Urgent.News

What's breaking now, across thousands of outlets.

AI

OpenAI's Math Papers Destroyed Careers Of Early Mathematicians: US Professor

New York University Professor Tristan Buckmaster raised concerns about potential conflicts of interest over OpenAI's findings on 377 mathematical problems and access to researchers' unpublished ideas.

OpenAI's Math Papers Destroyed Careers Of Early Mathematicians: US Professor

New York University Professor Tristan Buckmaster has expressed concerns that OpenAI's release of solutions to over 700 mathematical problems has negatively impacted early-career mathematicians. According to Buckmaster, the company's release has "wiped out" entire research programs that these mathematicians were working on.

The release, which was made available on GitHub, includes computer-checkable versions of many proofs and is the result of an internal frontier model. Some mathematicians, including Buckmaster, have raised questions about potential conflicts of interest and whether unpublished data from researchers was used to train the AI.

Buckmaster told CNBC that the release had upended work by early-career mathematicians, stating that "whole research programs" were wiped out. Other mathematicians, such as Bryna Kra of Northwestern University, have noted that while AI can advance research, releasing so much work at once could undermine academic norms.

Brief written by urgent.news from NDTV - India News, Business Insider — 2 reports on this story. Machine-written — may contain errors; check the original before relying on it.

This story

This is one outlet's version. Read the fullest account.

Read the original at ndtv.com →

More in AI

Train an 8B LLM on a 4GB Laptop GPU: Hands-On with Soup

Train an 8B LLM on a 4GB Laptop GPU: Hands-On with Soup TL;DR Soup eliminates the high VRAM barrier of LLM fine-tuning by introducing layer streaming, enabling you to fine-tune an 8B parameter model…

  • Soup model trains 8B LLM on 4GB laptop GPU using layer streaming
  • Declarative YAML config specifies training parameters like model, batch size, learning rate
  • Supports 4-bit and 8-bit quantization for model compression and speed

More from Saturday 10 October →