#134 WaveForm AI
Subscribe to get the latest
on 2023-04-20 00:00:00 +0000
with Darren W Pulsipher, Logan Selby,
In this episode Darren Pulsipher, welcomed Logan Selby, the co-founder and president of DataShapes, where they discuss a unique approach to Artificial Intelligence that is bucking the trend.
Keywords
#ai #machinelearning #edgecomputing #anomalydetection #technology
Logan is a former intelligence officer in the DOD and has a passion for AI and robotics, which led him to be a reservist in the defense community. He transitioned to DataShapes, where they have a unique technology that solves critical needs in the defense community using AI. While AI has been around for a long time, the vast amount of data available for training models and the advancement of computing resources have led to the development of smarter systems like chatbots and large language models.
Current techniques are resource-hungry and very costly to train and build a general-purpose inference solution. For example, running large language models like ChatGPT can cost up to $3 million a day, but AI is evolving rapidly and receiving more attention than ever. One concern in using neural networks is the inability to audit and explain how the AI arrived at a result. There are social, political, and legal challenges to trusting decisions made by these networks, particularly in fields where human life is at stake. However, society will eventually overcome these challenges and fully embrace AI. DataShapes’ approach, which uses traditional machine learning techniques to solve pain points in data analysis, has full auditability and discovery in their trained models.
DataShapes has a unique approach to solving problems quickly and efficiently in resource-constrained environments. While traditional neural network training requires a lot of labeled data and can be brittle, DataShapes uses methodologies to learn in real-time or near-real-time. Their technology focuses on waveforms and signals and is auditable, making it ideal for use in austere environments where people are getting dirty and carrying a server stack is impossible. While neural networks are great at language models and image recognition, Logan’s company is hyper-focused on waveforms and signals. Their approach is different and highly effective.
Detecting different types of waveforms and relationships in the waveforms is at the center of this new technology. This approach detects patterns harder to spoof compared to traditional techniques used by the Department of Defense. This has the potential for applications of their technology in electronic warfare, including gathering intelligence and analytics. Additionally, the platform can detect, analyze, and gather intelligence, which can be exported to edge devices. Their self-learning anomaly detection feature, Infinite Loop, was also mentioned, which establishes a continuous baseline based on the parameters the end-user prescribes. The technology can be used in the automotive industry, healthcare, and entertainment industries, where it could be embedded in every sensor.
DataShapes has a product called GlobalEdge, an intelligent agent that sits on or behind sensors to conduct ETL operations for data being collected. The machine learning component of GlobalEdge filters the data to provide relevant insights and anomalies in real-time, reducing the amount of irrelevant data being pushed back to headquarters. The product can also be used for data compression on the edge to the data center. The software can scale down to as little as 47 K, making it suitable for a variety of applications, including virus detection using UV waves.
To find out more about DataShapes and their approach check out https://www.datashapes.com