Philadelphia
Artificial Intelligence is all over the headlines, but it鈥檚 not all over K-12 curriculca鈥攜et. Many high schools are still struggling to offer computer science, much less instruction in cutting-edge systems that are designed to take in massive amounts of data and mimick the human brain by making predictions, finding patterns, and more.
One of a handful of exceptions: Whitney High School in Cerrittos, Calif., part of Los Angeles County, which for the first-time last school year offered a course titled 鈥淐omputer Programming for Solving Applied Problems Using Artificial Intelligence Honors.鈥� The teacher, Matt Johnson, presented at the International Society for Technology in Education鈥檚 conference here. (.)
Fifteen students signed up for Johnson鈥檚 course, including 10 boys and 5 girls. He had a high bar for entering the class. Students either had to have taken Advanced Placement Computer Science Principles, or demonstrated that they had deep programming knowledge.
Johnson, who worked in the tech industry before becoming an educator, initially considered teaching his kids how to program in AI applications with software that鈥檚 in vogue in Silicon Valley at the moment: Google鈥檚 TensorFlow. A friend assured him that if his kids could master that, they鈥檇 be offered high-paying jobs at companies like Facebook shortly after graduating from high school.
But, it would ultimately be doing them a disservice, the friend said. After a few years, the field may have moved on and the students could be unemployable. Instead, Johnson decided to 鈥渇ocus on the fundamentals. So that kids don鈥檛 end up in a box.鈥�
He divided his course into three areas. Those included:
Technical Skills: The class used 鈥楧eep Learning鈥� on Anaconda Platform and Jupyter, an open source web application. They also worked on concepts like the architecture of the brain which is relevant for the development of so-called neural networks, which are basically voluminous interconnected processing nodes used to help computers learn how to perform new tasks鈥�) and sorting and algorithmic design, and more.
Non-technical skills: Johnson reserved Fridays as a day to talk about AI鈥檚 strenghts and weaknesses and the way it may change the world. He and his kids discussed things like the history of women in coding and the rise of automation. It was important to Johnson that the students get exposed to the broader implications of the technology they were trying to master. 鈥淚 don鈥檛 want to send a bunch of whiz kids out there who don鈥檛 understand the repercussions for society鈥� of what they are doing, he said.
Projects: The class culminated with student projects on things like basketball, climate change, and more. One student tried to create a program that would be able to write original songs mirroring the style of the popular rapper .
How successful was the program?
鈥淚 don鈥檛 think Drake should be worried,鈥� Johnson said.
Here were three big takeaways from his first year:
Be Ready to Help High-flying Students Learn to Fail
Johnson is used to teaching physics and computer science, but even he was surprised by the differences in skill between the students who seemed to immediately grasp the high-level AI programming concepts he was teaching and those who flailed.
鈥淭he students who really struggled were used to working in boxes, with clearly defined goals,鈥� Johnson said. 鈥淭here was a big gap between the kids who were zooming ahead and those falling behind.鈥� That complicated his task, but was also tough for the kids who weren鈥檛 moving ahead as quickly, most of whom were used to being stand-out students. 鈥淲hen you鈥檙e struggling and you see that student [next to you] isn鈥檛 just done, but gleefully zooming along,鈥� it can be an ego blow, he said.
He鈥檚 tried to tell his kids: 鈥淐omparison is the thief of joy, do not compare yourself to other people. ... You鈥檙e slower, but you鈥檙e getting skills that most people don鈥檛 have at this age.鈥�
Textbooks May Be Necessary, For Now
Johnson tried to steer clear of using a textbook to teach AI, but that was difficult both for him and the kids who didn鈥檛 immediately grasp the concepts.
鈥淚 need to give them a traditional structure,鈥� he said.
He thinks though, that ultimately textbooks won鈥檛 be the way to teach these skills. The landscape is just changing too fast for that. 鈥淚 don鈥檛 think [the] textbooks [available] are that good,鈥� he said. 鈥淚 don鈥檛 think they are the answer in general.鈥�
Other tweaks for next year include bringing in guest speakers, so that students can start to get a sense of how their work relates to what鈥檚 going on in industry. He鈥檚 already excited to welcome the AI club from the local university. He鈥檚 also hoping students can get credit from the local community college for his class, down the line.
High School Kids Can Do AI
There鈥檚 nothing magic about being in your 20s that prepares you to study programming in AI applications Johnson said. Seventeen year olds can handle it. 鈥淪ome students are more than ready,鈥� he said. 鈥淚 can say honestly I had students more advanced than myself by the end of the year. I pointed them in the right direction and then they zoomed ahead. They could have done it at 16, they could have done it at 14.鈥� He thinks eventually the same principles he taught could be extended to middle and even elementary school classes.
Image: Getty
Want more on artificial intelligence? Check out these good Ed Week reads: