Razib Khan's Unsupervised Learning

Nikolai Yakovenko: the Singularity is not here

Jan 14, 2025
Nikolai Yakovenko, a former professional poker player and research scientist at Google, now CEO of DeepNewz, discusses the state of AI as of 2024. He highlights the seamless integration of AI into daily work, particularly in software development, and how this transformation counters fears of mass unemployment. Yakovenko reviews the evolution of large language models, touches on the potential of AI-driven news apps, and examines the competitive landscape in AI innovation, all while sharing insights on the public's shifting perceptions of technology.
Ask episode
AI Snips
Chapters
Transcript
Episode notes
INSIGHT

LLM Training and Scaling

  • Large language models (LLMs) are trained on vast amounts of text data, predicting the next word or sentence.
  • This simple task scales surprisingly well, allowing LLMs to perform complex language-based functions.
ANECDOTE

NVIDIA's GPU Bet

  • NVIDIA's successful bet on GPUs for AI contrasts with Intel's struggles, highlighting the difficulty of predicting technological advancements.
  • Even within NVIDIA, deep learning wasn't initially prioritized, showing how unexpected breakthroughs can emerge.
ANECDOTE

Overestimating AI Application Development

  • Early excitement about AI capabilities, like generating synthetic Wikipedia articles, often leads to overestimating the speed of application development.
  • Real-world applications often take years to materialize, as seen with LLMs before ChatGPT.
Get the Snipd Podcast app to discover more snips from this episode
Get the app