- IBM is known for its splashy demonstrations of new AI-powered research initiatives like 'Project Debater,' a robot that can debate humans, and IBM Watson.
- While those publicity stunts are helpful in garnering consumer interest and media attention, the key for IBM is translating the research into products it can sell to enterprises.
- On Wednesday, the tech giant announced it will begin offering clients applications based on Project Debater. The system can autonomously group together similar documents and understand complex aspect of the human language, like idioms.
- A key role in bridging the gap between the two sides of the business are the data engineers, according to chief research scientist Ruchir Puri.
- Click here for more BI Prime stories.
Sign up here to receive updates on all things Innovation Inc.
In June 2018, IBM pitted a robot against humans in two debates on telemedicine and space exploration — a publicity stunt meant to showcase the tech giant's AI prowess in an increasingly competitive market.
Events like "Project Debater" are progressively more common for Big Blue. The first major even was in 2011 when IBM Watson — the company's signature AI platform — beat famed Jeopardy champion Ken Jennings at the trivia game.
Alongside those two showcases, it is also preparing to launch an entirely robotic ship dubbed the "Mayflower Autonomous Ship," which is expected to cross the Atlantic in September with no humans on board.
But while these consumer-friendly demonstrations draw eyeballs and media attention, the real value is in the applications that IBM eventually sells to corporate clients.
The company is gradually perfecting how it translates buzzy research that can span up to 10 years into actual products, according to Rob Thomas, the general manager of data and Watson artificial intelligence at IBM.
"The relationship is figuring out the right point in time when you start to grab stuff from the 3-5-10-year [research] road map and bring that into the 3-month, 9-month, 18-month" timeline for product launch, he told Business Insider.
On Wednesday, for example, IBM announced that key parts of its Project Debater platform will now be available to clients.
The system can employ natural language processing to cluster together groups of related documents, as well as understand some of the more complex aspects of the English language, like idioms. Such a task is difficult, as the meanings of phrases like "beat around the bush" go beyond the definitions of the specific words.
Business Insider talked to Thomas and Ruchir Puri — a nearly 25-year veteran of IBM who was the chief architect of Watson and currently serves as the top scientist for the research arm of the organization — to learn how the tech giant bridges the gap between the two worlds.
Bringing in the business
At IBM, the conversation on the commercial side typically begins once the research arm publishes its technology outlook, an annual report that provides a high-level overview of all the different applications the company should be exploring.
In 2016, for example, the report highlighted the expected differences between consumer and enterprise AI applications.
While smart speakers, image analysis, and other tools were all the rage among consumers at the time, many companies were just exploring how they could adopt the advanced tech in their operations.
IBM narrowed in on the goals it thought corporations would need most: evaluating the trustworthiness of the algorithms, automation, and natural language processing.
"That was the broad canvas that was painted," Thomas said. "Research was fairly far down the path at that stage in terms of understanding those components."
The commercial team has since released products in each of those buckets.
A key role in bridging the gap between the two sides is the data engineer, according to Puri. And that's largely because those in the position understand how to take data, apply it to the applications under development, and make it useful for clients.
"It is not about the AI expert and the data scientist. [IBM] must ensure that the line of business people, the key stakeholders, are all there" making the case for why the tools are important, he said. And data engineers are "probably one of the most critical personas" involved in those discussions because they can understand how the technology meets the business need.
Learning lessons from Watson
It hasn't always been smooth sailing.
When IBM first started rolling out Watson to enterprises, for example, the company struggled to meet the hype it created in industries like healthcare.
One reason it was so difficult is because IBM tried to create platforms that were company-specific, according to Thomas. Now, it's pivoting to broader solutions that can be used by multiple customers.
"Once we have a feel for the core technology and the use cases, we turn that into much more focused client discussions," he said. Now, "for people that wanted their own version of it, we kindly said it's probably not the right thing for us. So it was really about focusing on the things we knew [the tools] could do well."
So even when Project Debater was still in the research phase, the commercial-side of the business was already thinking about the potential applications — but it can take years before such technology is proven out.
"There are often discussions before [launch], but they are very hypothetical. It was still a little bit theoretical. We didn't know for sure that [Project Debater] was going to work. We thought it was, but we didn't know for sure," Thomas said. "It took that maturing a bit to where we could actually start to formulate use cases."
KPMG, for example, was already using Watson Discovery — effectively a search engine for companies that relies on the unstructured data stored internally across different verticals.
But Thomas and the team knew a tool like Project Debater that could help cluster documents would amplify the platform for the consulting giant. So it went to KPMG and sold them on the latest upgrade.
As more tech giants tout the success they've had in using robots to replace humans in increasingly complex scenarios, the ability to turn research projects into actual enterprise products will become even more important.