An overwhelming majority of businesses today have either adopted or are in the process of adopting AI to drive more efficient business outcomes. Yet successful implementations remain elusive, largely constrained by an inability to effectively audit and re-train AI systems when performance issues inevitably arise.
On November 2, a distinguished panel of AI experts from Casper Labs, IBM, watsonx and Gartner® will address the following key challenges:
- How to successfully navigate the Brute Force problem with generative AI
- How to realize greater visibility into AI behaviors – and how to modify them
- How to implement effective AI governance
Zooming in On Key Challenges
For all of its immense promise, AI – especially generative AI – isn’t going to be a feasible option for most businesses until they can more reliably audit and modify AI systems. For a more detailed overview of what the ideal risk management model for AI looks like, it’s worth looking at the NIST’s comprehensive standards framework, which has emerged as a guiding north for the entire industry.
Realizing these standards ultimately comes down to achieving greater transparency into AI training data; today, that environment is essentially a black box. When an AI hallucination occurs, it’s prohibitively difficult if not impossible to verify why and where it happened – which makes addressing the problem a non-starter.
Blockchain technology offers the most cost-effective and tamper-proof approach to realizing this critical level of visibility. When AI is augmented with blockchain, it’s possible to reconcile and track which data caused which outcomes, when and why. It also unlocks version control: when a given AI system falters, there’s currently no reliable method to “restore” a previous, working iteration. Think of this as akin to Google Docs’ beloved “restore previous version” feature, just on a much grander scale.
While the economics of such an approach are challenging in a traditional public blockchain environment, the advent of hybrid blockchains eradicates that concern. By hashing key data and storing it on-chain, organizations retain a tamper-proof methodology, while maintaining sensitive and/or extraneous data in more cost-effective private environments.
Here’s a more detailed look into how blockchain technology is increasingly being used to augment AI systems:
If you’re interested in learning more about Casper Labs’ specific efforts to address this challenge, be sure to join our webinar, featuring IBM, watsonx and Gartner on November 2 at 12 PM ET. You can register here.
