Rethinking Learning Methods
Christian challenges the reliance on SGD in learning, highlighting the need to incorporate faster memory processes akin to human intelligence. He discusses the limitations of current models like transformers in capturing crucial aspects of intelligence, questioning the reasoning capabilities of GPT-3 towards achieving AGI.In this clip
From this podcast
Machine Learning Street Talk (MLST)
#50 Christian Szegedy - Formal Reasoning, Program Synthesis
Related Questions
Will large language models scale all the way to artificial general intelligence (AGI) as discussed in the episode Episode #183 ... Is ChatGPT really intelligent? and the clip The Limitations of Language Models from the episode MEGATHREAT: The Dangers Of AI Are WEIRDER Than You Think! | Yoshua Bengio and the clip Neural Networks Explained?
What do you think about the potential for Large Language Models (LLMs) to scale to Artificial General Intelligence (AGI) as discussed in the episode Ryan Greenblatt - Solving ARC with GPT4o, the clip Arc Challenge Reflections, and the episode Aravind Srinivas: Perplexity CEO on Future of AI, Search & the Internet | Lex Fridman Podcast #434?
Do you think that GPT-3 is on the way to AGI?