85% of Data is Unstructured and Not Ready for AI Use, Industry Experts Say

IBL News | Santa Clara, California

Artificial Intelligence (AI) technology in enterprises has become mainstream. This is one of the main conclusions of the AI & Big Data Expo North America conference, which started yesterday and will continue today.

This in-person and virtual event, hosted in the Santa Clara Convention Center, attracted thousands of attendees — over 5,000, according to the organization of the event. Engineers and business leaders explored the latest innovations and explored impact of AI across industry sectors.

The adoption of AI has surpassed the 60% of the corporate landscape. The average adoption rate across all geographies was 56% in 2021, up to 6% from 2020, according to data shown by Daniel Wu, Head of AI & Machine Learning Commercial Banking and JP Morgan Chase. The banking executive, who participated in an opening talk, agreed on the fact that data is still pretty siloed. He highlighted that the main challenge in this regard is legacy data with inconsistent quality, and a lack of proper data models increased with the fact that cloud and on-premise hybrid systems end up duplicating it.

Henry Ehrenberg, Co-Founder at Snorkel AI [in the first picture below], revealed that 85% of data is unstructured, unlabeled, and not ready for AI use.

Currently, data governance is top of mind for every business leader. Other trends are, according to Mohan Reddy, Co-Founder and CTO of SkyHive and Associated Director of Human Perception Lab at Stanford University :

  • Unified Analytics
  • Graph Neural Networks and Enterprise Applications
  • The emergence of no-code AI Platforms
  • Multilingual models. Transformers as a big part of Enterprise NLP
  • Federated AI Strategies in large enterprises
  • New business models

Mohan Reddy [in the second picture below] elaborated on how MLOps and AIOps are gaining traction. He also stressed how organizations are looking to build AI trust by establishing protocols for sourcing, handling, and using data for developing ethical AI solutions and preventing algorithmic bias in outcomes.