AI is biased. The White House is working with hackers to try to fix that

Tech companies are encouraging independent hackers to test their AI models for biases and inaccuracies in order to make the technology more equitable and inclusive, as demonstrated by the largest-ever public red-teaming challenge at Def Con.

npr.org

Relevant topic timeline:

8/21/2023

DEF CON's AI Village Pits Hackers Against LLMs to Find Flaws

Thousands of hackers participated in the AI Village event at DEF CON 2023, attempting to make large language models say something dangerous, with the top three winners, including Stanford student Truc Cody Ho, showcasing their ability to generate discriminatory code, credit card numbers, and misinformation.

8/27/2023

Can AI Increase Diversity On Tech Teams? The Jury Is Out

A survey found that most Americans believe there is racial bias in corporate hiring practices, and many believe that artificial intelligence (AI) could help improve equality in hiring, although skepticism remains, particularly among Black Americans; however, concerns about the ethical use of AI remain due to biases in AI systems that favor white, male, heterosexual, able-bodied candidates. Hackajob, a UK-based hiring platform, has introduced features to increase diversity and reduce bias in tech teams, while experts emphasize the importance of addressing bias in AI datasets through diverse data collection and involving underrepresented groups in AI system development.

9/1/2023

From Google To Nvidia, Tech Giants Have Hired Red Team Hackers To Break Their AI Models

AI red teams at tech companies like Microsoft, Google, Nvidia, and Meta are tasked with uncovering vulnerabilities in AI systems to ensure their safety and fix any risks, with the field still in its early stages and security professionals who know how to exploit AI systems being in short supply, these red teamers share their findings with each other and work to balance safety and usability in AI models.

Topics

Posted 8/26/2023, 9:00:49 AM