IBM Unveils TrueNorth, a Chip-Sized Brain

By September 03, 2014

The US multinational information technology and consulting corporation is looking to take computing power to the next level, a move which could help to advance artificial intelligence.

Some commentators are already claiming that the new cognitive computing chip unveiled by US IT giant IBM in early August is the most important advance in computer technology since the 1980s. TrueNorth, as it is known, which boasts an architecture modelled on the human brain, with ‘neurons’ and ‘synapses’ that manage respectively processing and memory, is more powerful and far more power-efficient than previous chips. The TrueNorth chip architecture approximates how the human brain works in that each ‘neurosynaptic core’ has its own memory (the synapses), a processor (neurons) and communication conduits (axons), which all operate together in an ‘event’-driven fashion. This is exactly what happens in the human brain, where only a part of the brain is being used at any given moment. Dharmendra Modha, lead researcher on the Systems of Neuromorphic Adaptive Plastic Scalable Electronics (SyNAPSE) project at the IBM Almaden Research Center, predicts that this major initiative could overcome the performance limits of the von Neumann architecture, the mathematics-based approach on which almost every computer built since 1948 is based.

Building super-powerful chips

IBM’s scientific paper explains how this new cognitive architecture allows the chip’s processing power to be freed up, enabling 46 billion synaptic operations per second per watt. “It’s a supercomputer the size of a postage stamp,” claims Dharmendra Modha. Another advantage is its low power consumption. While current processors use up to 140 watts to run 1.4 billion transistors, TrueNorth only needs 70 milliwatts for its 5.4 billion transistors. This enhancement should lead to the creation of extremely high-performance processors. TrueNorth’s 5.4 billion transistors are woven into an on-chip network of 4,096 neurosynaptic cores, producing the equivalent of 256 million synapses. IBM has also tethered 16 of these chips together in four four-by-four arrays, which collectively offer the equivalent of 16 million neurons and 4 billion synapses, showing that the design can be easily scaled up for larger implementations.

A huge field of applications

This future supercomputer will still not come close to the power of the human brain, which has 100 billion neurons and 100 - 150 trillion synapses. However, it should enables huge advances in artificial intelligence. The new processor will allow machines to capture information about the surrounding world and take appropriate action in consequence. Dr Modha invites us to “imagine traffic lights that can integrate sights, sounds and smells and flag unsafe intersections before disaster happens.” In fact there will be a huge field of applications for the new computing approach. Not surprisingly, the project is of great interest to the US army. Originally launched in 2008, the SyNAPSE project is part-financed by the Defense Advanced Research Projects Agency (DARPA), an agency run by the United States Defence Department.

Legal mentions © L’Atelier BNP Paribas