Quick quiz: what do you visualize when you hear the word robot? Some machine in human form, behaving like a human? Or perhaps one of the massive, spiderlike machines that assemble complex products ranging from automobiles to computers?
In reality, modern robotics follows both paths. However, in spite of the progress with humanlike robots, most business applications of the technology focus on machine forms that are specific to the task at hand, whether it be a Mars Rover or a medical robot assisting a surgeon.
There are exceptions to this rule, however: companies who focus on making robots as humanlike as possible, with the eventual goal of behavior so realistic it can actually fool people into thinking the machine is fully human.
To achieve such levels of verisimilitude, such robots must have sophisticated artificial intelligence (AI) built in. For example, take Sophia, the creation of Hanson Robotics that Saudi Arabia recently granted citizenship to, in a lavish publicity stunt.
Sophia has the ability to carry on limited conversations, requiring AI-driven natural language processing. ‘She’ can also recognize faces and respond to interactions emotionally, both via tone of voice as well as through a complex set of facial expressions – all technologies that depend upon AI.
In many ways, however, Sophia is less about AI and more about theatrics. ‘Her’ creators have combined AI, robotics, and elements of show business to prove that a fake person can be sufficiently realistic to be appealing rather than repulsive.
Not In Our Digital House
What they have not accomplished – and in fact, aren’t attempting to accomplish – is to create AI that can simulate human behavior in a way that solves true business problems.
Sophia the Robot at Web Summit 2017STEPHEN MCCARTHY/WEB SUMMIT.
What Do We Really Want from Human-Like AI?
In spite of decades of Hollywood influence, the real money in robotics today is in the massive armlike devices from industry. No witty banter, no subtle smiles here – instead, these machines focus on the tasks at hand.
When we talk about AI, however, we’re on shakier ground, as most of today’s applications of AI in business are not particularly humanlike. True, we have voice interactions like Siri, image recognition, and machine learning-driven predictions based upon crunching massive quantities of data – but if you think about it, these capabilities are only vaguely humanlike at best.
On the other hand, Artificial General Intelligence (AGI) – that is, true humanlike reasoning ability – is still well out of reach today. In the absence of AGI, then, what are the most advanced AI capabilities we can currently bring to bear to solve real business problems?
I recently spoke with Donald Thompson, Founder and CTO of AI-driven knowledge platform vendor Maana, to get his take on this question. “Maana supports decision making, reasoning, and answering questions,” Thompson explained. “We construct models that represent human expertise and combine them with data models in the context of optimizing a decision flow or equipment like an oil well.”
Where Sophia is able to mimic human expressions and conversations, Maana’s ability to mimic – or perhaps duplicate – human expertise provides the business value context that Sophia’s theatrics cannot.
The essence of Maana’s technology is what Thompson calls ‘digital knowledge.’ “Digital knowledge consists of models of the domain of business artifacts with the business goal in mind. These models digitize decision flows, and provide recommendations that help experts in the organization make better and faster decisions,” he explained.
The business artifacts in question could represent a wide variety of different types of information, from PDF documents to spreadsheets to emails to sensor data to information bottled up in various applications. From these artifacts, Maana is able to create data models, as well as computational models that mathematically model the human knowledge of a specific business process.
Given the rather philosophical questions about the nature of human knowledge, Maana puts a fine point on this conundrum. “We phrase every knowledge model as an answer to a business question,” Thompson said. “These are higher order questions, for example, ‘given the pump failure in a particular oil well, what are other oil wells with similar pumps?’”
The choice of oil wells in this example is no accident: Maana has found traction in the petroleum industry, listing companies such as Chevron CVX +0.66% Corporation, Royal Dutch Shell , and Saudi Aramco as customers as well as investors.
The reason that the oil business has taken to Maana’s AI-driven decision-making capability is due to the combination of the industry’s massive big data sets with its ability to distil its business goals into easily digested statements.
Thompson clarified the types of goals Maana’s technology targets. “The goals include minimization (for example, risks or costs) or maximization (for example, profits),” he said. “For example, given this list of constraints, what risks should we mitigate?”
In fact, Maana’s ability to interpret constraints gives the technology a leg up on modeling human expertise, as such expertise often boils down to dealing with constraints on a day-to-day basis.
For example, a person may know how to accomplish a goal, but also must know about all the various exceptions that might occur and how to deal with them. Such exceptions are a type of constraint.
Knowledge modeling in order to achieve certain goals can be thought of as decision support. “By decision support we mean that the technology can observe a particular domain, reason about the domain, justify and explain its reasoning, and then allow the user to hypothesize about different courses of action.”
The latter point is especially important: the technology does not generally make business decisions itself, but rather supports human decision making – in part by empowering people to simulate the results of different actions they may take.
Such decision support requires a mix of technical capabilities. “It’s a combination of natural language processing, constraint satisfaction, and simulation,” Thompson explained. “Maana adds constraint satisfaction and Bayesian probabilistic inference.”
The result is human-like reasoning, as different from simpler forms of AI like machine learning as the auto assembly robot is from Sophia. “Contrast Maana’s approach with machine learning, which is how to train machines to learn from data,” Thompson said. “Reasoning is the ability to answer a question.”
As with every vendor in the emerging AI marketplace, Maana’s technology will continue to improve over time. However, its ability to model human expertise to answer business questions in order to support mission-critical decisions gives the technology human-like characteristics that provide business value that other technologies cannot, the gimmicks and other theatrics of Sophia notwithstanding.
Intellyx publishes the Agile Digital Transformation Roadmap poster, advises companies on their digital transformation initiatives, and helps vendors communicate their agility stories. As of the time of writing, Maana is an Intellyx customer. None of the other organizations mentioned in this article are Intellyx customers. Image credit: Stephen McCarthy/Web Summit.
Jason Bloomberg is president of industry analyst firm Intellyx.
Jason Bloomberg is a leading IT industry analyst, Forbes contributor, keynote speaker, and globally recognized expert on multiple disruptive trends in enterprise technology and digital transformation. He is founder and president of Agile Digital Transformation analy...
No comments:
Post a Comment