By Sarisha Boodoo, TESI Environmental Leaders Network Intern
The University of Florida recently announced a plan to integrate artificial intelligence (AI) into every major, with the hopes of establishing it as a core competency across the curriculum. With this new AI-enabled workforce, UF hopes to fuel computer-based solutions for some of the most pressing issues of our time: rising seas, aging populations, data security, personalized medicine, urban transportation and food insecurity.
Now, the UF Thompson Earth Systems Institute's Scientist in Every Florida School Program (SEFS) is working to extend AI to Florida's K-12 educators and students. Their secret ingredient to getting them onboard with what can seem to be a daunting topic? Shark teeth!
From July 11-15, 2022, SEFS hosted a cohort of 12 Florida middle school teachers at UF for a workshop focused on using AI to identify fossil shark teeth. The workshop is just one component of a yearlong professional development program titled “AI Learning in K-12 with Fossil Sharks.”
Funded by the National Science Foundation, this three-year pilot program aims to connect K-12 students and teachers with UF’s cutting-edge technology, and hopefully spark further scientific and career curiosities in the classroom.
For more background on the program, read our press release.
Meet the Teachers
Using a branch of artificial intelligence called machine learning, the group worked closely with paleontologists, education professionals, and computer engineers to train computers to identify the teeth of the extinct Megalodon! At the same time, the teachers worked together to develop standards-based, hands-on lesson plans to bring back to their classrooms.
Read more below for a breakdown of each day.
Monday, July 11, 2022
Breaking Down Artificial Intelligence and Debunking Myths
The cohort jumped into the fundamentals of AI on Monday morning. Although the technology is widely used among us, it is not so widely understood.
Pasha Antonenko, an associate professor of educational technology in the UF College of Education, described the misconceptions of artificial intelligence being painted as “a big, blue cyborg with wires everywhere.” However, AI is not as obscure as it seems.
“Artificial Intelligence is a computer system able to solve problems and perform tasks that ordinarily require human intelligence," Antonenko said. "It uses sensors and prior data to perceive the world and analyze data. It then autonomously uses this perceived data and improves its intelligence based on the data it analyzes."
Jeremy Waisome, an instructional assistant professor in the UF Herbert Wertheim College of Engineering, summarized AI with five main points from AI4K12’s Five Big Ideas in AI.
After soaking up some information, teachers began developing their fossil kits, which included around sixteen different shark teeth, some of which were real, and others that were 3-D printed. The team explained that later in the week, the teachers would develop a computer model using Roboflow software, to help distinguish the teeth based on their unique characteristics.
Later that day, teachers learned more about machine learning and the importance of variability within a dataset with David Reed, director of UF's new academic center for Artificial Intelligence.
Reed emphasized that the accuracy of machine learning is dependent on the quantity and variation of input data.
“For example, developing the skills of identifying an animal as a dog requires the computer to view thousands of different dog photos," Reed explained. In this way, the computer would be able to detect that an image is a dog, even if it does not resemble the standardized image of a dog.
The group also learned that machine learning certainly has its limitations, which may lead to inaccurate output predictions.
Brian Stucky, an assistant scientist at the Florida Museum of Natural History, presented his research on using artificial intelligence to understand the mating and courtship behaviors of eastern redback salamanders. He demonstrated how crowdsourcing applications such as iNaturalist – an app for nature enthusiasts – use spatial, auditory, and visual data to identify animals.
Stucky's presentation showed that many environmental curiosities can be answered with the data that AI provides, but this data can sometimes be biased or inaccurate. For example, due to its bright red color, the Eastern redback salamander is perhaps more noticeable than other salamanders. Therefore, humans may be more prone to document eastern redbacks on apps like iNaturalist, causing biased output data showing more eastern redbacks than other salamanders in the area.
Stucky said the accuracy of applications like iNaturalist depends on frequent reinforcement. The questions and concerns the teachers posed about AI were good practices for the conversations teachers may have in their classrooms when implementing their new lesson plans.
Tuesday, July 12, 2022
Training Datasets for Machine Learning
On day two, the cohort took a closer look at machine learning and trained computer models themselves!
Victor Perez, an assistant curator of paleontology at the Calvert Marine Museum in Maryland, kicked his presentation off with a simple question that ignited conversation across the room: What is data?
Data is the info used to develop meaning or make decisions. It can be quantitative and qualitative. Training a dataset for machine learning involves organizing the input data into categories known as classes. The input stages include collecting data, preparing data, choosing a model, and training the model.
The cohort pulled out their shark teeth kits and started developing their input data. They organized the teeth by color, shape, texture, number of serrations, and size. While there is some contention about the best way to measure the size, Perez reminded the teachers that output data is subject to errors due to human biases, natural variation, image quality, sample size, and data sources.
