NeurIPS 2024: Impossible GenAI Questions With Xiaowen Zhang скачать в хорошем качестве

NeurIPS 2024: Impossible GenAI Questions With Xiaowen Zhang

What happens when AI researchers ask each other the toughest, most mind-bending questions they can think of? In this NeurIPS 2024 series, researchers attempt to answer the unanswerable. In this video, Xiaowen Zhang, Research Engineer at Manifest AI, takes his turn. Questions: 00:25 If we had unlimited compute power and used all existing data, could current AI training methods create machines with human-level intelligence? Or would we hit a performance ceiling? Explain your answer. 01:20 Why are foundation models dramatically less energy-efficient learners than human brains? 02:43 If you had unlimited compute, what problem would you tackle and why? 03:21 Will xLSTM be the future of NLP? 03:34 In your opinion, what should GPU-rich companies do to make research more accessible to GPU-poor labs (assuming that it is their goal)? Follow us & stay updated: ► Vist our website: https://buff.ly/NxcmKGi ► Follow us on Linkedin:   / neptuneai   ► Follow us on Twitter:   / neptune_ai   ► Check our Github: https://github.com/neptune-ai Connect with Xiaowen on Linkedin:   / seanxwz   #genai #generativeai #llm #llms #neurips