Sepp Hochreiter - LSTM: The Comeback Story?

Topics covered
Popular Clips
Questions from this episode
- Asked by 109 people
- Asked by 84 people
- Asked by 69 people
- Asked by 33 people
- Asked by 20 people
- Asked by 18 people
- Asked by 12 people
Episode Highlights
Industrial AI
Industrial AI is poised to revolutionize sectors like simulation and robotics by enhancing speed and efficiency. highlights the potential of XLSTM technology, which offers energy efficiency and rapid processing capabilities, making it ideal for industrial applications beyond language processing 1. This technology allows for simulations that were previously impossible due to computational limitations, such as simulating millions of particles in industrial processes 2.
We can simulate it and we can build the real thing. And this gives industry a big, big push.
---
Additionally, XLSTM's efficient memory systems enable its use in embedded devices, making it suitable for applications in robotics and autonomous drones 3.
Real-World Challenges
XLSTM addresses real-world challenges in industries like automotive and aerospace by offering fast inference speeds and fixed memory advantages. explains that XLSTM's ability to perform fast inference makes it 100 times faster than traditional methods, allowing for more efficient processing in robotics and other industrial applications 4. This speed advantage is crucial for applications requiring real-time decision-making, such as autonomous vehicles.
We are fast in inference. That would be better. But this fast inference speed also helps us to grow in industrial applications away from language.
---
Furthermore, advancements in XLSDM, a variant of XLSTM, have made it faster than Flash Attention in both training and inference, showcasing its potential for widespread industry adoption 5.
Related Episodes


Jürgen Schmidhuber - Neural and Non-Neural AI, Reasoning, Transformers, and LSTMs
Answers 383 questions

Open-Ended AI: The Key to Superhuman Intelligence? - Prof. Tim Rocktäschel
Answers 383 questions

#046 The Great ML Stagnation (Mark Saroufim and Dr. Mathew Salvaris)
Answers 383 questions

Jurgen Schmidhuber on Humans co-existing with AIs
Answers 383 questions

ICLR 2020: Yoshua Bengio and the Nature of Consciousness
Answers 383 questions

Ben Goertzel on "Superintelligence"
Answers 383 questions

#032- Simon Kornblith / GoogleAI - SimCLR and Paper Haul!
Answers 383 questions

#54 Gary Marcus and Luis Lamb - Neurosymbolic models
Answers 383 questions

MLST #78 - Prof. NOAM CHOMSKY (Special Edition)
Answers 383 questions

#50 Christian Szegedy - Formal Reasoning, Program Synthesis
Answers 383 questions

OpenAI GPT-3: Language Models are Few-Shot Learners
Answers 383 questions

Cohere co-founder Nick Frosst on building LLM apps for business
Answers 383 questions

047 Interpretable Machine Learning - Christoph Molnar
Answers 383 questions

Prof. Subbarao Kambhampati - LLMs don't reason, they memorize (ICML2024 2/13)
Answers 383 questions
