Danforth students heading to national science fair

Eat dots. Not too many. All greens.

To borrow from food writer Michael Pollan, that is basically how life ticks in the artificial ecosystem designed by Evan Farrell and Max Musing, two Grade 12 students in the MaST program at Danforth Collegiate and Technical Institute.

It looks simple, but their computer program tests a big idea. And it has earned Farrell and Musing two of just 10 spots for Toronto students at this year’s Canada-Wide Science Fair in Fredericton, New Brunswick.

Evan Farrell, left, and Max Musing run the simulated ecosystem program that won the two Danforth CTI students a gold medal at the Toronto Science Fair on March 28, plus an invitation to the Canada-Wide Science Fair in Fredericton, New Brunswick in May. Their software program is designed to test how quickly a simple artificial intelligence can learn under different stress conditions. PHOTO: Andrew Hudson
Evan Farrell, left, and Max Musing run the simulated ecosystem program that won the two Danforth CTI students a gold medal at the Toronto Science Fair on March 28, plus an invitation to the Canada-Wide Science Fair in Fredericton, New Brunswick in May. Their software program is designed to test how quickly a simple artificial intelligence can learn under different stress conditions.
PHOTO: Andrew Hudson

On screen, the program shows nine big dots with pointers that can turn, stop, or move ahead in search of smaller green dots – their food.

“When we start off the simulation, we don’t actually program them to do anything,” said Musing, watching as dots scurried across a laptop screen like single-cell organisms.

“We don’t even tell them, ‘You have to eat food to survive.’”

But the dots can learn.

Farrell and Musing programmed tiny ‘brains’ for the dots, each with 14 artificial neurons. (Human brains, for the record, have about 86 billion real ones).

Still, after three minutes, a lifetime in dot world, many of the big dots die.

Some find too little food to eat. Others eat too much.

What makes Farrell and Musing’s program so interesting is what happens when the dots survive.

Like DNA, surviving dots can pass on what

A screenshot shows Max Musing and Evan Farrell's artificial ecosystem in action.  The big dots with pointer lines represent organisms hunting for food, represented by the smaller green dots. PHOTO: Andrew Hudson
A screenshot shows Max Musing and Evan Farrell’s artificial ecosystem in action. The big dots with pointer lines represent organisms hunting for food, represented by the smaller green dots.
PHOTO: Andrew Hudson

ever food-finding strategies they learn to the next generation. Throw in a few random mutations, and their offspring can learn even more.

“It’s basically how evolution works – survival of the fittest,” said Musing.

Given enough time – about 800 generations or so – the dots can evolve to the point where more than half survive.

Farrell said they ran four basic scenarios to see how quickly the dots reach that peak – in each one, they set out a different amount of food.

“What we’re actually testing for is how an environmental pressure affects their rate of learning,” he said.

What they found is that given tons of food, the dots learn very slowly. They learn faster when food is harder to find.

But give them too little, and most dots starve no matter how smart they are.

In other words, necessity is the mother of invention – to a point.

Farrell said he and Musing started thinking about the project last summer. They were inspired by another Danforth science project that won a silver medal at last year’s city finals.

“They had a traffic light system, and the lights basically learned how to get traffic to flow best through a simulated city,” said Farrell.

“I thought that was the coolest thing ever.”

After their own project won gold at the Toronto Science Fair this spring, Farrell and Musing got invited to a research conference where they met a PhD student, Trevor Bekolay, whose own work involves using a simulated brain to study how sound gets translated into speech.

With one month before they compete at the Canada-Wide Science Fair in Fredericton, New Brunswick, Farrell and Musing are adding tweaks to their program and thinking more about the real-world applications it could have.

Many of the big prize-winners at nationals are things people can use — last year, an Ottawa teen won the best-in-fair award and $16,200 for designing a low-cost 3D medical scanner that can be used to make prosthetic limbs.

Farrell and Musing said their quick-learning dots might be a helpful model for building farming robots, or robot garbage-collectors.

They could even have off-world applications, helping engineers to design extra-terrestrial rovers that can search efficiently for meteorites on the surface of some far-flung planet.

“That’s what we’re gunning for — Mars robots,” said Farrell, smiling.

email
Click here for our commenting guidelines.

Leave a Reply

*