Fireside Chat with Richard Socher, Chief Scientist, Salesforce
summary
A research group integrates published neural techniques into products, improves accuracy by 20% on a sequence-tagging task for contact/email matching, and ships layers of support: advice, prototype code, reusable APIs and a new voice assistant that updates CRM fields from meeting dictation. A no-code product Einstein builder predicts one column from others but enforces minimum examples and warns about confounders; ethical training via trail head is required because biased features (zip, income) can proxy race or gender.
AI-written summary · about
chapters (9)
- 0:00Introduction and welcome
- 1:49Career journey and Salesforce
- 5:19Academia vs industry in AI
- 7:05Balancing research and products
- 11:59Collaboration with engineering teams
- 15:51AI accessibility and ethical challenges
- 22:08AI education and ethical AI
- 26:14Progress in NLP and deep learning
- 31:29Closing and Q&A
related talks
22mDiscover the Future of Salesforce Development
18mCh. 2: Innovation
15mCode, Clouds, and AI: A Quick Look into Salesforce Developer Insights
22mFireside Chat: AI and Data-Driven Sales
38mHow Salesforce and AWS accelerate the contact center
22mFireside Chat
36mCh. 1: Corporate Narrative
59mPanel: Artificial Intelligence: Who is Winning?Panel d4