Artificial Intelligence (AI) has changed much since I did my PhD in AI about 30 years ago. The goal of AI was to build programmes that could imitate human capabilities in thinking and problem solving. These capabilities included playing games like chess, proving theorems in mathematics, scheduling activities on a factory floor, designing the layout of a circuit, and so on. The programmes were hard to write, especially the ones that could solve a significant problem. The challenge then was two-pronged: how to get the programmes to solve the problem and how to get them to solve the problem in reasonable time.
The huge upsurge of interest in AI in recent years has come from solutions emerging from research labs that address both these challenges. Recent methods of problem solving and learning are able to solve very difficult problems, and recent advances in computing technologies, such as cloud computing and open source programming, have addressed many of the challenges of thirty years ago. Computing power and source code is available for those willing to learn and experiment.
I teach AI to business students at IIM Bangalore with two goals in mind: to communicate to them the excitement of this field, and the path through which successes have been achieved; and to communicate the risks and challenges associated with such a powerful, general purpose technology that has the potential to disrupt entire economies. Future businesses and business leaders have to grasp the essence of AI in order to channel this force for their own and society’s benefit. This is not easy. Even students with a technical background in computing or IT find it difficult to understand AI. Thus, the challenge is to communicate AI in a manner that is digestible.
Curriculum of AI in business schools
AI in the curriculum in business schools has to build upon some traditional ideas, and some cutting edge, possibly contentious ideas. Below are some issues that are important for the curriculum of AI in business schools.
• The basis for teaching AI has to be the principles and concepts of information systems. Students have to see AI as an evolution in information technology that builds on existing information systems within organisations and in society. This is distinct from teaching AI from a computer science and algorithms perspective.
• The fundamental basis of understanding AI algorithms is by understanding knowledge representation and search. The two pillars of AI are how problems and domains have to be structured, also called a representation, and how programs can find answers to problems within these representations, also called search. These fundamental concepts of AI programs can serve as the basis for grasping complex subjects such as planning, cognitive systems, expert systems, and also machine learning.
• Machine learning methods and techniques also follow the principles of representation and search. This manner of presentation is intuitive for business students who are exposed to problem solving techniques in many other courses in the curriculum.
• Students have to clearly understand that practical applications of AI are in narrow domains. AI programs currently display excellent performance in these domains but are severely restricted to ‘transferring’ knowledge. For example, a programme that can find a pattern in a particular time-series data set may not find anything in another data set. Programmes that can classify images of certain objects, say suitcases, may fail to recognise related objects, such as handbags.
• Machine learning and pattern recognition programmes are heavily dependent on very large quantities of data. For high performance levels, say for 95% accuracy on some classification tasks, the data sets required for training are usually massive and difficult to obtain. A key challenge for managers is to ensure that adequate data is available for AI programs, both for training the programs on tasks and also for testing them for accuracy.
• Owing to their narrow application domains, AI programs tend to be brittle. Their performance seems to be high in the domain in which they are trained. When the data changes a little, the performance does not degrade gracefully, many times it simply fails. Many industrial AI products have addressed this brittleness, mainly by ensuring that the data sets are adequate to cover most variations. Managers have to monitor this weakness carefully, as it could lead to severe problems.
• AI programs intervene and disrupt both cognitive and physical tasks. This affects traditional white-collar and blue-collar workers. For example, white-collar work such as making reports from data, writing summaries of reports, translating text, reading and analysing text and images from focussed domains, are all white-collar tasks that can now be accomplished by AI programs. Also, tasks such as moving boxes, replacing items on shelves, cutting vegetables and preparing food, serving tables, and delivering packets, which are typically blue-collar work are being replaced by robots. The challenge for managers is to understand the manner in which such robots can be integrated within businesses without removing productive workers or replacing them. The fear of widespread unemployment resulting from use of AI is palpable in the media. Managers have to create spaces where AI works alongside workers rather than replaces them.
• The field of AI has historically addressed questions about the nature of thought, the processes of comprehension, sense-making and understanding within brains, the mind-brain issue and also about the nature of consciousness. These remain possibly some of the most interesting aspects of AI. Recently, researchers and engineers have begun to construct artificial brains. These devices replicate the entire structure of brain cells and thus, imitate their behaviour. There is considerable speculation on how these designs will result in ‘superintelligent’ machines that will surpass human capabilities. The aspect of superintelligence, consciousness in machines & the role of machines in the future of humanity are exciting topics of discussion in classes. Students understand these are speculative issues, however, their potential to disrupt and overwhelm human affairs is strong.
Teaching AI requires engaging with issues of immense possibilities and also of substantial challenges and problems. These issues have to be grasped from the perspective of meeting the common problems of businesses – competition, market growth, branding, employee management, employee productivity, innovation, and staying relevant for customers. Students have to grasp that AI, as a powerful, general-purpose technology, has the potential to affect and disrupt all these aspects of business.