IBM chip 'senses' events to react like brain

International Business Machines Corp has developed a computer chip inspired by the human brain that may predict tsunamis and highlight risks in financial markets, the company announced today.

The technology, called cognitive computing, is programmed to recognise patterns, make predictions and learn from mistakes, has human-like capabilities not possible using today’s best computers. It’s a sharp departure from traditional chip design concepts, IBM said.

Systems built with the new chip can synthesise events currently occurring and make decisions in real time, the Armonk, New York-based company said. Sponsored by the US Department of Defense, IBM has received $21 million more in funding after today’s developments to bring the technology to applications. The advancement marks the end of a conceptual phase and initiates the first step to find out how to bring the chips to scale for production, said Kelly Sims, an IBM spokeswoman.

“We’re inventing a new system, changing the game,” Dharmendra Modha, the project’s leader, said in an interview. “It’s a new generation of computers, bypassing the hurdles faced by today’s computers.”

While current computers handle commands individually on a linear if/then basis, Modha said machines equipped with the new chips will “rewire themselves on the fly.” Without any set programming, the devices reach decisions through integrated memory, computation and communication cores that resemble synapses, neurons and axons, respectively, in the brain’s nervous system. The sensor network is three times larger than any ever created, he said.

REACTING TO SURROUNDINGS
Cognitive computers may react to taste, touch, smells and sound, Modha said, while consuming less power and volume than today’s technology.

The core of the device has 3.8 million transistors in a 4.2 millimetre square of silicon, a tiny scale for what the chip accomplishes, said Richard Doherty, research director at Seaford, New York-based Envisioneering Group, a technology assessment firm.

“Research like this has been going on for decades, but this marks the beginning of commerce, industry and profits for cognitive computing,” Doherty said in an interview. “They’re going to be more like humans and primates than ever before.”

Current chips using two billion transistors couldn’t perform similar tasks, he said, and they’d have to be 10 times as large.

“We’re giving birth to a new type of architecture-enabling applications to ask computers to help us with today,” Modha said.

POTENTIAL USES
Some of that help may come, for example, in the form of predicting tsunamis or creating financial opportunities, he said. By constantly recording and reporting temperature, pressure, wave height and ocean tides, such a system could issue a warning for giant waves based on its decision making.

“In a financial sense, it would be looking not for micro opportunities, but a macro pattern that spots anomalies, risks, and creates new possibilities,” he said.

The technology also might be used in a doctor’s glove, providing real-time patient data during surgery, based on the body’s texture, smell and temperature.

“The impact of this is inevitable, but the timing is unpredictable,” Modha said. “As a civilisation, we’re computing with half a brain and now we’re bringing forth that other half to complement it.”