IPsoft is a company founded by a mathematician and based on mathematics – and that becomes abundantly clear just a minute or two into a conversation with its founder and CEO, Chetan Dube. He talks enthusiastically of using algorithms to map the inner workings of the human brain, to enable computers to process information like humans do – only better, because computers can do things faster, without errors or coffee breaks. IPsoft is putting the technology to create more efficiencies in IT outsourcing (ITO) and business process outsourcing (BPO).
NTT Com: IPsoft provides a number of different services and tools. In a nutshell, how do you describe what the company does?
Dube: IPsoft’s mission statement is to power the world with expert systems. We believe we can create a more efficient planet by leveraging expert systems, whether it’s in IT outsourcing or in business process outsourcing.
NTT Com: Can you explain the idea that IPsoft is based on mathematics and talk about the connection between IT and mathematics?
Dube: In a previous life, I was assistant professor at New York University and conducted research in mathematical sciences centered around the deterministic finite state machine. We had a bunch of faculty from NYU and DMTS [Distinguished Members Technical Services] people from AT&T Bell Labs who were of the opinion that the way infrastructure management practice and business process management practices were being orchestrated was fundamentally flawed, based on replacing labor with cheaper labor. We felt labor needed to be replaced by expert systems. That would be how you could deliver sustainable and quantum gains in efficiency. Our founding tenet was the infrastructure of tomorrow would not be managed by people but by expert systems. Mathematics is fundamental to modeling human neural activity [and building expert systems].
There are very few other markets that can hold a candle to that market size and growth rate. NTT America has ambitious plans for starting to take a dominant position in this explosive market space.
NTT Com: What are deterministic finite state machines?
Dube: They are machines that move from state space to state space based on inputs coming in, and the transitions are deterministic – that’s why they’re called deterministic finite state machines. You and I are in a certain state currently. You asked me a question. Based on that input coming in and based on a set of variables around me, I answered the question and move to another state. So a human brain and an engineering brain similarly examine problems with network interfaces, it studies all the inputs that exist in the network environment and the variables and based on some of the inputs coming in, it makes a transition to another state where it emits an action sequence.
That was primitive. Since then we’ve spent about a decade plus 3 or 4 years of research on this and the cognitive models have evolved significantly in that time. Around 1950, the father of computing science, Alan Turing, said the age of machine intelligence will have arrived when you cannot discern between human intelligence and machine intelligence – so you couldn’t determine whether you were having a conversation with a human or a machine. We are on the precipice of that.
NTT Com: That puts me in mind of IBM’s Watson and Jeopardy! challenge.
Dube: We have studied Watson’s algorithms in depth. Watson is a fantastic Q&A answering machine. In game playing, the universe can be normatively bounded. A large part of the engine is based on statistical Q&A analysis and co-referencing and semantic role processing, which are all centered around pinpointed atomic questions.
The human brain doesn’t work like that. Neural processes are fairly complex, they are not just statistical, elaborate, exhaustive search engines, which can probe multiple petabytes of storage and exhaustively search statistical databases to find a hit. The human brain continually builds neural images over time. Why is it that a child who has never been taught a single line of English and who has never gone to school can still speak English, or any language? The human brain observes and is an adaptive learning machine that can build neural images rapidly and adaptively. The two approaches are very different. Jeopardy is a classic example. It took 4 years for a team of IBM scientists to take all the Jeopardy trivia and distill it, distill it, distill it into elaborate knowledge representation ontologies. This is taking an elaborate amount of information and breaking it down to represent it in knowledge structures that can be queried by statistical search engines. The way to approach the answer to Turing’s challenge would be not to look at an adult brain and try to distill all that information, but to look at a child’s brain and study how it learns, and to emulate that process. One is a top down approach, the other is bottom up. We believe bottom up is only the sincere, scalable approach to emulate human intelligence.
NTT Com: Does IPsoft strictly offer services or is its technology available for customers to use on their own premises?
Dube: We offer services. Some customers have chosen to adopt our technology and put it in their own data centers but IPsoft engineers run the platform as a service. One of the largest banks has chosen to adopt IPsoft autonomic technology for serving its fixed income and derivative desks. IPsoft will be responsible for the availability and uptime and services rendered through that autonomic platform. We have a simliar agreement with NTT America where we are providing the platform for them to be able to provide autonomically managed infrastructure services to their customers. We believe it’s a transformative partnership because NTT has the promise of the global reach of the customer base. As opposed to traditional methods where most infrastructure was managed by human processes, NTT will be able to render services where the infrastructure will be self-managed, self-healing and self-governing.
NTT Com: What kind of benefits will that provide for NTT customers?
Dube: NTT will be leveraging our technology to service their customers in a more efficient way. If you look at the ITO market, it’s about $77 billion with BPO about $421 billion. It’s a combined market of about half a trillion growing at 9% to 14% year over year, depending on which analyst you talk to. There are very few other markets that can hold a candle to that market size and growth rate. NTT America has ambitious plans for starting to take a dominant position in this explosive market space. To win big, a couple of things are essential. One is differentiation. You need to approach the marketplace with a differentiation that makes your services stand out from services rendered from every other company, even an offshore provider that’s just replacing labor with cheaper labor. In this case, NTT will be able to offer customers a transformative proposition of not just replacing labor with cheaper labor, but replacing labor with expert systems.
The second proposition NTT will be able to advance is a significant improvement in the quality of services rendered. No manual processes can come close to competing with the times to resolution that are offered by automated process. The business outcome and the quality of the customer experience will be radically improved as compared to the general populace. So NTT is planning on distinguishing itself on both of those fronts in partnership with IPsoft and autonomic technologies.
NTT Com: I love the “transformational minds” feature on your web site, listing accomplishments of some famous thinkers like Pascal and Fermat. But I could not get my head around how Georg Cantor proved that one infinity could be bigger than another infinity. How can that be?
Dube: Cantor came up with aleph-naught, a number that counts all the integers — whole numbers without fractions — that there are in existence. Aleph-naught has to be infinity, since there are an infinite quantity of whole numbers. But then what about real numbers? Real numbers include rational numbers, and irrational numbers, like the square root of five, and integers. This has to be a greater infinite number than all the other infinite numbers. He put boundaries on something so inherently unquantifiable.