Lex Fridman Podcast

#151 – Dan Kokotov: Speech Recognition with AI and Humans

Jan 4, 2021
Dan Kokotov, VP of Engineering at Rev.ai, shares his expertise in automatic speech recognition technology. He discusses the challenges of real-time transcription, including accuracy issues with accents and pacing. Kokotov emphasizes the role of user feedback and data quality in improving ASR systems. He also explores the future of transcription services in the gig economy and highlights the importance of bridging human and machine efforts. Their conversation touches on the evolution of podcasting and the need for standardized transcripts to enhance accessibility.
Ask episode
AI Snips
Chapters
Books
Transcript
Episode notes
ANECDOTE

Dune's Philosophy

  • Dan Kokodov considers Dune the greatest sci-fi novel.
  • He finds the philosophical idea of needing pressure and suffering for progress fascinating.
ANECDOTE

Rev's Simplicity

  • Lex Fridman praises Rev for simplifying the transcription process, unlike his past experiences with Upwork.
  • He compares it to Isotope RX, another product that streamlined his audio editing workflow.
INSIGHT

Rev's Origin

  • Rev aimed to improve the Upwork model by standardizing service categories and simplifying the user experience.
  • They started with translation services and later added audio transcription.
Get the Snipd Podcast app to discover more snips from this episode
Get the app