OpenAI's Google Search Competitor 🔍 // SoundHound x Perplexity 🤝 // Fishing for Glitch Tokens in LLMs 🎣

OpenAI plans to challenge Google search with a new search feature for ChatGPT, which could have a significant impact on the AI industry. SoundHound AI and Perplexity have partnered to improve the accuracy and complexity of voice assistants across cars, apps, and phone assistants. "Fishing for Magikarp" addresses an issue with "glitch tokens" in large language models, while "CuMo" proposes a new approach to improving multimodal LLMs. "Does Fine-Tuning LLMs on New Knowledge Encourage Hallucinations?" highlights the risk in introducing new factual knowledge through fine-tuning and the importance of pre-training for large language models. Contact:  [email protected] Timestamps: 00:34 Introduction 01:38 OpenAI plans to announce Google search competitor today 03:02 SoundHound AI and Perplexity Partner to Bring Online LLMs to Next Gen Voice Assistants Across Cars and IoT Devices 04:48 Homoiconic Python 06:06 Fake sponsor 07:49 Fishing for Magikarp: Automatically Detecting Under-trained Tokens in Large Language Models 09:00 CuMo: Scaling Multimodal LLM with Co-Upcycled Mixture-of-Experts 10:34 Does Fine-Tuning LLMs on New Knowledge Encourage Hallucinations? 12:13 Outro