Episode 100: Multi-Agent and Energy Efficient AI Models are the next big thing

Our guest in this episode of AI-podden is Noam Slonim, Distinguished engineer and Initiator and leader of the Debater project, IBM

Noam Slonim shares with us his best boughts on the recent developments of large language models (LLMs)and we discuss how this new technology relates to his previous breakthrough on building an AI that can debate humans successfully and without false claims.

Noam and Ather discuss the level of intelligence in the new LLMs and whether they have reasoning abilities yet. Noam believes that it certainly has some capabilities in reasoning but that there is still room for improvements and that he also believes in smaller but more specialized language models that have different knowledge, rather that one central LLM.

Ather and Noam also discuss the important topic of inventing more energy efficient algorithms to train AI systems, both from a cost as well as an environmental perspective.