Elliot Shmukler (@eshmu, Co-Founder/CEO @anomalo_hq) talks about the impact of data quality on AI, how unstructured data can be improved, and how monitoring of data lakes can help prevent model drift and give organizations confidence with predictable results.
SHOW: 945
SHOW TRANSCRIPT: The Cloudcast #945 Transcript
SHOW VIDEO: https://youtube.com/@TheCloudcastNET
CLOUD NEWS OF THE WEEK: http://bit.ly/cloudcast-cnotw
NEW TO CLOUD? CHECK OUT OUR OTHER PODCAST: "CLOUDCAST BASICS"
SPONSORS:
SHOW NOTES:
Topic 1 - Elliot, welcome back! It’s hard to believe it has been 3 years since we spoke! Give everyone a brief introduction.
Topic 2 - Here’s the problem I see when it comes to AI adoption today. There isn’t an “off the shelf” AI model with an organization's data built in; that’s impossible. So, you must bring this data, often unstructured, to the model, often with mixed results. Do you agree?
Topic 3 - I see data quality in two ways… the quality of the data before ingestion is one way, we want the data to be clean going in. But, we also need a way to detect, mitigate, and do a root cause analysis for quality checks along the way, correct? Give everyone an idea of what this life cycle looks like.
Topic 4 - What are you seeing as the barriers to adoption? Is it the tools, the models, the need for RAG pipelines, the lack of data scientists, and AIOps?
Topic 5 - We have this crossroads where proprietary data makes an organization unique, but exposing that unique data puts the organization at risk. How much of a factor does this play, and how do you advise organizations around this complex intersection
Topic 6 - There is always this concept of predictable results. This answer should be consistent and repeatable. We’ve seen things like model/data drift and hallucinations hinder this concept, leading to a lack of confidence in the results. How do you advise organizations to tackle this lifecycle management and predictability over time?
FEEDBACK?