Computers these days they just don’t learn to wish you could be as smart as a computer well computer engineers wish they could build a machine as smart as you.
Even your dog okay so there are certain tasks that computers are obviously better at than we are for example quick what’s the square root of seven thousand thirteen now while all of you were trying to cheat by opening up a calculator app whatever device you’re actually watching this on could have calculated the answer to a thousand decimal places but can that same computer take a picture of me.
Recognize what it’s looking at with dedicated facial recognition software written by humans I must add it probably can but that requires a lot of work on the front end to make it happen you see when it comes to recognizing visual images or a style of a piece of music the natural architecture of a human.
Even an animal brain is at a huge advantage so what if we could make computers think more like brains now a computer is based on von Neumann architecture that’s where you have a CPU and a separate memory bank that CPU performs operations based on binary data.
That’s essentially an electrical on or off switch and it can really only do one of those at a time until an entire program is fulfilled now your brain is also an information unit but it works in a very different way it’s all based off the neuron neurons can communicate with other neurons through electrical and chemical signals.
You have a web of around eighty to a hundred billion neurons in your brain they’re constantly making and breaking and reinforcing connections as you form new ideas now it’s true that your brain executes individual computations at our rate much slower than a computer chip but your brain actually has some enormous advantages, for one thing, it’s massively parallel see a traditional computer has to solve operations one at a time while.
Your brain can be working on million at the same time it also is really energy efficient it uses a fraction of the wattage that IBM’s Watson supercomputer used you know that’s the one that beat the human champions on Jeopardy and finally your brain is naturally designed to learn or in computer terms to reprogram itself as it does work it redesigns itself to do work better check this out in 2012 researchers with the Google brain project created a neural network of 16,000 processors then they fed it 10 million random unlabeled YouTube thumbnail images.
The computer learned to recognize cats, of course, it did essentially it had to invent the concept of a cat which is pretty much what a human naturalist would have to do if they went to a Balian world and saw alien life-forms you classify stuff based upon visual similarities.
You make the rules up from there meanwhile on a slightly smaller scale the semiconductor company Qualcomm has a neuro-inspired chip in development that could have some amazing implications in machine learning in one demonstration they had a robot working on this system navigating to different white spots on a floor covered with colored tiles now here’s the crazy thing no one programmed the robot to do that the robot just behaved on its own and then got positive reinforcement essentially a good robot message every now.
Then when it did the right thing you know in the short term these brain-inspired computers may help us with everything from energy efficiency to sensory input to big data but in the long term that’s where things get really interesting maybe we’ll no longer have to program computers maybe we’ll just have to teach them and now I’ve got a question for all of you out there if you had a robot with brain-inspired learning Hardware what would you teach it.
What would you not teach it to let us know in the comments below?