Synthetic Data with Alex Watson, Founder of Gretel AI

Watch Episode Here

Video Description

In this episode, Nathan interviews Alex Watson, founder and CPO of Gretel AI, about the company's work in synthetic data. They discuss why we need synthetic data, Gretel’s new pre trained tabular LLM that creates synthetic data on a zero shot basis, privacy techniques to prevent LLM memorization, and more. If you need an ecommerce platform, check out our sponsor Shopify: https://shopify.com/cognitive for a $1/month trial period.

SPONSORS:
Shopify is the global commerce platform that helps you sell at every stage of your business. Shopify powers 10% of ALL eCommerce in the US. And Shopify's the global force behind Allbirds, Rothy's, and Brooklinen, and 1,000,000s of other entrepreneurs across 175 countries.From their all-in-one e-commerce platform, to their in-person POS system – wherever and whatever you're selling, Shopify's got you covered. With free Shopify Magic, sell more with less effort by whipping up captivating content that converts – from blog posts to product descriptions using AI. Sign up for $1/month trial period: https://shopify.com/cognitive

With the onset of AI, it’s time to upgrade to the next generation of the cloud: Oracle Cloud Infrastructure. OCI is a single platform for your infrastructure, database, application development, and AI needs. Train ML models on the cloud’s highest performing NVIDIA GPU clusters.
Do more and spend less like Uber, 8x8, and Databricks Mosaic, take a FREE test drive of OCI at https://oracle.com/cognitive

NetSuite has 25 years of providing financial software for all your business needs. More than 36,000 businesses have already upgraded to NetSuite by Oracle, gaining visibility and control over their financials, inventory, HR, eCommerce, and more. If you're looking for an ERP platform ✅ head to NetSuite: http://netsuite.com/cognitive and download your own customized KPI checklist.

Omneky is an omnichannel creative generation platform that lets you launch hundreds of thousands of ad iterations that actually work customized across all platforms, with a click of a button. Omneky combines generative AI and real-time advertising data. Mention "Cog Rev" for 10% off.

X/SOCIAL:
@labenz (Nathan)
@AlexWatson405 (Alex)
@Gretel_AI
@CogRev_Podcast

TIMESTAMPS:
(00:00:00) - Intro to Alex Watson and Gretel AI's focus on helping build better data
(00:03:02) - Origins of the company name Gretel AI and initial vision around enabling data sharing while protecting privacy
(00:05:16) - Alex's background in data privacy and compliance from his previous startup Harvest AI, acquired by AWS
(00:06:37) - Early experimentation with language models in 2020 to recreate data distributions and improve ML model performance
(00:07:24) - Using synthetic data to create additional examples and improve detection of rare diseases
(00:12:50) - Why use synthetic data?
(00:17:02) - Sponsors: Shopify | Omneky
(00:19:00) - Training models to recreate real-world data distributions and using validators to detect unrealistic outputs
(00:21:30) - Generating tabular data row-by-row with transformers vs token-by-token with language models
(00:24:40) - Pre-training the Gretel tabular LLM on diverse internet data sets to learn good data
(00:26:27) - Challenges of limited context window size relative to full tabular data sets
(00:30:40) - Sponsors: Oracle | Netsuite
(00:34:00) - Using an agent planning architecture to break down large data generation requests
(00:37:40) - Having the agent determine when to use code vs the LLM for different parts of the data
(00:39:41) - Example use case of adapting models with synthetic data samples for long-tail cases
(00:43:00) - Using reinforcement learning to intentionally generate more diverse and representative synthetic data
(00:46:24) - AI Regulation: Biden’s Executive Order
(00:48:20) - The importance of alignment checks and controls while still providing model openness and flexibility
(00:51:16) - The potential of efficient, lightweight models compared to massive LLMs like GPT-4
(00:56:00) - Analogizing model specialization to specialized parts of the brain rather than ever-larger general models
(00:59:22) - Stochastic parroting vs reasoning
(01:02:18) - Focusing on solving data problems for users and iterating based on their feedback
(01:06:04) - Using differential privacy techniques to prevent memorization and exposure of private data
(01:14:37) - Adding noise during training to blur memorization while still allowing model convergence
(01:18:42) - Optimism that synthetic data quality issues reflect details not fully understood yet vs inherent problems


The Cognitive Revolution is brought to you by the Turpentine Media network.
Producer: Vivian Meng
Executive Producers: Amelia Salyers, and Erik Torenberg
Editor: Graham Bessellieu
For inquiries about guests or sponsoring the podcast, please email vivian@turpentine.co