Building Trust in AI Systems from the Start
Sanjay Rao
Generative AI | prev McKinsey, Microsoft, Norwest Venture Partners | Strategy, Product, Venture Capital
As artificial intelligence becomes embedded into corporate operations, outputs from software are increasingly model driven versus programmatic.? In the model world, data utilized to train an AI system, fine tuning, and hyper parameterization among, other items, can significantly influence output. Ensuring the software engineering model of functional and unit tests are updated to assess model output is key.? Bringing transparency to the input factors of AI models will build long term trust.
1)????? End Users are Concerned About Trusting AI
According to a 2023 KPMG survey (Trust in Artificial Intelligence: A global study (kpmg.com), 61% of those surveyed were concerned about trusting AI systems. At the same time, there is a tremendous opportunity to build trust from the start - 71% of those surveyed were either willing to trust or ambivalent about trusting AI systems, while only 29% were likely not to trust AI.
2)????? Learnings from Regulations such as GDPR
Regulations such as GDPR led to a number of privacy and compliance focused startups. ?Indeed, according to the same survey 71% of believe that AI will likely be regulated.
Companies leveraging AI today, can help reframe the conversation by self-regulation.
3)????? Visibility in Data Sources and Allowing 3rd Party AI Audits
Self-regulation today can include describing data sources and leveraging a 3rd party AI audit service.
Tau Ventures is an early stage AI focused venture capital fund. This is a purposely meant to be a short article to start discussions.
Image Pixabay CC.
CalmWave CEO | Founder #healthcare #ai4good #nurses #patients
1 年Couldn’t agree more which is why we do what we do at CalmWave.
SENIOR HEALTHCARE INDUSTRY EXECUTIVE | HEALTH TECH STRATEGIST | PRODUCT BUILDER | CLINICAL DATA & AI SME
1 年Fully concur—a couple of thoughts. - As AI-based end applications explode, there is a massive opportunity for the proverbial shovels, i.e., tools and infrastructure, to ensure trust, quality, fidelity, and other enabling attributes. - As organizations push to leverage AI's market potential with their customers, it is super critical that they put equal focus, if not more, on leveraging AI in their internal value chains and decision-making. I often see that organizations miss this crucial step and, in the process, miss out on highly valuable learning opportunities.