Humanising AI: The 30 year project to build a brain

Since 1984, Cycorp has been discreetly working on an advanced artificial intelligence system, Cyc, that could “humanise” robots and objects with highly developed understanding and reasoning capabilities.

From C-3PO in the Star Wars franchise to Bender in the Futurama television series, the idea of having super-intelligent robots that behave like humans has long fascinated people. Apple’s introduction of Siri gave us a taste of what it’s like to have daily interactions with an intelligent system, but anyone who has used Siri knows that (s)he has limits and flaws.

So what is it that makes Cyc’s artificial intelligence seem less, well, artificial?

The key is a focus on building Cyc’s ability to make inferences so that it can execute commands without needing every specific action pre-coded.

“It’s the difference between someone who understands what they’re doing and someone going through the motions of performing something,” said Cycorp president and CEO Doug Lenat in an interview with Business Insider.

Bendorama-(1)

Cycorp has been quietly building Cyc for the last 30 years in a process that is more like educating than programming, with the goal of implementing the system with human knowledge and logic.

The idea of a computer like Cyc immediately conjures images of robots that can complete daily tasks for you, à la Rosie the housemaid from The Jetsons. While these robots would be immensely useful, Cycorp envisions even more.

In a preview for the Ginormous Systems conference that was held in Washington DC last year, Lenat discussed a future that is revolutionised by intelligent systems, saying “every door, every bicycle, every bridge will have the suitable sort of RFID tip, will have its own address and you’ll be able to go up and have a conversation with it, sort of like you do with Siri today.”

Clearly, Lenat sees Cyc as part of this future. If Cyc could be installed into these everyday objects, “humanising” things could become a reality.

shutterstock_103176044

For now, Cyc is still being developed into the brain-like system that Cycorp envisions, but it is already being put to use as a sixth grade maths teacher. Cyc works with students by acting as one of them.

The student tries to help it understand how to solve the problems, and through this process Cyc learns what the student is confused about and adjusts its behaviour accordingly.

Cycorp’s inference-focused artificial intelligence promises a whole new kind of robot, one that can take many forms but “think” with an intelligence that is unprecedentedly human.


Featured image courtesy of Roger Schultz, first body image is a screenshot from Futurama S06 E17, “Benderama” .


Bio-bot breakthrough: Tiny machines with muscle tissue take a walk

Today, robots can be powered by all kinds of energy sources, from solar cells to chemical fuels to rechargeable batteries. What’s next? A generation of machines powered biologically, using muscle cells that expand and contract just as they do in the human anatomy.

These “bio-bots,” less than one centimetre long, can be precisely controlled because their muscle cells move in reaction to an electric field. The bio-bots walk faster or slower in correlation with the rate of the electric pulses signalled by engineers.

Each bio-bot is comprised of a flexible 3D printed hydrogel base, living muscle cells and two posts that serve as legs. Like the muscle-tendon-bone system in the natural world, the hydrogel structures the bio-bot as a backbone, the cells provide the muscular support and the posts act as tendons.

While biological machines have certainly been engineered before, this new group is the most efficient to date. Bio-bots were first made to walk using heart cells from rats, but because of the heart’s autonomous beating, scientists were unable to control the expansion and contraction that powered the movement.

The research team at the University of Illinois at Urbana-Champaign discovered that skeletal muscle cells provide a better alternative, as they allow the engineers to power the bots on and off and control speed by varying the electric current.

Rashid Balir, leader of the bio-bot study, explained further: “Skeletal muscles cells are very attractive because you can pace them using external signals. For example, you would use skeletal muscle when designing a device that you wanted to start functioning when it senses a chemical or when it received a certain signal.

“To us, it’s part of a design toolbox. We want to have different options that could be used by engineers to design these things.”

So far, muscle power seems to be a promising technology with a wide range of potential medical and environmental uses.

“It’s exciting to think that this system could eventually evolve into a generation of biological machines that could aid in drug delivery, surgical robotics, ‘smart’ implants, or mobile environmental analyzers, among countless other applications,” said Caroline Cvetkovic, co-first author of the study’s publication.

shutterstock_166300766

To make these applications a reality, engineers will continue to hone their control over the bio-bots, implementing neurons within their structures to enable steering capabilities.

Since 3D printing allows for the speedy production of differently shaped hydrogels, the researchers can easily experiment with various models.

Soon, they hope to create a hydrogel backbone that can change directions as a result of different kinds of signals, further expanding the functionality of these “living” machines.