In early January IBM announced the launch of its new Watson business unit.

Watson is of course IBM’s supercomputer that is named after the company’s founder. With the ability to think and communicate in a human-like way, Watson is also the computer that beat two reigning Jeopardy! champions in 2011.

It is the ability to think and communicate like humans do – dubbed cognitive computing – that brings with it great challenges for the Watson software development team.

The programming behind a cognitive computer like Watson is enormous. All the possible actions that the computer can take must be programmed in advance. This means the performance we might see on Jeopardy! is just the tip of a programming iceberg – with literally hundreds of hours of programming lying beneath the performance.

One of the tricky things for cognitive computers, that software developers have to deal with, is the ability to deal with uncertainty. Drawing inferences and developing insight from situations (which is how we cope with uncertainty) takes a lot of complicated programming that relies heavily on advanced algorithms and analytics.

One technique that developers in this area are working on is stochastic optimisation. It marries sequences of random variables (as you might find in potential scenarios) with probability theory. Pervasive and predictive analytics are also being used by developers to turn unstructured data about the computer’s environment into something meaningful that predictions can be made from.

While Watson currently performs a ‘showcase’ role for IBM, there are serious uses for cognitive computers. IBM, for example, is working to create ‘cognitive assistants’ that can augment human capabilities not just by data crunching at ridiculous speed, as most computers today do, but by additionally being able to read human scenarios, draw conclusions and advise on outcomes. This would ultimately deliver benefits in trade, medicine, crime solving and humanitarian intervention to name just a few.