Connect with us

Hi, what are you looking for?

ContributionAmericans.comContributionAmericans.com

Tech News

OpenAI’s flagship AI model has gotten more trustworthy but easier to trick

Illustration of Microsoft and OpenAI logos
Image: Microsoft

OpenAI’s GPT-4 large language model may be more trustworthy than GPT-3.5 but also more vulnerable to jailbreaking and bias, according to research backed by Microsoft.

The paper — by researchers from the University of Illinois Urbana-Champaign, Stanford University, University of California, Berkeley, Center for AI Safety, and Microsoft Research — gave GPT-4 a higher trustworthiness score than its predecessor. That means they found it was generally better at protecting private information, avoiding toxic results like biased information, and resisting adversarial attacks. However, it could also be told to ignore security measures and leak personal information and conversation histories. Researchers found that users can bypass safeguards…

Continue reading…

You May Also Like

Editor's Pick

Vanessa Brown Calder Commercial‐​to‐​residential zoning reforms have become a popular focus of regulatory reform legislation in states, with at least nine introducing legislation in 2023...

Editor's Pick

On this week’s edition of StockCharts TV‘s StockCharts in Focus, Grayson discusses the most important chart on all of StockCharts – your default ChartStyle! In addition...

Editor's Pick

We are so excited that StockCharts.com is now offering you our ACP plugins, we had to write about it for today’s Daily, especially since...

Tech News

Illustration: Beatrice Sala 23andMe posted a blog yesterday saying that data from users of its genetic testing and analysis platform has been circulating on...