Friends,
I hope everyone enjoyed a relaxing Memorial Day. Thank you to the families who allow us every year to pay tribute to their loved ones sacrifice. We can never thank them enough.
This month I want to tackle a question that is coming to us frequently. With all the advancements in AI, how can I trust what the AI produces? Can I trust Large Language Models (LLMs)? I should be clear that Straylight Systems doesn’t build LLMs, but more on that next newsletter.
During the last couple of months as we have attended industry conferences and meetings, there is a theme we continue to hear, “How does my organization trust the AI?”. Great question! You should be demanding an understandable or accountable framework so that the Head of Engineering and/or the CEO knows exactly what the AI is doing and how it is doing it.
Here is our approach, summed up as “Show Your Work, Cite Your Sources”:
Documentation (Show Your Work). Every prediction and decision Straylight AI makes is thoroughly explained and documented. This means that decisions can be traced all the way back to their source data – a literal answer to the question “where did you learn that?” This is in stark contrast with most LLMs that do not give sources or explain their reasoning.
Assessments (Scoring / Cite Your Sources). As well as tracing predictions back to source data, that data itself is scored along several dimensions of quality and reliability. We all know that predictions made from high-quality, authoritative sources are more trustworthy than those made from mixed-quality, less reliable sources.
Straylight is developing these frameworks and metrics to help our customers not only understand what the AI is doing but also shape the ways regulators or government are contemplating how the governance of AI could be developed. As our proprietary AI “engines” tackle real-time, structured and unstructured data to normalize, harmonize and operationalize that data, ensuring it is valid or qualified is paramount.
I wanted to share some updates and highlights:
Automating Data Quality: Want to see Straylight Systems in action? Check out the amazing video shown at SOF Week!
Customers: Straylight conducted a demo last week to a large Fortune 500 Insurance company that was eager to see how our AI automates the parsing of PDFs and flat files (CSVs) to accelerate their processing of claims. It was encouraging to see one of the largest insurance providers get stunned by our AI. Again, more to come!
Ask:
Know any growth stage or enterprise technology companies unable to find AI engineers or engineering projects taking months to build? Help us by introducing us to Directors, VPs and Heads of Engineering that need to accelerate their software development and lack AI/ML resources or capabilities.
The Straylight Systems team is excited to see how we can help. Thank you!
Straylight Systems is an Artificial Intelligence company and our core product is an AI-powered Intelligent Automation Platform that helps organizations automate data engineering and automate workflows to accelerate software engineering product builds for business users in both commercial and government.
Sincerely,
Sergio
Co-Founder