the CEO of Meta (ex Facebook), are virtual spaces that will create new forms of interactions between users.
It is a new form of Internet that is closer to reality. We can attend virtual events completely recreated in 3D (like concerts or weddings), we can watch sports events in a more immersive way, work with colleagues, study from home with a VR headset, etc.
It’s true that it can be scary said like that, but well, if we don’t create our metaverse we’ll be walking around in someone else’s metaverse (like Baidu for example 👀).
How is AI used for blockchain?
There are tons of applications of artificial intelligence for all technologies that rely on blockchain.
AI is involved in cryptocurrency trading, NFT generation or even metaverse construction.
Cryptocurrency trading
I hesitated to start with this application, I don’t want to feed the blockchain=cryptocurrency shortcut.
Crypto-currencies are known for their high volatility which makes predictions very difficult. Nevertheless, some time series analysis techniques such as LSTM can predict market trends, even if the models are less powerful than for the stock market.
Automatic generation of NFTs
One of the applications of artificial intelligence uses automatic generation of digital art objects. Although the technique has been around for a long time, it has gained momentum with the explosion of NFTs, which make it much easier to sell digital creations.
GAN networks are often used for this.
3D reconstruction of objects for the metaverse
In addition to generating NFTs, computer vision makes it much easier to model real-world objects in 3D.
The metaverse will recreate environments that has to be as realistic as possible. To enable this, we must be able to reproduce complex objects and shapes systematically and quickly, without having to use overwhelming softwares at each time.
Today, the best 3D reconstruction models are able to model an object in a rather realistic way from 2 or 3 photos.
How is blockchain used for AI?
The use cases of artificial intelligence for blockchain were easy to find, as AI is already quite mature and research on this topic is already in place.
The use cases of blockchain for AI are less obvious, although an entire field of artificial intelligence, which is called decentralized AI is emerging.
Decentralized AI stems from the need to secure the data used to train a model. Decentralized and encrypted storage ensures data security, prevents tampering or loss and make model training and usage more inclusive.
Data traceability
Blockchain can improve data traceability.
The strength of blockchain today relies on its security. Which makes it usable for large financial transactions.
When training an AI model, we need to validate the origin of the dataset. On the one hand to better assess the quality of the data and avoid training a biased AI, on the other hand to make sure that the data has been collected in an ethical way.
I don’t think the solution already exists, but I can imagine a decentralized dataset exchange platform that would work with NFT. When we transfer a dataset, we transfer an associated NFT.
This operation would allow at any time to check the origin of the dataset and to go back to its creation. Each action performed on the dataset could be recorded and visible.
Sharing Updatable Models on the blockchain
Still with the aim of making the use and training of models more ethical, the blockchain can offer more transparency regarding the models that are proposed on different platforms and facilitate collaboration.
Microsoft has created a framework, Sharing Updatable Models, for the collaborative maintenance of machine learning models.
The machine learning model is initially built via a smart contract on the Etherum blockchain and it is freely available, you can use it by paying the gas fee of the network.
The user can then modify the model, or add training data to make it more robust. The modifications are then tested in an automated way, from an initial test dataset which remains unchanged. If they have contributed positively to the performance of the model, the user is rewarded, otherwise he is penalized.
The rewards are given either in the form of points on the platform, or through financial compensation. The principle of points is inspired from Stackoverflow, and relies on the choices of other users to validate or not the proposed changes. The financial compensations are calculated via another public smart contract, and are based on several criteria such as the performance gain or the size of the dataset provided.
As often in the world of blockchain and web3, the success of this type of project relies on the ability to create a dedicated and trusted community.
For more info I invite you to read the blog post on Microsoft’s website or look at the research paper directly, Decentralized & collaborative AI on blockchain.
The stakes linked to artificial intelligence and blockchain go beyond the purely technological framework and will challenge the entire web economy. The end of cookies implies the end of targeted advertising. Therefore, several web giants will have their business models challenged.
AI, as in all fields, will be an important tool in the new world. Both for its construction and for maintaining it.
Conversely, blockchain will make AI more secure, ethical, and collaborative. We will now talk about decentralized AI. Serious problems, such as data anonymization for many industries, which are still difficult to handle today, could be solved via blockchain.
