Comments on: Can you trust ChatGPT and other LLMs in math? https://bdtechtalks.com/2023/03/06/chatgpt-llm-mathematics/?utm_source=rss&utm_medium=rss&utm_campaign=chatgpt-llm-mathematics Technology solving problems... and creating new ones Tue, 07 Mar 2023 14:53:34 +0000 hourly 1 By: Logan Higgins - Conversion Blitz https://bdtechtalks.com/2023/03/06/chatgpt-llm-mathematics/comment-page-1/#comment-35510 Tue, 07 Mar 2023 14:53:34 +0000 https://bdtechtalks.com/?p=15900#comment-35510 [Mathematics under the scrutiny of ChatGPT and other LLMs] This article sheds light on the capabilities and limitations of language models such as ChatGPT in solving mathematical problems. While these models have proven useful in several domains, including math, their performance in this field is often below that of an average graduate student. However, studies like these help us understand the problem-solving gap between humans and deep neural networks, highlighting the importance of informing non-mathematicians using ChatGPT of its limitations.

]]>
By: Casey Dorman https://bdtechtalks.com/2023/03/06/chatgpt-llm-mathematics/comment-page-1/#comment-35507 Mon, 06 Mar 2023 18:09:37 +0000 https://bdtechtalks.com/?p=15900#comment-35507 Can a LLM to generate a procedure for solving a mathematical problem when that procedure is logical derivable from the mathematics it has been trained on but was not contained in the training corpus and could not be generated using statistical probabilities between elements in the corpus (i.e. the way it generates language)? Do deep-learning devices trained primarily on mathematics do this?

]]>