As most people that follow this blog know, I’ve been involved in the world of startups for about two decades. Focus has been on technologies that can transform markets solving hard(er) problems. Several of these involved semiconductor-based hardware systems while others were pure software to make at scale compute more efficient and more performant. 

Over the past decade or more many of the software companies that I’ve been assisting have been AI/ML related. AetherBio (using AI/ML to optimize genetics for biological factories), Bitfusion.io (providing a platform for AI/ML workloads across CPU, FPU, and FPGAs/Hardware accelerators), OnSpecta (3X optimized performance for AI/ML Inference), a SBIR Phase-II for the US Navy, Salience Labs (photonics-based compute), and currently Austin-GIS (vision processing for security and advertising delivery). The technologies applied within AI/ML neural networks and processing systems have not only made made great strides in data set size and processing, but also some significant advances in formulation of problems. The first neural network I personally constructed was very, very limited and was based on a switched capacitor structure back in my university days and only mimicked a couple of synapses. The advances have been enormous and it is not hyperbole to say exponential with hundreds of thousands times the compute available today.

However, even as these advances have occurred and we are seeing broader applicability and adoption of AI/ML, we still face several monumental challenges that have not been addressed because ultimately AI/ML systems rely on content, data, and appropriate training and governance of operation. For a while and certainly during 2023 the challenge of trustworthiness of operation, authenticity, and providence of training content have emerged as key factors for AI/ML adoption and use. Even with all other hype and advances in AI, these factors have not been adequately addressed and only recently gained the awareness and concern that they warrant.

It was for this reason, that during 2023 I set out to study and attempt to address the issue of measuring the impact and value of content within an AI/ML neural network. The result of that study / work was an invention and associated IP filing titled “CONTENT ATTRIBUTION SYSTEM FOR ARTIFICIAL INTELLIGENCE AND MACHINE LEARNING SYSTEMS“. This algorithm provides the ability to look with the structure and weights of a neural network over time and determine how the neural network is affected by new training data sets, operation, and tuning. 

Authentrics.ai was formed to commercialize this technology and bring this capability to market. Importantly, it is not a new method of creating a neural network but rather it allows periodic measurement and monitoring of neural networks to determine the value of training sets, impact of each training set on the computation of results, and determine the contribution of each data set to the results that the neural network produces. The company will provide this as a SaaS and as an on-premise software solution and is applicable to any system that utilizes neural networks. If not obvious, this has application to companies who want need to trust their AI, optimize training processes, and value the content that they use.

More information about Authentrics.ai is available at our website and Linkedin company page. You can also reach out directly to John.Derrick@Authentics.ai.

Thanks for reading and if you have questions about the offering, the company, or how you could get involved, please reach out.