Did Microsoft just take us one step closer to the point when robot overlords dominate humanity? The firm announced yesterday that its Project Adam seeks to emulate the human brain by creating a network of interconnected serves that mimic a human brain’s neural network by being able to.
So what can this human brain-like network of supercomputing do?
So far it’s only been used as an image recognition system that is able to tell which breed of dog is shown in an image with better accuracy than most humans can do. In one of the benchmarking tests used to test Project Adam’s process, called ImageNet, it was able to correctly sort millions of images into around 22 000 categories 29.8% of the time compared to around 20% accuracy for the average human.
That might not sound like useful ice breaking trick at parties, but apparently in AI research terms it’s huge.
Project Adam divides the calculations it needs to do over several different nodes of the network which then combine their results to get an answer using a technology called HOGWILD. It allows the independent operation of different processors which can write their results and even overwrite the results of other processors to the same memory locations which would logically make you think that data would get lost, slowing things down, but in fact has the opposite effect in that these data collisions happen very infrequently.
So Project Adam isn’t close to becoming the human race-destroying SkyNet from the Terminator movies just yet but it isn’t too far fetched an idea to imagine the image recognition powers of Adam being integrated into the Cortana digital assistant that is coming soon to Windows Phones with the 8.1 update of the operating system.
Check out the video below to get more of an idea of what Project Adam is about and wonder whether a computer really needed to be taught the difference between the only two breeds of corgis?[Source – Microsoft, Via – Wired]