The BYU Artificial Intelligence Association hosted a Campus AI Day on Tuesday, November 5, to help students network with AI professionals and learn more about advances in AI.
During the day, various companies based on AI applications were invited to network with students and professors.
Some of these new tools included DonkeyChats, an AI tool that was originally developed to facilitate political debates and has now become a way to help couples communicate better with each other. Passive Logic, another AI tool that uses predictive models to heat and cool a building, and Tarriflo that helps automate commerce across global borders, were also present at the event.
A panel discussion with industry professionals Andrew Carr, Jeremy Fillingim and David Wingate gave students the opportunity to ask questions about the progress and future of artificial intelligence.
The panel was moderated by Jeffrey Olmo, vice president of the Artificial Intelligence Association, who opened the conversation by asking the panelists how surprising it is to them that language models work as well as they do.
Fillingim, founder of Passive Logic, recalled the shock many felt when AI finally took off.
“I’m really pleasantly surprised by what it’s good at, and still surprised by what it’s not so good at,” Fillingim said.
The next question for the panelists was whether society is moving into an era of diminishing returns for training AI based on more data, or whether scaling will provide more and more capabilities to AI models.
Carr, co-founder of Cartwheel, responded by explaining that the problem with scaling isn’t whether there will be diminishing returns, but whether there will be enough data, power and computing to get to that point. The other point Carr made was that this scalability may not be important for the practical application of AI.
Wingate, a BYU professor who studies language models, answered the next question about whether AI will reach a bottleneck by explaining that every time an AI is developed, it starts with poor performance that is slowly improved until it is in very short order. time improves exponentially. . He said he expects this trend to continue in other applications for AI.
Olmo asked the panelists for their opinions on the future of education in combination with AI, given its possibilities.
“Block and tackle is still important… having an engineering mindset, the ability to tackle a problem, that is a skill that is transferable no matter what set of tools you have at your disposal,” Fillingim said.
Carr encouraged students to “just do things,” meaning the more things a student does outside the classroom, the more skills they will develop to influence the world around them.
Sharing his perspective as a teacher at BYU, Wingate explained that it is up to students to be active and mature in their pursuit of educational experiences because AI makes it easier to cheat.
In closing, Olso asked the panelists for their most contrarian views on AI.
“Are you thinking enough about AI for the poor? Are we thinking about the homeless population? What about our prison populations? What about single mothers and orphans? Will AI bless them or will it just make rich people richer?” Wingate asked.
Wingate emphasized that it is the responsibility of people who understand AI to choose how they will use it for the benefit of the world.