Cognitive Computing Chips

Advancement in computer chips over the last few decades have been focused on getting things faster and more compact. Fundamentally the chip design hasn't changed over the last 50 years, is still based on Von Neumann architecture of sequential execution of instruction sets.

IBM Research recently announced a new generation of computer chips that are designed to emulate human brain's abilities for perception, action and cognition. Cognitive chips are designed to process large amounts of data using very little power. Unlike traditional computers it does not require any programming rather learns from experiences, finds correlations, creates hypotheses - and learns from - the outcomes, emulating the human brain.

To mimic the cognitive abilities of brain, IBM brought together the disciplines of neuroscience, nanoscience and supercomputing. The digital circuits inspired by neurobiology constitute “neurosynaptic core” with integrated memory (replicated synapses), computation (replicated neurons) and communication (replicated axons). There are two working prototypes - one core contains 262,144 programmable synapses and the other contains 65,536 learning synapses.

Why is this important? Such a system can find wide ranging use. By monitoring the world water supply it can analyze ocean temperature, pressure, wave height and develop a decision model to accurately predict tsunamis. By monitoring financial transactions it can prevent fraud and spurious activity.

This does sound science fiction. Are we closer to building truly intelligent systems that can replace humans? Only time will tell, for now I am excited to be living in such exciting times.

To read more about the technology, click here.

Incentivized Innovation using Crowdsourcing

Heritage Provider Network, a California physicians group announced a competition, the goal of which is to develop a predictive algorithm that can identify patients that will be admitted to a hospital within the next year using historical claims data. The competition started on 4 April 2011 and will run until 3 April 2013, grand prize money is $3 million along with milestone prizes totaling $230,000.

Contest entrants will be provided with deidentified Heritage patient records collected over a period of 3 years. The contestants can participate individually or as a member of a Team.

Heritage believes this can be game changing in our efforts to fixing American health care system. In 2006 more than $30 billion was spent in unnecessary hospitalizations. By predicting how many days the patient spends in the hospital, health care providers will be able to take better preventive measures thereby reducing unnecessary hospitalizations. This will be a win-win for both patients and the providers.

Heritage isn't the first to use such incentivized competition approach, applying Internet enabled crowdsourcing to solve complex scientific problems. In 2006, Netflix the online movie rental company held a contest, the goal was to improve their Cinematch recommendation engine accuracy rate by more than 10%. The dataset comprised of 100 million movie ratings from scrubbed clean Netflix customer data. According to the website, 51051 contestants on 41035 teams from 186 countries participated with the winner walking away with a prize money of $1 million in late 2009.

This is truly remarkable as it allows the scientists and researchers from around the world to apply their expertise in solving large-scale predictive modeling to solve problems in various fields of science, commerce and politics.