We have touched a few times (here, here and here) on the topic of ethical sourcing and regulatory compliance. Today, we will delve into the recent news of Accenture’s initiative: testing a novel technology designed to navigate the landscape of artificial intelligence regulation. This technology, developed by EQTY Lab, employs cryptography and blockchain to bring transparency into the origins and workings of large language models (LLMs), marking a potential milestone in responsible and safe AI innovation. Or not?
Focus On: Rethinking Transparency
The initiative shifts from traditional AI model assessments, which typically focus on outputs, to a more nuanced approach that scrutinises the models’ construction and operational integrity. This shift from output to input and process is not merely technical but represents an intentional move towards ensuring that AI models function as designed, thereby instilling a higher degree of certainty among developers and users alike. The ongoing evaluation of EQTY Lab’s AI Integrity Suite in Accenture’s AI lab demonstrates the scalability of such a solution, potentially benefiting a broad spectrum of corporations.
The Regulatory Context
This move towards enhanced transparency and accountability in AI development is in direct response to a growing global regulatory emphasis on AI. With various countries, including the US and the EU, actively proposing frameworks to mitigate the risks associated with AI while harnessing its potential, the industry is at a turning point. The challenge now is to bridge the gap between advocating for responsible AI and implementing concrete measures that ensure it.
A critical aspect of this discussion is the inherently dynamic nature of AI models. Unlike static products that can be easily inspected and deemed safe or unsafe, AI models are continuously evolving. They are constantly being updated with new data, algorithms, and usage patterns, which means their behavior and output can change over time. In short, every time an user inputs a prompt, even if just to say “Hallo”, the model changes in ways we cannot fully control or investigate. This fluidity poses a significant challenge for compliance and oversight, as traditional regulatory approaches may not be agile enough to keep pace with constant permutations of the model.
EQTY Lab proposes a solution that could balance the need for transparency with the protection of proprietary information. Through cryptographic signatures, developers could offer insights into the construction and evolution of AI models without disclosing sensitive details. This approach could serve as a blueprint for addressing the legal and regulatory complexities surrounding AI, particularly as models become more sophisticated and integrated into various applications.
Whether this approach will work or not remains to be seen, although it is clear that the regulatory will soon intervene in one way or the other to regulate this space, being careful to balance the need for safety and the one not to stifle innovation.
How to prepare?
Although we do not know how this regulation will look like, there are a few preparatory actions that can be taken by corporate leaders:
1. Prioritise Transparency: Look for ways to incorporate AI models due diligence to enhance transparency before purchasing decisions are made, especially for client-facing applications.
2. Stay Ahead of Regulatory Developments: Keep abreast of global regulatory trends and anticipate how they might impact your AI initiatives. Early preparation can serve as a competitive advantage.
3. Foster Responsible Innovation: Cultivate a culture that values ethical considerations in AI development, recognizing the importance of balancing innovation with accountability. To this extent, it is often better to go for a less capable AI model trained by a trusted provider than to chose the best one trained on dubious data sets. Adobe Firefly vs. Midjourney come to mind.
4. Contribute to Industry Dialogue: Engage in discussions about AI regulation to ensure that the perspectives and needs of your industry are represented in policy-making processes.
Follow me
That’s all for this week. To keep up with the latest in generative AI and its relevance to your digital transformation programs, follow me on LinkedIn or subscribe to this newsletter.
Disclaimer: The views and opinions expressed in Chronicles of Change and on my social media accounts are my own and do not necessarily reflect the official policy or position of S&P Global.