Opinion: Blockchain reliance on AWS won't bring transparency to AI

Author: Dominic Williams, CoinDesk; Compiler: Songxue, Jinse Finance

The rapid development of artificial intelligence (AI) has captivated the world, with many asking what's next for this technological breakthrough. **While AI has demonstrated its potential to transform various industries, it faces a major hurdle to mass adoption: a lack of trust and transparency. **

Decentralized computing via blockchain could alleviate the current trust problem, but there is a catch.

Dominic Williams is the founder and Chief Scientist of the DFINITY Foundation, a non-profit research and development organization and a major contributor to the Internet Computer.

Right now,** people have limited insight into AI models, and no real way to verify the sources of the data the models were trained on, what data the models actually collected, and how that data informs the models and their accuracy. **

Until there is a fundamental change in the transparency of AI programs and the infrastructure they are built on, users at all levels will not feel safe leveraging these models due to a lack of trust and overall skepticism.

The cross-application of artificial intelligence and blockchain technology offers synergies that will enhance both technologies and drive widespread adoption through their integration.

**Currently, since artificial intelligence requires massive computing resources and data sets, most blockchains lack the necessary infrastructure to support artificial intelligence models due to limited computing power. The limitation in computing power is partly due to the fact that most blockchains are not fully decentralized. **

**In contrast, many of the most popular blockchains in the world today rely on centralized cloud infrastructure (i.e. Google Cloud and Amazon Web Services), which prevents blockchains from supporting processing and storage at the speed required for artificial intelligence data capabilities. **

Despite the negative headlines surrounding current attempts to integrate artificial intelligence with blockchain, it’s not what it seems. The current integration results in AI running alongside the blockchain, rather than the intended goal of running AI on the blockchain.

**The core infrastructure and underlying technology of these "blockchain artificial intelligence" projects are mainly run on centralized servers, and use plug-ins to connect centralized artificial intelligence models to blockchains running on centralized cloud networks. This defeats the purpose of using blockchain technology to advance artificial intelligence, as it does not address the fundamental issues of trust and transparency. **

**A fully/fully decentralized blockchain, such as the Internet Computer (ICP), the network I helped build, that provides computing power that matches or exceeds Web2 cloud servers, would allow AI models to run entirely within smart contracts. This will make training parameters and inputs for large language models both open-source and tamper-proof. ** In order to achieve AI integration on the blockchain, we need a blockchain that can process data at a speed comparable to the Web2 cloud, and this can only come from complete decentralization.

Hosting the AI model itself on the blockchain allows the AI system to leverage the inherent decentralization to increase transparency in all aspects of the model. Therefore, AI on the blockchain is the next logical step for long-term success, as blockchain will enhance the trustworthiness, accountability, and security of AI, thereby enhancing trust among users.

However, until misconceptions about how exactly the two technologies work together are cleared up, the growth of the AI ecosystem will not reach its full potential.

** Realizing the full potential of AI on the blockchain requires a truly decentralized network. ** It must be able to store and process data so that the full model can run unhindered in the smart contract. These decentralized systems like ICP will change the landscape of AI development by enabling AI to function like an autonomous cloud.

Build Truth and Trust

For example, consider an artificial intelligence model designed for medical professionals. The model is widely used, but ends up producing untrustworthy responses. This is because there is no easy way to validate what training data the model is based on and how it is used.

This centralized model only produces outputs without insight into inputs. However, in a decentralized environment, large language models for AI can be constructed from only well-known medical textbooks and databases of reputable medical research papers.

When the doctor interacts with the AI, the hidden process is completely transparent, and cryptographic proofs guarantee what content the AI has been trained on. As a result, the generated responses can be verified and doctors can trust the results.

This example is just one of many that demonstrate that decentralization is critical to building trust in AI models. By operating in a fully open and public environment, AI on the blockchain ensures transparency of data processing, enabling users to understand how their data is being utilized.

Additionally, on-chain AI applications can all access and contribute to the same dataset, creating a collaborative ecosystem within the blockchain. The tamper-resistant and security features of the blockchain ensure that this data is also less susceptible to malicious misuse.

The collaboration between artificial intelligence and blockchain presents an excellent opportunity to advance both technologies and facilitate a more trustworthy and reliable exchange of information.