This Week in Startups

Liquid AI's Ramin Hasani on liquid neural networks, AI advancement, the race to AGI & more! | E1928

22 snips
Apr 9, 2024
Guest
Ramin Hasani
Ramin Hasani, CEO and co-founder of Liquid AI, dives into the world of liquid neural networks, drawing inspiration from the adaptability of worms. He discusses practical applications in fields like autonomous driving, showcasing how these networks can outperform traditional models with far fewer parameters. Hasani highlights the transformative potential for AI efficiency on compact devices and emphasizes the importance of responsible development in the race toward artificial general intelligence. Tune in for insights that blend biology with cutting-edge technology!
Ask episode
AI Snips
Chapters
Transcript
Episode notes
INSIGHT

Liquid Neural Networks vs. Traditional AI

  • Liquid neural networks, inspired by worm brains, are adaptable to new inputs after training.
  • Traditional AI models become fixed after training, limiting their flexibility.
ANECDOTE

Autonomous Driving Demo

  • In a driving demo, a small liquid neural network with 19 neurons outperformed a traditional network with 500,000 parameters.
  • The liquid network focused on the road better, demonstrating robustness.
INSIGHT

Applications of Liquid Neural Networks

  • Liquid neural networks excel at modeling time series data like video, audio, and text.
  • This makes them applicable to diverse fields, from finance to medicine.
Get the Snipd Podcast app to discover more snips from this episode
Get the app