USC’s Information Sciences Institute talks the future of AI

USC’s Information Sciences Institute (ISI) held a press conference Wednesday morning in Marina Del Rey to address AI advancements and concerns, covering topics ranging from password protection to online sex trafficking advertisements.

Public Communications Manager Magali Gruet introduced the organization, saying, “We do a lot of research in artificial intelligence, in cybersecurity. We do some quantum, we do computing, we even do some space work.”

The ISI research team, consisting of over 400 scientists and Ph.D students, then presented their AI research, new programs, and other industry-changing developments during the two hour conference.

ISI computer scientist Genevieve Bartlett kicked off the conference by discussing her work on automating anti-phishing software and how she plans on using AI to take down scams systems more efficiently.

“I think something we all know is that the weakest link is going to be humans,” she said regarding online security. “So a lot of our work is in detecting scams, understanding the risk of things like an email or a social media contact”.

Jelena Mirkovic, a computer scientist and project leader for ISI, then gave an in-depth presentation about the relationship between people and passwords and how the human affinity for personal and memorable passwords affects our likelihood of getting scammed.

Her studies showed that one way to encourage the use of secure passwords that remained memorable was to “leverage existing memories that users have and then elicit those memories through a series of questions,” creating a password that is long and complex but more attractive to users than a randomly generated sequence.

Another technological development presented during the conference is a program that uses AI to prevent sex trafficking, persecute predators and even recognize those who are more at risk for online sex crimes.

“It’s there in almost every city, it’s there in almost every country, you know, developed, developing, and we just don’t know enough about it,” said ISI research lead Mayank Kejriwal on what drove him to study how to combat sex trafficking.

Through evidence collected by a program he helped create called Domain Specific Inside Graph (DIG), which uses AI is used to identify and report online sex trafficking advertisements, multiple traffickers have been prosecuted. According to Kejriwal, the program is used by more than 200 law enforcement agencies.

However, with the rapid advancements like DIG in the AI industry comes a need for ethical evaluation of these systems. Research Assistant Professor Fred Morstatter’s studies focus on bias and fairness in AI, and the ways that we can prevent the reproduction of harmful human mindsets by artificially intelligent entities.

The tools Morstatter and his team build to audit these trained biases look for datasets that discriminate against marginalized groups such as people of color, women, and members of the LGBTQ+ community.

“People like to criticize the AI systems for having bias in them.” says Morstatter. “Really what they’re doing is they’re propagating the biases that people take.”