Faculty Shine at SXSW

March 26, 2025

South by Southwest (SXSW) annually unites innovators of industry and creative thinkers in Austin, Texas, for a global assembly of technology, film, television, music and more.

Carnegie Mellon University’s rich expertise in technology and artificial intelligence, complemented by a plethora of talent in the arts and entertainment, makes SXSW an excellent venue for community members to showcase their skills. 

During SXSW 2025, Carnegie Mellon experts brought their groundbreaking work to new audiences in Austin.

The School of Computer Science faculty delivered five key considerations for individuals utilizing large-language models (LLMs) in their personal and professional lives.

“We really just want people to think twice before they throw anything to the model and make sure they also think twice before they take anything from the model,” Wu said.   

At another session, Sarah Fox and Nikolas Martelaro presented their collaboration with transit operators and their unions aimed at understanding the impacts of future technologies on transportation workers. The two explained the potential to leverage AI in "Creating Safer, More Equitable Public Transit Systems."

Their ongoing research involves partnership with the Transport Union Workers of America, the Amalgamated Transit Union and the American Federation of Labor and Congress of Industrial Organizations (AFL-CIO) Technology Institute. 

“We have been really excited to work with our union partners and to learn from real operators what is all the complexity and things that are happening on the road in real transit operations,” Martelaro said. “How can we understand and learn from operators to bring that knowledge into thinking about new technologies?” 

Fox explained that the role of bus operators often goes beyond driving. They are often responsible for balancing the environment of the vehicle and the wellness of passengers. 

“Bus operators regularly described unexpected circumstances, both in navigating the road and managing the social environment of the bus. These circumstances often require split-second decision making and nuanced social judgments,” she said. 

 

How to Be a Smarter AI User

• LLMs struggle to infer what you mean.
  • Solution: Be more structured with your instructions to LLMs.
• LLMs hallucinate and are overconfident.
  • Solution: Double verify and think critically about model outputs.
• LLMs are privacy liabilities.
  • Solution: Be careful what information you provide to LLMs.
• LLMs appear more human than they are
  • Solution: Beware of attributing human likeness to AI and avoid anthropomorphic language.
• LLMs don’t work for everyone.
  • Solution: Be thoughtful and careful about when/how you use LLMs.
     

This is a shortened version of the article about the CMU community at SXSW 2025. To view the full article, visit CMU News.