Liquid AI's Ramin Hasani on liquid neural networks, AI advancement, the race to AGI & more! | E1928
Liquid AI’s Ramin Hasani joins Jason to discuss the mission and the concept of Liquid AI's liquid neural networks. They dive into liquid neural networks’ applications, transition from theory to execution, their efficiency on small devices, and more!
Key Points
- Liquid AI is a company focused on creating efficient and scalable artificial intelligence systems that are rooted in biology and physics, drawing inspiration from the nervous system of the C. elegans worm to invent liquid neural networks which remain adaptable after training.
- While current AI systems like GPT-4 are large, parameter-heavy, and fixed after training, Liquid AI's liquid neural networks are smaller, require significantly less computational power, and can continue to adapt and learn from incoming inputs.
- Liquid AI's technology aims to understand and deploy AI systems efficiently in society, bridging the gap between statistical models and causal models to create more explainable and controllable AI, with applications across various industries like finance, healthcare, and autonomous driving.
Chapters
0:00 | |
1:09 | |
7:06 | |
8:34 | |
16:07 | |
20:07 | |
21:37 | |
27:30 | |
34:15 | |
35:24 | |
41:18 | |
43:57 | |
50:25 | |
57:01 | |
1:00:12 |
Transcript
Loading transcript...
- / -