Latest Report Suggests Blockchain Can Monitor Bias in AI Models

• Blockchain Council
• September 13, 2024

In the ever-evolving landscape of artificial intelligence (AI), the quest for safety, fairness, and accuracy has led to an unexpected ally: Blockchain technology. Renowned for its foundational role in cryptocurrencies like Bitcoin, Blockchain’s potential to revolutionize AI has been largely underestimated, until now. Tech giants like FICO, a data analytics firm, and the Blockchain-focused startup Casper Labs are harnessing the power of Blockchain to oversee the development and training of AI algorithms, addressing critical issues of bias and transparency in this burgeoning field.

As AI continues to permeate various sectors of the business world, doubts about the reliability of its outputs have become increasingly apparent. In tandem, regulatory bodies have begun exerting pressure on companies to enhance transparency and accountability in their AI algorithms. Enter Blockchain, the immutable ledger technology, which could hold the key to creating a secure and verifiable record of AI model development.

Scott Zoldi, Chief Analytics Officer at FICO, elaborates on the potential of Blockchain in AI. “Blockchain can be used to track exactly what data an algorithm was trained on, when, by whom, as well as what other steps were taken to vet and verify that data,” he explained. While companies typically attempt to follow this data trail, Blockchain offers the promise of shared, consistent, and trustworthy records, significantly streamlining the process. It doesn’t prevent algorithms from exhibiting bias or malfunctioning, but it does provide an auditable record that sheds light on why such issues may arise.

Zoldi further emphasizes Blockchain’s ability to break down processes into smaller, transparent contracts. “There’s a level of transparency and honesty in having that immutability,” he noted. It’s not the ultimate solution for the ‘black-box’ problem of AI models, but it undoubtedly offers better records for diagnosing and addressing issues.

However, not everyone is convinced of Blockchain’s role in AI governance. Scott duFour, Chief Information Officer of Fleetcor Technologies, acknowledges the importance of AI governance but likened the use of Blockchain to “a hammer looking for a nail.” DuFour argues that while Blockchain could boost trust in AI systems, it should complement existing tools aimed at helping designers understand and interpret AI model predictions.

FICO’s Blockchain tool is currently in internal use, with plans to make it available to customers later this year. Meanwhile, Switzerland-based Casper Labs is collaborating with IBM to develop its Blockchain-based AI tool, offering a feature called “version control.” This function records the data and parameters influencing a particular model at any given time. If companies detect bias or inaccuracies in their models, they can revert to a prior version, according to Mrinal Manohar, CEO of Casper Labs.

Correcting bias in AI algorithms remains a challenging and time-consuming process for many companies, often due to a lack of the necessary systems and tools. Casper Labs’ tool, presently in beta testing, is expected to integrate with IBM’s Watsonx AI governance platform in the third quarter of this year. However, it won’t be exclusive to Watsonx, signaling a broader application within the AI governance landscape.

This innovative use of Blockchain in AI governance seems to encounter fewer hurdles than previous Blockchain applications, such as supply-chain tracking. In the supply-chain sector, established methods for tracking goods existed, making the transition to a complex Blockchain solution a hard sell. In AI governance, however, no such established methods are in place, providing Blockchain with a unique opportunity to establish itself as an industry standard.

Avivah Litan, Vice President and Distinguished Analyst at Gartner, suggests that while Blockchain in AI governance is a compelling idea, it might be ahead of the market. Many companies have not prioritized AI governance and risk management historically, considering it less important than other aspects of their projects. However, as the number of generative AI projects multiplies, the importance of AI governance is beginning to gain prominence.

Contrary to this view, Manohar asserts that the Blockchain-based AI tool is not ahead of the market. He highlights a positive response from corporate tech leaders who have tested the tool. He argues that companies might not prioritize AI governance today simply because they don’t know where to start, not due to a lack of interest or awareness of its future importance.

Nicolás Ávila, Chief Technology Officer for North America of software company Globant, acknowledges the need for the tools to prove themselves. Nevertheless, he recognizes the value in the convergence of Blockchain and AI technologies. He envisions a future where “AI is going to solve the problems of Blockchain and Blockchain is going to solve the problems of AI.”

As the adoption of AI continues to grow, so do concerns related to bias, transparency, and accountability in AI models. Blockchain technology, initially celebrated for its role in cryptocurrency, is now poised to address these pressing issues. FICO and Casper Labs, among others, are pioneering the integration of Blockchain into AI governance, offering the promise of transparent, trustworthy, and auditable AI models. While some skeptics question the readiness of this innovative approach, the potential for Blockchain to become the backbone of AI systems of the future is a tantalizing prospect. As the AI landscape continues to evolve, the marriage of Blockchain and AI may indeed prove to be a match made in technology heaven.