Paul Guzzo, University Communication and Marketing
In the 1980s, a visionary professor of computer science and engineering at the University of South Florida worked with ambitious doctoral students to explore technology that was still in its early stages.
They taught artificial intelligence systems to identify chairs.
Today, such work may seem rudimentary. But at the time, it was groundbreaking for global artificial intelligence research and USF.
USF continues to lead in AI. In the fall, Florida’s first Bellini College of Artificial Intelligence, Cybersecurity and Computing will be launched. It is also the first university of name dedicated to Fields, thanks to historic $40 million gifts from Arnie and Lauren Bellini.
“This is a huge step for the university,” said Louise Stark, a USF alumnus who worked on the chair project. “Congratulations to USF.”
This step is part of a much larger journey that began 40 years ago, with a group of pioneering professors and students who specialize in artificial intelligence when the field was seeking public acceptance and support.
“AI wasn’t a national conversation at the time,” he said, and has been studying artificial intelligence since joining USF in 1986, serving as co-director + X at the USF Institute.
Those pioneers had to start a conversation.
Hall has arrived with Kevin Bowieer, the professor who worked on its chair project, dated the 1984 USF artificial intelligence research.
Perhaps that might be accurate, but he wasn’t sure anyone hadn’t come before him.
“The University of Notre Dame has a great opportunity to learn about the world,” said Bowyer, a professor at the Computer Science Engineering Department. “We decided to build a USF AI research block for each block.
But the same pioneer had very few skills to support these ambitions.
“There was no Amazon,” Bowyer said with a laugh. “Computer science artifacts were much more expensive back then, so if you wanted something like a 3D scanner, it wasn’t easy to get. You had to write a special grant.
There’s a good reason for that, says Sudeep Sarkar, interim dean of Bellini College, who studied artificial intelligence at Ohio State University before joining USF in 1993, initially as an assistant professor of computer science and engineering. Artificial intelligence was not moving forward at the same speed as today. Researchers struggled to teach artificial intelligence, such as how to find roads on maps and identify boundaries for individual objects in photos.
“We’ll talk to friends about our interest in artificial intelligence. They’ll laugh at how it’s a useless research that never leads to a lot of things, but we were excited by the possibility.”
Provisional Dean’s sudeep sarkar
“Excitement comes from pushing boundaries. If there’s no one who thinks you’re doing crazy things, you probably aren’t pushing those boundaries enough,” Boweer said.
But it wasn’t just that USF researchers were unable to get technology all the time. Some of the things they needed didn’t exist.
“AI has been around since the 1960s and began gaining traction in the 1980s, but we faced some important challenges,” said Prasant Mohapatra, Executive Vice President of USF Provost. “Some of my friends who went to the area at the time couldn’t get a job and had to change the area. Can you believe that? But that was because technology wasn’t ready to support their work.”
For example: In the early 1980s, there was no World Wide Web.
“A lot of what we do today requires a big data set,” Hall said. “Today, we can scrape the entire internet for data. We can find a million examples of data and fit a store-bought disk. So, without these huge data sets, we were pretty much looking for technology to improve performance.
That’s how Stark and Bowyer taught artificial intelligence systems to identify chairs in 1988 and 1989.
We created approximately 250 3D models of objects that could become chairs using computer aided design software.
Other early artificial intelligence researchers have created datasets of only objects that the system normally wanted to learn to identify through memorization. Stark and Bowyer wanted to teach artificial intelligence to use logic to identify chairs between groups of similar objects.
“That was more common sense than anything else,” said Stark, who is now retired. “We put rules into the system and tested seatable surfaces with stable support.”
The first rule was to align the objects on the flat surface into the dataset and then put them in a group. From the following groups, the system identified objects with stable bases: However, since that collection might have included a trash can, the system chose objects with back support and put them in another new group, doing so until only the chairs remained.
This research paper was published in 1992 in “IEEE Transactions on Pattern Analysis and Machine Intelligence.”
“Then it came to recognition,” Bowier said. “The national and internationally renowned AI figures have noticed that research. It led to the recognition of other AI research at USF. It was a breakout moment.”
Over the next few years, Stark said he was wondering as others improved her job. “One of the criticisms of my research was that the system could not determine whether the objects were made of paper mache or sturdy, so other students added that.
Bowier is somewhat plagued when he hears people talk about how smart artificial intelligence has become. The system itself is not wise, he said, which is why it is called artificial intelligence. Rather, they just know what it is programming.
“What we enjoy today is the result of dozens of people, all built on each other’s work for decades,” he said. “Some of them had enormous insights. Some have modest insights, but none of them have done it all by themselves.”
“The USF AI community is not only taken seriously, but is at the forefront of a growing industry.” – Interim Dean Sudeep Sarkar
The same can be said about Bellini University.
“It took a lot of resilience for the early AI researchers at USF to stick to it despite what others were saying.
