I came across a recent story at BBC News, one which makes me both hopeful and fearful. It seems that a team of researchers, working for Google, have completed work on an artificial neural net that is capable of recognizing pictures of cats. Designed and built to mimic the human brain, this may very well be the first instance where a computer was capable of exercising the faculty of autonomous reasoning – the very thing that we humans are so proud (and jealous) of!
The revolutionary new system was a collaborative effort between Google’s X Labs division and Professor Andrew Ng of the AI Lab at Standford University, California. As opposed to image recognition software, which tells computers to look for specific features in a target picture before being presented with it, the Google machine knew nothing about the images in advance. Instead, it relied on its 16,000 processing cores to run software that simulated the workings of a biological neural network with about one billion connections.
Now, according to various estimates, the human cerebral cortex contains at least 1010 neurons linked by 1014 synaptic connections – or in lay terms, 10 trillions neurons with roughly 1 quadrillion connections. That means this artificial brain has one one thousandth the complexity of the organic, human one. Not quite as complex, but it’s a start… A BIG start really!
For decades – hell, even centuries and millennia – human beings have contemplated what it would take to make an autonomous automaton. Even with all the growth in computer’s processing speed and storage, the question of how to make the leap between a smart machine and a truly “intelligent” one has remained a tricky one. Judging from all the speculation and representations in fiction, everyone seemed to surmise that some sort of artificial neural net would be involved, something that could mimic the process of forming connections, encoding experiences into a physical (i.e. digital) form, and expanding based on ongoing learning.
Naturally, Google has plans for an application using this new system. Apparently, the company is hoping that it will help them with its indexing systems and with language translation. Giving the new guy the boring jobs, huh? I wonder what’s going to happen when the newer, smarter models start coming out? Yeah, I can foresee new generations emerging over time, much as new generations of iPods with larger and larger storage capacities have been coming out every year for the past decade. Or, like faster and faster CPU’s from the past three decades. Yes, this could very well represent the next great technological race, as foreseen by such men as Eliezer Yudkowsky, Nick Bostrom, and Ray Kurzweil.
In short, Futurists will rejoice, Alarmists will be afraid, and science fiction writers will exploit it for all its worth! Until next time, keep your eyes peeled for any red-eyed robots. That seems to be the first warning sign of impending robocalypse!
3 thoughts on “The Future is Here: The Google Neural Net!”
I’m seriously considering going Sarah Connor on Google’s headquarters. Who wants to be the Terminator?
so this is how it will end huh? a smart search engine that just snaps one day, kind of like all those mail carriers.