Matthias Bastian

3 days ago | the-decoder.com | Matthias Bastian

ChatGPT can now access company data through a SharePoint connector. The beta feature is available for ChatGPT Plus, Pro, and Team users. With the new integration, ChatGPT can analyze and summarize content from multiple SharePoint sites, complete with source references. OpenAI says use cases include cross-department summaries of strategy documents or building customer profiles by combining internal data with information from the web.

3 days ago | the-decoder.com | Matthias Bastian

Summary Jakub Pachocki, who leads advanced model development at OpenAI, says AI's ability to generate knowledge autonomously marks a turning point for business and research. According to Pachocki, so-called reasoning models are on track to generate knowledge autonomously. He describes this as a form of "reasoning," though it's fundamentally different from how humans think.

4 days ago | the-decoder.com | Matthias Bastian

Summary Many language models are more likely to generate incorrect information when users request concise answers, according to a new benchmark study. Researchers at Giskard evaluated leading language models using the multilingual Phare benchmark, focusing on how often they "hallucinate," a term for when models produce false or misleading content, under realistic usage conditions.

4 days ago | the-decoder.com | Matthias Bastian

U.S. President Donald Trump has dismissed Shira Perlmutter, head of the U.S. Copyright Office, shortly after her office released a report opposing broad "fair use" exemptions for AI training purposes. The report's stance conflicts with the interests of Trump ally Elon Musk and much of the AI industry. Perlmutter had served since 2020. Rep.

4 days ago | the-decoder.com | Matthias Bastian

Summary The US Copyright Office has pushed back against one of the AI industry's most common legal arguments: that training AI models on copyrighted material generally qualifies as fair use. In a new report, the agency rejects several of the industry's key justifications—like comparing AI training to human learning or claiming that it's a "non-expressive" use.