AI Is Taking Over Science Fairs
With a blog title like this, I suppose there could be a lot of room for different interpretations here. But, for those of you who are worried, I don’t mean that a sentient robot from the future has come back in time to win this year’s ISEF (International Science & Engineering Fair) so that an alternate dimension could be created in order to save the human race from AI. But, with the way student science fair projects are evolving, this could very well happen. And I think that’s actually a good thing.
What I’m really saying here is that science fairs around the world are increasingly being taken over by AI-related science projects. It’s no longer the case that the occasional algorithm-focused AI project is relegated to computer science or math category. Medical, physics, energy, environmental, chemistry and other subjects are now flooded with AI projects at an ever increasing pace. It’s happening at the regional and state level of science fairs as well as the International.
Last year, I did a quick study of the projects containing the terms artificial intelligence or machine learning at the middle school and high school international fairs. I presented the following data on this research at the AAAI (Association for the Advancement of Artificial Intelligence):
Part of this global craze to rope children into the world of AI is due to international competition. Two years ago China fully adopted AI into their national curriculum and we are just beginning to see some of the research indicating that the programs implemented thus far have been quite effective. In addition Russia and other countries throughout Europe have released AI standards over the past several years, in various stages of adoption. NSF has funded a project known as AI4K12 which is planning to release the United States’ first set of standardized curriculum for K-12 AI education this year (and some of it is already out for public comment, if you visit their github page). One of the key elements of this new standard is the fact that it contains a major section (Big Idea#5) that prioritizes the teaching of social/ethical considerations in AI (a topic I will cover in another article in more detail) which will hopefully help us all avoid any future scenario where a robot has to go back in time to save humanity from AI.
What’s behind all this?
From autonomous vehicles to curing diseases to helping make industrial work easier and faster, AI has the potential to transform the way society works. With such a rosy possibility for the future, there is no doubt it could change the future landscape of the global economy. AI is infiltrating nearly every discipline of science and engineering, and the education world is beginning to realize that AI skills will be critical for all students to know in college and the general workforce, in the very near future. Over the past decade the terms “21st Century Skills” and “computational thinking” have been fairly popular among STEM educators. However, while computational thinking skills teach people about the basic logic and how to think like a computer, AI thinking is the opposite — teaching computers to think like people. Related, but very different skillsets, applicable to numerous fields. In order for a computer to learn how to predict the prognosis a COVID-19 patient, for example, the computer might need to use convolutional neural networks (CNN) to read chest X-ray image data, and cross-compare it to a recurrent neural network (RNN) model which reads and interprets the associated doctor’s notes. The system might also need to use/interpolate other biometric data, and any gaps in the data might have to be filled in with a GAN (Generative Adversarial Network). It’s pretty complex stuff, but then again, so is the process of real human learning. And this example is not someone’s post doctoral thesis; it is being taught right now in 2020 to young students at the border of high school/college in after-school programs in Massachusetts. Different types of data need different types of AI — and there is a whole planet full of natural science and engineering data out there that has not yet been touched by AI. In a world where AI thinking is only in its infancy, and so many fields that can benefit from a little AI, that’s why the possibilities for AI science fair projects are nearly endless.
While the concept of AI has been around for decades, the possibility of doing anything useful in AI has been limited by the physical constraints of the semiconductor world. Since the days of Bob Noyce’s invention of the silicon integrated circuit and founding of Intel (which for many years was the key sponsor of ISEF), the number of transistors in a computer chip the size of a coin have multiplied from mere dozens to over 5 billion. Those transistors in a way are equivalent to the neurons in human brains, and they allow computers to begin to think like we do. Over the past decade, we’ve seen computers beat humans in various games of intellect, and their recognition capabilities rise to the point where they are equivalent to (or better than) human perception. This opens up a wide range of possibilities for AI to be used in places previously thought possible. When conducting a drone strike or guiding a medical laser, you shouldn’t be 70–80% sure of the outcome — you need to be as close to 100% as you can. And pretty close to 100% accuracy is precisely the mathematical territory we are in nowadays, in many fields of machine learning. The fact that MIT has used AI to find an antibiotic that can cure highly resistant bacteria earlier this year, and the FDA has recently approved AI based diagnostic systems that are more accurate than human doctors, it feels like we are on the cusp of some major AI-driven changes in the world. Young students are beginning to see the applications of this potential in everything — and if you read the titles/abstract lists of just about any middle or high school science fairs these days, you’ll see lots of evidence of this.
And to me, it feels very promising for the future.