Computer Vision: Revolutionizing Multimodal Analytics and Insights

With the rise of visual data In the realm of inbound telemetry, computer vision has the potential to transform analytics. It automates the extraction of information from videos and images.

By providing real-time analysis and visual insights into data, computer vision is a transformative technology that will drive decision-making, automate data extraction and optimize processes in businesses, according to Ayush Kumar (photo), associate principal data scientist at IBM Corp.

“I think we are moving toward a much more multimodal future than today,” Kumar said. “We will see more information appear in the form of computer vision OCR that we perform on our own enterprise data. On the business intelligence side, I think there is a significant shift in the way information is consumed. Traditionally, we have more reporting, more visual, but at the same time we will have more agents, more granular information and deeper analytics that we will also pull from these systems.

Pradhan spoke to CUBE Research John Furrier to the Media Week NYC: the CUBE + NYSE Wired 2024 eventduring an exclusive broadcast on theCUBE, SiliconANGLE Media’s live streaming studio. They explained why artificial intelligence sectors, like computer vision, are game changers.

How computer vision will shape a multimodal future

With increased user engagement and diversity of communication channels, such as interactive formats, audio, video and text, the future is expected to be multimodal. As a result, computer vision will play an instrumental role based on the automation of manual tasks, with model training techniques, such as quantization and pruning, being helpful, Kumar pointed out.

“We have trained these LLM models in terms of in-context learning or retrieval-augmented generation, and training or fine-tuning these models is becoming more and more expensive,” he said. “I think in the future the cost of training will really come down. We’ve seen effective techniques like pruning and quantization that bring the size of these models to a point where we can train them.

Since generative AI has materialized as a new category, disruptive enablement, process change and technology stack adoption are required. As a result, IBM is enabling the transition of this cutting-edge technology through data curation and multi-model and multi-cloud approaches, according to Kumar.

“There are a lot of changes in terms of generative AI,” he said. “You see a lot of development in different places. The data side where we are looking at more data curation to power these models and LLMs. We’ve kind of figured out how data lakes, data warehouses work, how to enable businesses to have structured data, but when it comes to unstructured data, we’re still trying to figure that out. With IBM, we have multi-cloud and multi-model, but we also have a data curation process and accompanying tools to make that journey easier.

Here’s the full video interview, part of SiliconANGLE and theCUBE Research’s coverage on the Media Week NYC: the CUBE + NYSE Wired 2024 event:

Photo: SiliconANGLE