Researcher uses ‘fuzzy’ AI algorithms to help people with memory loss

A new computer algorithm developed by the University of Toronto Parham Arabi can strategically store and recall information – just like our brains.

Department of Electrical and Computer Engineering Associate Professor Edward S. Rogers Sr., Faculty of Applied Science and Engineering, also created an experimental tool that leverages the new algorithm to help people suffering from memory loss.

“Most people think AI is more robotic than human,” says Aarabi, whose framework is explored in a paper presented this week at the IEEE Medical and Biological Engineering Society Conference in Glasgow. “I think that has to change.”

Parham Arabi

In the past, computers relied on their users to tell them exactly what information to store. But with the rise of artificial intelligence (AI) techniques such as deep learning and neural networks, there has been a move towards “fuzzier” approaches.

“Ten years ago, IT was all about absolutes,” says Aarabi. “CPUs processed and stored data from memory in an exact way to make binary decisions. There was no ambiguity.

“Now we want our computers to make rough conclusions and guess percentages. We want an image processor to tell us, for example, that there is a 10% chance that an image contains a car and a 40% chance that it contains a pedestrian.

Aarabi has extended this same fuzzy approach to storing and retrieving information by copying several properties that help humans determine what to remember – and, just as critically, what to forget.

Studies have shown that we tend to prioritize more recent events over less recent ones. We also focus on the memories that are most important to us and compress long stories to the essentials.

“For example, today I remember I took my daughter to school, promised to pay someone back, and promised to read a research paper,” says Aarabi. “But I don’t remember every second of what I went through.”

The ability to ignore certain information could supercharge existing machine learning models.

Today, machine learning algorithms scour millions of database entries, looking for patterns that will help them correctly associate a given input with a given output. It is only after countless iterations that the algorithm finally becomes accurate enough to handle new problems that it has not yet encountered.

If bio-inspired artificial memory allows these algorithms to highlight the most relevant data, they could potentially arrive at meaningful results much faster.

The approach could also support tools that process natural language to help people with memory loss keep track of key information.

Aarabi and his team implemented such a tool using a simple email-based interface. It reminds participants of important information based on algorithmic priority and a relevant index of keywords.

“Ultimately, it’s suitable for people with memory loss,” says Aarabi. “It helps them remember things in a very human, very gentle way, without overwhelming them. Most task management aids are too complicated and useless in these circumstances.

The demo is free and accessible to everyone; just email [email protected] for instructions.

“I use it myself,” says Aarabi. “The goal is to get the demo into people’s hands – whether they’re dealing with severe memory degradation or just daily pressures – and see what feedback we get. The next steps would be to build healthcare partnerships to test more comprehensively.

“These days, AI applications are increasingly found in many human-centric fields,” says the professor. Deepa Kundur, director of the electrical and computer engineering department. “Professor Aarabi, in researching ways to better integrate AI into these ‘softer’ areas, seeks to ensure that the potential of AI is fully realized in our society.”

Aarabi says this algorithm is just the start.

“Biologically inspired memory could very well bring AI closer to human capabilities.”

Sharon D. Cole