Usually it is a matter of systematic and patient collection of data, testing hypotheses that consolidate our knowledge in the vicinity of what we already know. We record the effect of altering this or that experimental condition, and gradually the area of scientific terra firma encroaches on the ocean of ignorance: slowly but surely the shoreline extends, creating the branched causeway we regard as Truth.

Structurally, the human brain is a mess. The problem is the way in which it has evolved: the bulk of what fills our skulls – the telencephalon, and especially the cerebral cortex – is relatively new, but it has not displaced the older and simpler structures seen in reptiles. Rather, the brain has evolved through accretion (Sarnat and Netsky 1974) (Figure 5.1): the older areas are still functional – indeed they are much the more important – but they have come to be supplemented and regulated by the newer areas, which provide improved overall integration and prediction through the massive bands of associational fibres that link every part of cerebral cortex to every other. In addition, these newer and ‘higher’ areas send and receive huge nerve tracts that have elbowed the older structures aside and distorted their shapes, making the relationships between their parts hard to discern.

There are fashions in science just as in everything else, and the popularity of measuring the time required to make a response to a stimulus has waxed and waned over the years. The earliest investigators were motivated in part by a practical problem in astronomy, the difficulty of estimating the exact moment at which a star or planet reached the cross-hairs of a transit telescope (Wolf 1865, Mollon and Perkins 1996). This work revealed two striking kinds of variability. On average, different observers had very different reaction times (Henmon and Wells 1914) (the ‘personal equation’); in addition, reaction times varied greatly from one occasion to another. Later, with the rise of quantitative psychophysics in the nineteenth century, reaction time came to be seen as an objective parameter of psychological performance that could be obtained without very advanced recording technology (Boring 1929). But these data proved easier to obtain than to explain, and reaction times fell out of favour.

The goal of this chapter is to explore and review the role of artificial intelligence (AI) in scientific discovery from data. Specifically, we present AI as a useful tool for advancing a No-Boundary Thinking (NBT) approach to bioinformatics and biomedical informatics. NBT is an agnostic methodology for scientific discovery and education that accesses, integrates, and synthesizes data, information, and knowledge from all disciplines to define important problems, leading to innovative and significant questions that can subsequently be addressed by individuals or collaborative teams with diverse expertise. Given this definition, AI is uniquely poised to advance NBT as it has the potential to employ data science for discovery by using information and knowledge from multiple disciplines. We present three recent AI approaches to data analysis that each contribute to a foundation for an NBT research strategy by either incorporating expert knowledge, automating machine learning, or both. We end with a vision for fully automating the discovery process while embracing NBT.