Keynote Summary of the Data + AI World Tour from databricks

Arsalan Tavakoli and team roll out their vision for AI

If your company is called “databricks,” one might expect a data-centric vision of AI at their conference, and you would be correct. But that doesn’t make it the wrong approach. As we have been saying for a long time, “garbage in, garbage out,” meaning you need to start with good data as a necessary, but not sufficient, condition for achieving good results.

Executive Summary

• Arsalan Tavakoli, SVP of engineering, and the databricks team advocated for a data-first approach as essential for quality AI outcomes.

• The conference emphasized the importance of a “single source of truth” and stringent data governance for AI's scalability.

• Tavakoli's keynote highlighted the democratization of AI, likening the proliferation of large language models (LLMs, like ChatGPT) to historical tech advancements like PCs and internet browsers.

• Stan Guzik of S&P Global stressed efficient data management through an "enterprise data catalog" for informed corporate decision-making.

• The Unity Catalog aims to democratize data across roles while maintaining security and compliance and enabling integration with external data.

• The event concluded with discussions on secure data sharing protocols, a new databricks marketplace for data exchange featuring industry giants, and the concept of a data clean room to prevent data exfiltration.

The Data+AI World Tour provided insights into data management and artificial intelligence (AI). Over 1,000 industry experts attended the conference, which highlighted the long-term implications of data and AI at scale and underscored the necessity of a single source of truth in data and solid data governance.

Arsalan Tavakoli, databricks' co-founder, gave the keynote address. Tavakoli's speech focuses on the democratization of technology through innovations like ChatGPT. He draws historical comparisons, such as the development of computers from the 1950s, to personal computers in the 1980s, and the internet's spread from its inception in the 1970s to widespread accessibility with browsers in the 1990s.

Stan Guzik, CTO of S&P Global, offered a corporate view on leveraging data for decision-making. Guzik's presentation illustrates aligning and standardizing complex data from various commodity types for efficient delivery. He introduced the concept of an "enterprise data catalog" to improve data access and management.

Returning to the stage, Tavakoli expanded on the methods to make data and AI democratic within organizations. He emphasized the necessity of making data available to every employee, turning English into the newest programming language. Tavakoli warned against overly simplifying this integration due to the distinct challenges each company's jargon and structure present. He talked about a "knowledge engine" that comprehends context, security, and privacy, and that can evolve over time.

Tavakoli explained the Unity Catalog's role in rendering data usable for various groups, such as engineers and marketers, focusing on preventing data breaches and ensuring regulatory compliance. He spoke about the ability to merge external data into databricks, spot model drift, and perform root cause analysis on data quality issues.

The conference ended with a spotlight on the evolution of data sharing and protection. Tavakoli introduced an open protocol for data sharing and a marketplace for data and its consumers, including key players like Bloomberg, Dun & Bradstreet, and ICE. He also discussed a "data clean room" as a way to securely share data with a third party while reducing the risk of data exfiltration.

Overall, while the conference centered on databricks’ technology, it was well worth that in order to hear from a number of accomplished professionals in the field.

Copyright © 2023 by Alec Crawford         

#ai  #airisk  #innovation #STAYblog