Artificial Intelligence from a Technology and Innovation-Management Perspective
Artificial Inteligence (AI) is a very broad topic with a multitude of different perspectives, living in the intersection of mathematics (esp. statistics and logic), computer science, engineering (esp. robotics), economics, psychology (esp. cognitive science), linguistics, biology (esp. neuroscience) and philosophy. Depending how you define AI, its history goes back either to very recent research, starting around 1956 with a “Summer Research Project on Artificial Intelligence”, or to the great ancient Greek philosopher Aristotle himself. Thus, an extensive view on AI is naturally impossible within a few pages. Here, AI is introduced on a high level with emphasis on two point of views: technological and innovation-theoretical.
In its very core, AI “attempts not just to understand but also to build intelligent entities” (Russell et al., 2010). To be able to do so, Minsky (1961) stated five different subdomains of problem solving that science needs to solve. For a given problem instance, a machine or entity needs to use (1) Search techniques to traverse the solution space, (2) Pattern-Recognition for improving time and space efficiency, (3) Learning for generalizing from past experience, (4) Planning for allocating scarce resources like computational time and finally, (5) Induction for gaining knowledge out of all findings. But there are still plenty of dimensions in which one can distinguish different types of AI or respective fields of AI research. These include:
Should this entity be rational, or human-like (so per se irrational to some degree)?
Is it enough to be able to simulate thinking for decision-making or should it also act autonomously in a certain environment?
Is it able to fulfill only one task very well (“applied” or “narrow” AI), or any given task (“full” AI or “Artificial General Intelligence”)?
And finally, does the AI truly possess a conscious mind or just pretend to have one (“strong” vs. “weak” AI)?
For most of the business applications, the type of AI is weak, applied, mostly rational and thinking-orientated. Particularly, one sub-area of AI, namely Machine Learning (ML), is the main driver of the core problems of Pattern-Recognition and Learning.
As an innovation, AI and ML is still far from being widely adopted in business. On the one hand, one can argue with the general diffusion pattern of new technology, which often follows an S-curve. This reflects the fact that diffusion rates of any innovation usually first rise and then fall over time. The popular “Epidemic Model” explains this behavior by arguing that the limiting factor of technology diffusion is the “lack of information available about the new technology, how to use it and what it does”. This might be true to a certain degree for AI, since even today society and businesses are struggling to understand AI. On the other hand, AI history is specifically characterized by a series of hypes, quickly followed by phases of disenchantment between 1970s until the mid-2000s, so-called “AI winters” (Bengio, 2016). Bengio argues that missing computational power, a lack of data and insufficiently sophisticated algorithms yielded poor practical results, which is the reason for the decline in AI research.
Nevertheless, since 2005, AI got an accelerating wave of interest. This hype is mainly driven by one special ML technique, called Deep Learning (DL). DL is based on a relatively old technique, known as Artificial Neural Network (ANN), which is in turn inspired by the architecture of the human brain. The hype around DL rests upon multiple success stories, where the algorithms archived human or even super-human level performances, especially in the field of computer vision (e.g. for autonomous driving, face recognition) natural language processing (e.g. for chatbots and virtual assistants), or board games (e.g. Go, chess). Cockburn et al. (2017) state that DL will have a large impact on the whole economy, because it represents an Invention of a Method of Inventing (IMI) as well as a General Purpose Technology (GPT) at the same time. IMIs “enable a new approach to innovation itself, by altering the ‘playbook’ for innovation in the domains where the new tool is applied”. They argue that DL is indeed an IMI since it may be able to substantially "automate discovery" and "expand the playbook" in the sense of opening up the set of problems that can be feasibly addressed. Furthermore, DL is also a GPT. This means that DL can be characterized by the potential for pervasive use in a wide range of sectors and the role of an "enabling technology", opening up new opportunities rather than offering complete, final solutions. Ultimately, since AI is a GPT and a IMI at the same time, it is arguably one of the most important innovation enabler of our time.
Keesiu Wong is Co-Founder & CEO of Design AI, a start-up focusing on agile AI development and use case identification through Design Thinking. He is a trained Data Scientist with an academic background in Mathematics, Management and Data Engineering at the Technical University of Munich. In addition to 5+ years of experience with AI projects, he has entrepreneurial experience in 4 start-ups, as well as experience in Design Thinking, top management consulting and as a start-up coach at UnternehmerTUM.