The potential of AI is boundless. Currently, there is a lot of buzz around how it will change industries like transportation, entertainment and healthcare. Less known but even more revolutionary is how AI could change science itself. In a previous blog, I speculated about the impact of AI on academic research through text mining. The implications of automated research described here are even more far-reaching.
Recently, I came upon an article in Aeon that described exactly that. In it, biologist Ahmed Alkhateeb eloquently makes his argument in the excerpt below:
Human minds simply cannot reconstruct highly complex natural phenomena efficiently enough in the age of big data. A modern Baconian method that incorporates reductionist ideas through data-mining, but then analyses this information through inductive computational models, could transform our understanding of the natural world. Such an approach would enable us to generate novel hypotheses that have higher chances of turning out to be true, to test those hypotheses, and to fill gaps in our knowledge.
As a good academic, the author says a lot with a few words in the paragraph above. Let me unpack his statement a bit.
His first point is that in the age of big data, individual human minds are incapable of effectively analyzing, processing and making meaning of all the information available. There was a time where all the knowledge about a discipline was in books that could be read or at least summarized by one person. Furthermore, traditional ways of doing research whether through a lab experimentation, sampling, controlling for externalities, testing hypothesis take a long time and only give a narrow view of reality. Hence, in a time where big data is available, such approach will not be sufficient to harness all the knowledge that could be discovered.
His second point is to suggest a new approach that incorporates Artificial Intelligence through pattern seeking algorithms that can effectively and efficiently mine data. The Baconian method simply means the approach of discovering knowledge through disciplined collection and analysis of observations. He proposes an algorithmic approach that would mine data, come up with hypothesis through computer models then collect new data to test those hypotheses. Furthermore, this process would not be limited to an individual but would draw from the knowledge of a vast scientific community. In short, he proposes including AI in every step of scientific research as a way to improve quality and accuracy. The idea is that an algorithmic approach would produce better hypotheses and also test them more efficiently than humans.
As the author concedes, current algorithms and approaches are not fully adequate for the task. While AI can already mine numeric data well, text mining is more of a recent development. Computers think in numbers so to get them to make sense of text requires time-consuming processes to translate text into numeric values. Relevant to this topic, the Washington Post just put out an article about how computers have now, for the first time beat human performance in a reading and comprehension test. This is an important step if we want to see AI more involved in scientific research and discovery.
How will automated research impact our world?
The promise of AI-assisted scientific discovery is remarkable. It could lead to the cure of diseases, the discovery of new energy sources and unprecedented breakthroughs in technology. Another outcome would be the democratization of scientific research. As research gets automated, it becomes easier for others to do it just like Windows has made the computer accessible to people that do not code.
In spite of all this potential, such development should cause us to pause for reflection. It is impressive how much of our mental capacities are being outsourced to machines. How comfortable are we with this inevitable meshing of bodies and electronics? Who will lead, fund and direct the automated research? Will it lead to enriching corporations or improving quality of life for all? I disagree with the author’s statement that an automated research would make science “limitlessly free.” Even as machines are doing the work, humans are still controlling the direction and scope of the research. As we ship more human activity to machines, ensuring they reflect our ethical standards remains a human mandate.
