Our pursuit of AI (Artificial Intelligence) has recently taken a giant leap forward, with the help of a tiny little worm.
You may have seen headlines in the last few days, trumpeting the success of a project undertaken by the OpenWorm project, whereby they built a Lego-robot and controlled it with the brain of a nematode. This is a truly incredible thing, but it requires much explanation.
AI has been at the forefront of our minds of late, what with multiple block-buster films centered on the subject in the last couple years. And with prominent scientists, such as Steven Hawking, offering chilling if not somewhat alarmist warnings of the potential for AI to quickly decimate, replace, or simply make obsolete, the human form. This breakthrough comes at a time when we are both most in need of such solutions, and most wary of its moral implications.
None the less, the breakthrough has occurred, and here are the details.
The OpenWorm project is an international, open-source effort by various scientists and organizations to map the neural connections of the brains of worms. Their primary focus is in mapping, understanding, and successfully simulating neural function. Per their website, they admit that the field of neuroscience and its complimentary subjects are advancing at a rate too fast for anyone to keep up with all of the findings, experiments, projects, and papers being churned out around the world. As such, they’ve chosen to ignore most of it, so to speak, and instead focus on the simplest neural systems on the planet – those of worms.
Now, we’re not talking about your standard bait worm, in this case they’re working with nematodes, or roundworms; specifically Caenorhabditis elegans (more commonly known as C. elegans). Nematodes and other types of worm, such as planaria (which are fascinating in their own right), are commonly used in neuroscience because their brains work, fundamentally, the same as ours, but they are much, much smaller and simpler. Thus scientists are able to identify and work with the most basic brain functions on a primary level.
OK, on to the good stuff.
Contrary to what some of the headlines you may have seen on this are claiming, the OpenWorm project scientists didn’t connect a worm’s brain with a Lego robot. Not at all. What they did was create the most detailed and accurate map of neural connections ever, and they then recreated those connections inside a computer. After which they connected that simulated nematode brain to a robot (which happened to be made of Lego for ease of use) that contained sensors which were analogous to the sensory stimulus of which the worm’s brain would already have been familiar. Then they watched.
Amazingly, without receiving any programming input, the simulated brain began causing the robot to move, and not only that, it responded to environmental stimulus with appropriate action. It would stop short of bumping into objects, and reverse. It responded to food stimulus (simulated), and danger stimulus.
What the OpenWorm researchers have achieved is virtually identical to what was portrayed in the movie Transcendence, starring Johnny Depp. If you recall the plot of that film, Depp’s character was dying, and in order to save him, he and his wife uploaded his consciousness into a prototype computer system, wherein he could live on as an Artificial Intelligence simulation. They did so, not by physically transferring his “consciousness”, but by recreating the neural connections in his brain, in a virtual framework. This was why the prototype computer system was so important; it had to be capable of managing the massive number of connections in his brain. See? Identical, except in scale.
The human brain has somewhere in the neighbourhood of 100 billion neurons, supported by 100 trillion neural connections. If you wanted to count each connection by hand, it would take about 3 million years. Yeah, that’s a lot of connections. Simulating a human brain in the same fashion is out of our reach, at the moment, simply due to the sheer size of the job.
C. elegans, however, has only 302 neurons, and roughly 900 connections. Which is obviously much more manageable. But as mentioned, the procedure is much the same and the outcome identical. Except that worms – nematodes, planaria, what have you – also have much less cognitive ability, and therefore the results seen in this experiment aren’t necessarily scalable, which is really the point of all this research. The very basic motor control achieved here is virtual child’s play compared to managing the interaction of billions of neurons. It may not ever be possible.
Here’s the thing to keep in mind though, we have achieved AI. This is, by its very definition, a simulated intelligence that is housed and operated in an entirely artificial environment. There is no biological component to this worm-Lego-robot. It is exactly what so many of our stories and movies have conjured, if not a good deal more frail and non-threatening.
This is the moment future historians will mark as being either our salvation or our doom. The Transhumanists among us are likely celebrating this as a long sought achievement, a notch on the bed post of their trek to immortality through technology. The rest of us, and likely actual living-cyborg Steven Hawking, are left to wonder where this will actually lead. As with most such breakthroughs, we again have more questions than answers.
So, when the machines rise up, and humanity is on the brink of extinction, we’ll have worms to thank for our eventual doom. Though, if they’re made of Lego, how hard can they be to kill?