The TWIML AI Podcast (formerly This Week in Machine Learning & Artificial Intelligence) What’s Next in LLM Reasoning? with Roland Memisevic - #646
33 snips
Sep 11, 2023 In this discussion, Roland Memisevic, Senior Director at Qualcomm AI Research, explores the future of language in AI systems. He highlights the shift from noun-centric to verb-centric datasets, enhancing AI's cognitive learning. Memisevic delves into the creation of Fitness Ally, an interactive fitness AI that integrates sensory feedback for a more human-like interaction. The conversation also covers advancements in visual grounding and reasoning in language models, noting their potential for more robust AI agents. A fascinating glimpse into the evolving landscape of AI!
AI Snips
Chapters
Transcript
Episode notes
Language's Role in AI
- Language is crucial for human-like AI, enabling shared concepts and common sense.
- Training networks on labeled data, including verbs, adjectives, and adverbs, creates richer representations.
Fitness Ally: A Research Platform
- Fitness Ally, a virtual fitness coach, was developed to explore real-time, language-driven interaction.
- It combines visual input, language processing, and virtual embodiment for personalized fitness coaching.
LLMs vs. RNNs: Recurrence Trade-off
- LLMs, while enabling parallel training, lack true recurrence, hindering length generalization.
- RNNs could theoretically achieve similar performance but require significantly more training time.