How Aussies are evolving IBM's Watson

IBM Australia is looking to maintain its first-mover advantage in the burgeoning enterprise artificial intelligence (AI) space by putting more resources behind its cognitive drive.

The vendor has spent 2016 arming its commercial operation with a new team that has the industry domain knowledge to drive Watson harder into specific verticals.

Its local Research team has revealed involvement in cutting-edge projects for IBM’s nascent Watson Health business, as well as around neuromorphic computing, which is seen as the next evolution of Watson.

Since IBM created the US$1 billion ($1.3 billion) Watson Group in 2014, the vendor has been selling business on the benefits of cognitive computing, while at the same time trying to create as many applications and use cases for the technology as it can.

While these efforts were initially internally-driven, the growth of Watson over the past two years has increased the number of people and organisations working with the technology, with a growing amount of this work being done behind the scenes.

One of the challenges with tracking Watson’s progress in the enterprise right now is that there are many frontiers for the creation of new applications for the technology, and many of them sit outside IBM.

iTnews reported in July that Australia’s big four audit firms are racing to build cognitive and artificial intelligence (AI) practices.

But there are other frontiers, and IBM’s executives acknowledge that emerging applications for Watson are increasingly defined by customers and partners, potentially without IBM’s input.

“Often people are just doing things by themselves now,” Watson Asia Pacific director Jason Leonard said.

“There might be banks with their own innovation teams doing lots of projects that we don’t even know about [because] the Watson technology is now on [IBM’s cloud app development platform] Bluemix.

“They may be prototyping or have gone further than that and built applications that are even in production right now, and we don’t know about it.”

That means the user base for Watson in Australia is likely to be much higher than the handful of publicly-known early adopters - ANZ Bank, Defence, Woodside, and Deakin University.

New team, internal 'Kickstarter'

But IBM Australia is by no means taking a backseat approach to the growth of Watson in Australia.

At the start of this year, the company embarked on “a real expansion phase” for its internal capabilities, creating a new cognitive solutions team.

“The idea of that was not just to have people running around talking about the technology, which is the way we were treating it,” Leonard said.

“[For example, before 2016] I might go to a telco but the next day I might go to a bank or an insurance company or a government agency. But when you talk to a social security agency versus a telco their requirements are quite different.

“So we built out our industry specialisation to complement that. In Australia, our focus is on the financial sector, government and natural resources.”

IBM Australia has built a team of between 50 and 60 people that work directly with Watson.

Excluded from that “are all the people in IBM Research Australia and also IBM Global Business Services who may be working on cognitive projects or proposals as well”, Leonard said.

“In some ways a lot of the ‘other’ parts of IBM are starting to inject cognitive elements into all of their proposals as well to differentiate themselves in the marketplace,” he said.

More interesting applications for Watson are likely to emerge in the coming months, courtesy of Cognitive Build, which has been likened to an “internal Kickstarter” for Watson applications.

All 400,000 IBM employees worldwide were asked to form teams of a half-dozen or so employees and “come up with some cognitive idea”, Leonard said.

“There was a big effort two-to-three months ago to get everyone to form their [Cognitive Build] teams, come up with their ideas, and create their pitches.

“The idea might be something to help IBM be more efficient or productive, it might be more customer-oriented or it could be something that was just good for the world at large.

“Then everyone was given ‘monopoly money’ and everyone became a bit of a venture capitalist and voted on what they thought were the best ideas. That then bubbled up to a top ten that received further funding.”

Cognitive Build “created a whole spectrum of ideas”, Leonard said. IBM is now working to “get the best ideas built out”. He did not provide detail.

Pushing the state-of-the-art

Further applications for Watson are being driven out of IBM Research Australia, including via industry partnerships.

Watson owes its origins to IBM Research – and may well owe its next generation to work presently occurring in the labs, including in Australia.

“The Research side [of IBM] are always continuing to advance the state of the art of what we can do,” Leonard said.

Australian researchers are already leading several cognitive health projects, such as melanoma detection in partnership with Melanoma Institute Australia.

They are also playing a part in international projects, like the creation of a cognitive assistant to help ophthalmologists diagnose eye conditions, and more broadly around cognitive applications for medical imaging.

“About three years ago we started doing a piece of work that within Research we described as a grand challenge,” vice president and lab director for IBM Research Australia, Dr Joanna Batstone, said.

“It was in essence looking at how we enable cognitive technologies to understand and interpret medical images.”

Australia is one of three IBM labs worldwide that participated in the project; other participants were the Haifa labs in Israel and Almaden labs in San Jose.

“As that body of research spread across our labs, we then looked at how do we transform IBM with some of these investments in medical imaging?” Dr Batstone said.

“In partnership with IBM and our broader Watson Health team we acquired Merge Healthcare to become our Watson health imaging portfolio of capabilities, which then provides the route to market for the medical imaging technologies that flow out of the Research labs.”

The next generation

While much of the work undertaken to date has been around the development and application of software algorithms and APIs, the future of Watson may lie in hardware – and it’s a research area in which Australia has an active role.

“One of the interesting opportunities [for cognitive] is in rethinking the way that we design computers,” Dr Batstone said.

“So in addition to the software side of cognitive – the algorithm development and the convolutional neural networks that we build up – what we’re also doing within the IBM Research labs is developing what we call brain-inspired computing techniques.”

The present focus of the research is on creating computer chip architectures that are “designed to model the human brain and synaptic information flow within the human brain.” The first of these – TrueNorth – was first detailed in mid-2014 in the journal Science.

Dr Batstone believes the chips – which are part of a research effort under the name “neuromorphic computing” – will form the basis of “the next generation of cognitive technologies”.

“Within the Research labs here in Australia, we’re looking at how we can leverage some of these brain-inspired computer architectures to do pattern recognition, for example, to medical images or other types of visual analytics as opposed to leveraging, for example, the conventional computer architectural chip that we would find in our computers today,” Dr Batstone said.

TrueNorth is also being put through its paces by “an ecosystem of university collaborators who are now running their own experiments on our neuromorphic chip.”

“So at the moment we are actively working within a broader ecosystem to build up new ideas, applications, patterns, scenarios - in essence, business challenges - that would leverage a neuromorphic approach of solving them as opposed to a traditional network architecture,” Dr Batstone said.

 “We’re actively engaged with universities and we’ve got some commercial partners that are working with us directly on deployment of that technology.”