The Turing Test Gets Murkier

Computers can’t “think” like humans largely because the way their processors are designed isn’t the way our brains process information. Computer chips are linear — electrons flow from one on-off switch to the next.

That’s great for adding numbers and performing mathematical sequences. An Excel spreadsheet is designed to take advantage of that. But the human brain is a far better multitasker.

Each brain neuron is usually in contact with thousands of other neurons, so pathway possibilities are almost infinite. That’s why, despite our best attempts to create software that will recognize faces, we’re not very successful. The human brain, by contrast, is masterful at recognizing faces. (Nonetheless, the federal government keeps pouring millions of dollars into devices that contractors insist will one day be able to pick out bad guys at airports.)

IBM has been working on a more human-like computer for six years, but not by trying to tackle the problem through software. It’s changing the machine itself at the very level at which it computes — with chips that communicate like brain cells. The chips are called neuromorphic, and IBM has a brand name for them — TrueNorth. Your brain has at least 100 billion neurons, and by being linked to each other chemically via synapses, they can create 100 trillion connections.

One TrueNorth chip can produce 46 billion synaptic operations per second, says IBM. Each TrueNorth chip has a million neuron-like circuits that can be individually programmed. That means there are 256 million “synapses” on a chip that can be programmed.

To do this, IBM places 5.4 billion transistors onto each chip, the second largest number ever put on a chip. Dharmendra S. Modha, an artificial intelligence guru at IBM and head of cognitive computing at the company’s Almaden Research Center, recently wrote that “TrueNorth has a parallel, distributed, modular, scalable, fault-tolerant, flexible architecture that integrates computation, communication and memory and has no clock.

It is fair to say that TrueNorth completely redefines what is now possible in the field of brain-inspired computers, in terms of size, architecture, efficiency, scalability and chip design techniques.” One chip is not even close to what a human brain can achieve, but IBM is building a machine with 4,096 of these chips that researchers believe can solve complex problems like driving a car, a relatively easy task for humans that despite the extraordinary success of Google’s self-driving autos, is still a challenge.

Chips like TrueNorth could revolutionize robotics (think RoboCops that replace the uneven services of human police), medical imaging machines (who needs a radiologist to look at the scan with chips like these?), facial recognition (this could really speed up TSA lines), detection of alien signals from across the galaxy and multimedia applications like whole-wall 3-D movies in your home.

But chips like these are frightening, too. We are fortunate in many ways that artificial intelligence programming has mostly failed. Machines are not smarter than we are and cannot take over the planet — yet. When the hardware itself, not the software, can grow and learn, we could be in real trouble.

To a bright future,

Stephen Petranek
for The Daily Reckoning