Program a Brain on Wheels With SpikerBot

A new way to think about thinking

Robots are an exciting way to learn engineering, but writing your first lines of code can drain out the fun. Fortunately, SpikerBot from Backyard Brains makes it incredibly easy to give a robot life-like behaviors, and learn valuable neuroscience while you do.

We met with Co-Founder & CEO Greg Gage and Designer Alex Hatch to learn about the new bot.

Instead of learning about computer logic and if-then statements, SpikerBot’s app builds sketches entirely out of virtual neurons. As they say, “there’s no code, no LLMs, just neurons, synapses, and behavior emerging from a circuit a kid designed.”

If you’re familiar with the book Vehicles by Valentino Braitenberg, then you’ll have some understanding of the philosophy behind this bot. The bot responds directly to sensors, and lifelike behaviors emerge from the simplest sketches.

Grounded in science

Backyard Brains isn’t just making a lifelike robot that’s easier to program than before, they’re also filling a gap in everyday neuroscience education.

The problem is that students learn how a single neuron works in elementary school, but the thread is dropped until college, where (if they choose) students learn how millions of neurons work together. That’s a pretty big gap in between. It’d be like learning what a transistor is and then jumping to Python with no other electronics knowledge. So they want to bridge this gap to help end neuroscience illiteracy.

A new programming language

SpikerBot is programmed by dragging neurons and axons around a brain on screen. It’s less like traditional programming with pages of code to scroll through and more like moving cables on a patch synthesizer. Everything happens at once; you can see the whole picture in one pulled-back overview.

There’s a funny inversion to programming things this way. Behaviors that were once out of reach for beginner robotics, like following a person around, become extremely easy to do. You can make a creature that chases objects with only two neurons.

On the other hand, making a robot count in binary, something simple for normal computer logic, actually ends up being a complicated web of connections.

The focus on natural behaviors and intuition, I imagine, will keep people engaged longer than similar bots. Rather than having to learn the fundamentals of how a computer thinks before making a robot that acts like a dog, you can build a robot that follows humans around right off the bat, then slowly build in dog-like responses at a natural pace.

Another wise move: every change is instantaneous. There’s no code to compile and no waiting to upload a new version. As soon as you change a neuron from excitatory to inhibitory, you see the effect immediately. It happens at the pace of thought and experimentation.

The app is clearly the work of a lot of development and refinement. Usually when something groundbreaking comes along, it ends up being done by one company first, but takes off when another comes along and does it stylishly. In this case they’ve put the time in to make the brand new thing beautiful as well.

There’s a library of sample brain models to start with. You can pick each one apart, edit, and learn from it. Plus, the patterns and techniques used are patterns that show up in actual neuroscience. The bot was developed and designed with feedback from a combination of child educators and professional neuroscientists. Users can discover real neuroscience principles by playing around and trying to get SpikerBot to do what they want it to do.

You can learn simple neurological patterns like recurrent pairs and lateral inhibition, and learn how things like working memory work.

It supports peripherals too, including their existing Spiker:bit board, so you can build a life-like brain robot and control it using muscle signals from your own arm.

One chassis to support

None of this would be possible, of course, without the hardware. The SpikerBot itself is a brain-shaped robot modeled after Backyard Brain’s logo. It recognizes and reacts to its environment in the real world, because it’s filled with sensors: a camera, microphone, and a distance sensor. Also has lights and sounds, both of which were crafted to react more naturally than a typical machine. Sound effects are charming and quirky. Lights pop on and fade out like bioluminescent cells. Connection points enable 3D-printed attachments, so owners can customize their bot to fit its personality.

The lessons are valuable beyond traditional STEM applications. 1 in 5 people have a neurological condition. Giving people a shared language and understanding of building-block brain patterns helps people understand what’s happening with various neurological conditions.

Waddling forward

Backyard Brains has also embraced designing for the long term. The app is free and doesn’t require cloud-based authentication to work. The robot runs on easily-swappable AA batteries instead of constantly-decaying non-replaceable lithium cells. And they support recharging by offering their own custom, high-performance NiMh AA’s.

The campaign ends June 14. The bot will retail eventually at $299. It’s incredible to see what a $299 robot can do, and early pledges to the Kickstarter campaign can get one for $219.

Backyard Brands has delivered wonderful projects in the past, and while crowdfunding comes with risks, knowing that the team has a successful track record inspires confidence.

There’s even more features and capability planned, but part of the development is learning how the general public wants to use it. Despite having a lot of existing behaviors, there’s even more undiscovered. As Alex put it, “a lot we haven’t seen because you guys haven’t built anything yet.”