Perez recommended that teachers improve their input data by increasing the sample and variability size, but also find another way to classify data used by experts – in this case, by taxonomy, ecology, and anatomy. In the last part of the lesson, the teachers moved on to enter all their qualitative and quantitative data into a database.
With the amount of time the teachers spent examining shark teeth, the group eventually stepped into some Megalodon jaws at the Florida Museum of Natural History’s exhibit Florida Fossils: The Evolution of Life and Land! Bruce MacFadden, TESI director and UF distinguished professor, lent the group his paleontology expertise on the exhibit tour. Teachers also had the chance to visit the Museum's temporary exhibit, Science Up Close: Fantastic Fossils, where paleontologists prepared fossils in front of visitors.
MacFadden mentioned ways AI can assist museum researchers in quickly identifying common fossils, the same idea as the model the teachers were developing to identify shark teeth. The teachers became inspired to make their lessons more creative by engaging their students with the premise that they are paleontologists themselves, using AI to help them identify the fossils they “discovered.”
While they did not identify fossils inside the museum exhibits, the cohort trekked out to the UF Natural Area Teaching Lab to identify plants and bugs using the apps iNaturalist and Seek with Sadie Mills, TESI educator and coordinator, and Alan Ivory, SEFS scientist in residence. The team learned how to use the apps, and even took part in a challenge to find specific butterflies.
After, the group reconvened to discuss the challenges and benefits of machine learning in both applications. Some teachers expressed their image quality was not always up to par and the identification was either unable to be made or not accurate. However, as Stucky had mentioned, one of the main features of these apps is that humans usually proof the submitted image. From this, the application can then improve its identification through reinforcement learning.
Overall, the team enjoyed the exercise, especially with the daily challenge provided by the Seek app, which encourages users to photograph specific plants, insects, and animals and earn badges – this demonstrates a way in which artificial intelligence can be harnessed to exercise greater environmental stewardship among users.
For the rest of the day, the teachers worked on data input using GIFs! To see how object detection works in machine learning, the teachers were tasked with finding a GIF of a megalodon, saving the image, and then uploading it to Roboflow and Teachable Machine. In this exercise, the teachers confronted possible issues students may have pertaining to digital literacy. The team had to input megalodon and non-megalodons photos. They also learned about fixing images within a scale and annotating.
Wednesday, July 13, 2022
Machine Learning Capabilities
The cohort switched gears and headed into the UF Herbert Wertheim College of Engineering on Wednesday morning. Ph.D. candidates Connor McCurley and Ritesh Chowdhry presented on the inner workings of machine learning models.
McCurley's presentation focused on dataset features and ways in which artificial intelligence models learn from distinguishing features such as color, age, and width. He drew upon real-life environmental monitoring that requires machine learning such as root segmentation, coral reef health, early wildfire detection, and plant and animal identification.
Chowdhry’s presentation focused on the information processing systems of machine learning. “Reinforcement learning is equivalent to the way humans and animals learn from mistakes. Over time, the computer will learn. Artificial neurons work as a brain,” Chowdhry explained.
After the presentations, the group toured the College of Engineering's 3-D printing lab before heading to UF's Lake Wauburg for some social time and BBQ.
Thursday, July 14, 2022
On Thursday, teachers learned more about specific applications of AI and how scientists use it during a Q&A session with Emma MacKie, an assistant professor of machine learning, glaciology and geophysics.
Later, teachers learned more about Camp Dialogues, a summer camp to teach 7th and 8th graders how to use AI to build conversational apps like Siri and Alexa from Dr. Maya Israel, an associate professor of educational technology and computer science education in the UF College of Education. The presentation helped them understand how their students might react to AI concepts.
Friday, July 15, 2022
The week culminated in a showcase of the lesson plans teachers and scientists worked on throughout the week and how they planned to implement them in their classrooms. Participants were able to give feedback on other projects as well as learn new ideas. At the end of the showcase, the group came back together to share reflections of what they learned.
"This workshop was just the first element of this professional development program," said Brian Abramowitz, SEFS K-12 education and outreach coordinator. "I am so excited to continue this collaboration among the first cohort of teachers and look forward to seeing how students react to their new AI knowledge."
Learn more in this video!
Workshop Instructors:
- Bruce MacFadden, TESI Director and Distinguished Professor, Florida Museum of Natural History
- Pasha Antonenko, Associate Professor, School of Teaching and Learning, UF College of Education
- Jeremy Waisome, Instructional Assistant Professor, UF Herbert Wertheim College of Engineering
- Victor Perez, Assistant Curator of Paleontology, Calvert Marine Museum
- Stephanie Killingsworth, K-12 Education and Outreach Coordinator, Scientist in Every Florida School
- Brian Abramowitz, K-12 Education and Outreach Coordinator, Scientist in Every Florida School
- Sadie Mills, Educator and Coordinator, UF Thompson Earth Systems Institute
- Alan Ivory, Scientist in Every Florida School K-12 Smallwood Foundation Fellow