Published on the 26/09/2016 | Written by Donovan Jackson

Cognitive computing depends on big data automation…

On a recent visit to New Zealand, one of SAS Institute’s top executives provided insights into where BI is headed and it should come as little surprise that automation, artificial intelligence and machine learning are front and centre.

Oliver Schabenberger is the company’s CTO; he made the distinction between ‘weak’ and ‘strong’ AI – and said cognitive computing depends on leveraging big data and machine learning.

Schabenberger played a key role in the development of its ‘next generation’ cloud platform, SAS Viya, which he took pains to point out will enjoy parallel development with its existing SAS 9 suite. (He quipped that Viya is, of course, powerful, “because slow and lame isn’t a good marketing message.”)

Putting product punts aside, Schabenberger offered his insights into the future of BI; first up, artificial intelligence. “AI, as we see it now, is weak. With weak AI, we are only able to solve human tasks in specific ways; take the self-driving car. It does nothing more than what it is programmed to do, which makes that AI limited in its application.”

By contrast, he said ‘strong AI’ means systems that think like a human. “Previous periods of industrialisation have been about manpower and brawn. Now, automating a brain is a far more personal thing. We accept readily industrial automation, but there is an adjustment necessary for automation of cognitive functions.”

Sound scary? It’s an adjustment already well underway, even if it is by stealth. “There is already silent AI that we use very day and it doesn’t bother us. Whether it is Google autocompleting a search or Siri answering a question, we expect and receive a quality of response which continually improves.”

Underpinning AI, weak or strong, is machine learning. Under no illusions, Schabenberger in his presentation to the SAS User Group of New Zealand, said it is “the biggest buzzword ever. Bigger, even, than ‘big data’. But the difference is that it works. Think of that Google search. How did it finish my sentence? Machine learning is everywhere already.”

In turn, machine learning depends on having the raw material to learn from – the ‘only-just-less-hyped-than-machine-learning’ big data. The secret to big data, he added, is to intelligently ignore most of it (though he did note he was paraphrasing someone else’s observation there).

Delivering strong AI, said Schabenberger, depends on enabling ‘deep learning’ with high-performance computing to achieve greater automation and accuracy, with a particular focus on the interactions between computers and people. It’s a move towards ‘knowledge augmentation’ and he said that will rest on being able to take advantage of the power of unstructured data.

“Data on own is not valuable; the issue now is how much time we need to invest to go from seeing data to being able to intelligently ignore most of it. We need to reduce the time to analytic action,” said Schabenberger.

Data scientists today, he added, spend too much time tinkering with the machine – determining which sets of input values might give the best outputs, for example – rather than getting insights.

It is here, he said that machine learning and automation has a primary purpose; the data should do the programming to reduce the time to action.

“The new area is to use modern machine learning to perform those tasks [of preparing information and identifying valuable information from data] while being less dependent on business rules. Classical approaches need too many rules and that can limit the ability to get valuable insights from data. Analytical techniques which learn from the data itself and doesn’t require rules, except for a basic set which establish boundaries, can enhance and amplify cognitive computing.”

And if all that talk of computing which moves towards emulating the human mind still sends a shiver up the occasional spine, some further words of comfort. “This is automation without the need for deep subject knowledge. It can achieve things like figuring out who talks the most at dinner parties so you don’t invite them back,” Schabenberger joked. “This replaces frequent, high volume tasks, but it will not replace creativity.”