Can blockchain technology help mitigate the black box phenomenon of AI applications?

Abstract

In the rapidly evolving landscape of cutting-edge technologies, AI-driven business cases and blockchain solutions are like two relatives who don´t know each other. The reason is that the AI-inherent black box phenomenon stands in contrast to the transparency offered by blockchain technology. The term “black box” in the context of AI refers to the intransparency of AI systems where the internal workings and decision-making processes are not easily understandable. The general opacity of neural network-derived models, and particularly the recent wave of generative AI models (sometimes referred to as “black box” models) can lead to challenges in trust, accountability, and explainability. Users and developers often find it difficult to interpret how specific inputs are transformed into outputs, making it hard to diagnose logical and semantic errors (including hallucinations) or biases in the results. On the other hand, blockchain technology is celebrated for its transparency and immutability. Every transaction is recorded on a public ledger, accessible to all participants, ensuring a high level of trust and security. This transparency is particularly appealing in applications where accountability and traceability are paramount.