As the field of artificial intelligence grows increasingly complex, creating AI models can be a daunting task for beginners. One University of Saskatchewan student is looking to address this challenge with a new software called clevrML.
Second-year computer science major Landon Garrison developed the online platform with the idea of “[bridging] the gap” for those who are less familiar with AI. He says deploying artificial intelligence models typically requires writing a lot of code and going through a complicated learning process, but his software provides an alternative.
“Regardless of your experience or knowledge in AI, clevrML aims to make it easier, and most importantly, more affordable,” Garrison said.
At its core, clevrML is an online platform to build and deploy AI. Garrison says that neural networks, which are algorithms modelled after the human brain, are often expensive and require copious amounts of time, data and power to generate accurate models. However, clevrML uses a method called Active Memory Learning instead, developed by Garrison.
“What Active Memory Learning tries to do is take away a lot of the shortcomings of neural networks,” Garrison said. “Overall, it’s just a different technique used to make intelligent machines.”
Developing AML was the most challenging part of the process for Garrison. Implementing AML into a workable program was also difficult, but the result of over four months of work is a software that operates in text and image classification as well as forecast modelling.
cleverML is an alternative to other large cloud platforms that make machine learning models but come with variable costs, no model editing options, and longer build times. Additionally, Garrison says that clevrML offers a variety of different uses depending on the user’s needs.
For example, for people in the business industry, clevrML’s forecast modelling tool can be used for forecasting business workflow and inventory. For developers, the program can be integrated with other applications to lessen the learning curve and costs often associated with using AI models. Garrison says his software is designed for those interested in AI but lacking the skills or experience to develop models.
“If you’re interested in AI at all and you want to learn … this is a good place to start because you won’t break the bank,” Garrison said. “I tried to make it a lot cheaper and it’s very intuitive.”
Garrison says he spent the summer working on the project. Creating a workable demo and building off of previous versions were the bulk of the work.
“I genuinely want to see more people involved with this and more people be able to use some of the technologies that are out there,” Garrison said.
As for what’s next, although he cannot speak about them just yet, Garrison says the public can expect updates for clevrML.
“I’ve got to keep it quiet for now but in a couple months, I’ll be pushing updates that will even push the boundary a little more on what’s possible,” Garrison said.