When you think of automation, images of sentient robots performing human tasks may come to mind. While fictional androids such as R. Daneel Olivaw, a crime-solving partner created by science fiction writer Isaac Asimov, are not yet reality, automation is beginning to replace humans for some kinds of labor. Driverless vehicles are starting to take to the roads, and robotic arms have long stood on the factory assembly line. More recently, these arms have been programmed to cook with the finesse of master chefs.
When connected to artificial intelligence (AI), automation can assess and act on vast amounts of data, whether it’s planning a complex science experiment or simply setting the thermostat. With robots and algorithms influencing how we conduct our daily lives, we have truly stepped into the age of automation.
It is estimated that automation will replace 800 million jobs worldwide by 2030. Although technology is expected to disrupt occupations in all industries, jobs that are especially susceptible to being replaced are typically labour intensive and repetitive. On the other hand, jobs which involve significant amounts of human interaction as well as creativity are the least likely to be made obsolete.
A report by the Brookfield Institute for Innovation + Entrepreneurship at Ryerson University estimated the probability of various occupations going extinct as a result of new developments in AI and robotics. Assemblers, inspectors, and attendants were predicted to be almost certainly replaced by automation in the next 10 to 20 years. Biologists and related scientists were predicted to have a 15.6% chance of being replaced by automation. While the report considers this a low risk, it is still higher than that for managers, artists, and educators, who have a smaller than 5% chance of being replaced by automation.
Automating the life sciences
At first glance, these statistics may seem counterintuitive. Why would the risk for biologists – whose complex work requires intense intellectual input – be higher than that for other occupations that at first glance seem to have more components that can be automated (think calendaring for the high-level manager)?
It may be easy to forget that many tasks performed by scientists working in the biological sciences can be automated – and actually need to be automated to free up scientists’ time to do the hard thinking. Pipetting liquids, making serial dilutions, counting the number of bacterial colonies on plates – menial tasks once part of the life scientist’s repertoire – are now automated. Automated colony counters like those produced by BioLogics Inc. are saving thousands of man-hours in laboratories across the world. Liquid handlers such as the Labcyte’s Echo 555 and Opentrons’ OT-2 robot negate the need for hours of tedious manual pipetting while enabling high throughput screening.
Experimental measurements can also be optimized and automated using the Internet of Things (IoT). Internet-enabled sensors can be integrated with, or connected to, almost any piece of equipment to collect data, which is then stored in the cloud. Likewise, in the field, drones and unmanned aerial vehicles (UAVs) provide low cost high resolution imagery for tracking species occurrence, thus reducing the need for field biologists to trudge into inaccessible habitats to count or collect samples.
Algorithms have also entered biological workflows, heralding the fields of bioinformatics and computational biology. Sequence alignment software programs such as Molecular Evolutionary Genetics Analysis (MEGA) enable scientists to align multiple DNA or amino acid sequences to identify conserved regions – which are then used to construct phylogenetic trees showing evolutionary relationships between organisms. Protein structure visualization software such as RasMol generate 3D protein structures from amino acid sequences.
Just as Illumina DNA sequencers catapulted biologists’ ability to read DNA, lab automation plus AI will accelerate the biological design-build-test cycle in unimaginable ways. Image courtesy of Magnus Manske on Wikimedia.
In the pharmaceutical sector, the application of AI to predict molecule behaviour and drug target suitability has accelerated drug discovery. Machine learning algorithms have been used by companies such as ATUM to screen thousands of candidate genes and molecules rapidly. And companies like Zymergen and Codexis are using platforms that integrate machine learning, genomics, and chemistry to identify chemical building blocks and create novel proteins for materials and products. Algorithms are also required to analyze the large data sets generated by ‘-omics’ technologies which enable biological systems to be investigated on an unprecedented scale.
Automation in data analysis
With automation and algorithms facilitating high-throughput, increasingly parallelized laboratory experiments, biologists are inundated with mountains of data. It can be nearly impossible as well as a huge time sink to pour through data searching for the meaningful result. Automated workflow systems, like Riffyn’s cloud-based Scientific Development Environment (SDE), use machine learning to identify noise and errors in biological data sets, allowing biologists to identify true associations and correlations.
Workflow systems also improve reproducibility and standardization. By including the parameters, inputs and source codes in the platform, workflows allow users to keep track of the methods which are used to analyze the data, ensuring more accurate conclusions and facilitating reproducibility.
Scientists and automation are inextricably linked
Already, the biologists’ job is shifting from data collection to data analysis. This begs the question – are scientists who are primarily trained in biology still needed? Biology is an immensely complex and rapidly developing science encompassing signalling networks, development, wiring of the nervous system, ecology, and many others. Until AI – which has yet to surpass human intelligence in the ability to think creatively and critically – can draw connections between these vast areas of information, scientists’ knowledge will be critical to interpret data and understand the scientific, societal, and economic implications of their findings. Algorithms are only as good as their input data, and with many details of biological components and systems remaining a mystery, conclusions drawn from models built by algorithms with existing data may not be entirely accurate. This can yield surprising and unexpected results when experiments are done to verify these predictions.
The future of biological researchers stands only to benefit from automation. Far from being displaced by automation, scientists will have greater tools and time to tackle our most complex questions in biology. Technology will light the path to solving ever more complex biological challenges, and chances are biologists will be there to illuminate the dim frontiers of understanding of the natural world.5