Will Artificial Intelligence Replace Bench and Computer Scientists?

One popular topic buzzing around laboratory corridors is the debate on whether artificial intelligence (AI) could replace hands-on bench work. While AI has proven its worth in various domains like economics, customer service, and climate science, it makes you think – what are the real limits to these technologies? To what extent will human work and expertise remain indispensable, especially in places like laboratories or hospitals?

Here I discuss how AI is pushing forward ongoing research and creating new fields of study. I also explore if research jobs are at risk of being replaced by these automated technologies, and I delve into predictions as to where AI is expected to make significant advancements in 2024.

AI for the benefit of humanity:

We do not need to look too far to witness how AI is making a difference in health sciences and research. In March 2023, Weill Cornell Medicine unveiled the Institute of Artificial Intelligence for Digital Health (AIDH). This initiative aims to enhance patient care, drive discoveries, and improve teaching by integrating AI into healthcare practices. Likewise, the Englander Institute for Precision Medicine (EIPM) launched the AI-Extended Reality (AI-XR) laboratory to bridge augmented, virtual, and mixed reality with AI. Through this effort, scientists could visualize and interact with collaborators and the data in real-time just as if they were in-person. 

Unlike traditional AI, which analyzes information and makes predictions based on predefined instructions and structured data, generative AI (Gen-AI) has the unique ability to transform the same data into entirely new outputs, such as more human-like creation of content through text or images (e.g. ChatGPT or DALL-E). Gen-AI can help alleviate the workload of medical practitioners by assisting with clinical documentation as well as by aiding radiologists and pathologists in efficiently navigating through large sets of results, facilitating the identification of patterns within the data, and ultimately producing diagnoses. 

One example of applied Gen-AI is Augmedix. Produced in collaboration with Google, this technology captures the natural conversation between a physician and patient, transforming it into accurate and comprehensive medical notes. Physicians can then review and transfer these notes in real time to the hospital’s electronic health records, ultimately saving time, reducing burnout in clinicians, and enhancing overall patient care. 

Another application commonly used by wet-lab scientists to aid in experimental design is BenchSci, which uses AI to screen literature for published antibodies testing different experimental variables. Recently, BenchSci launched the ASCEND platform in collaboration with Google to produce knowledge graphs pulling results from an extremely large number of experiments. These graphs enable scientists to depict and understand complex connections in biological systems such as biomarkers, detailed biological pathways, and interconnections among diseases. 

These examples illustrate how advancements in AI are playing a pivotal role in driving significant improvements in the fields of science and medicine. But to what extent will humans remain indispensable? Can AI replace the workforce in laboratories and hospitals?

Domain of knowledge:

“If AI wants to make all my buffers and do mammalian and parasite cell culture, be my guest” says one Reddit user when asked if scientists should be worried about being replaced by AI.

There is concern that AI is going to put people out of jobs. “If you’re still using your hands, you won’t be doing science,” said Max Hodak, the co-founder of the biotech company Transcriptic in an interview with Science. “But the brain of the biologist won’t be replaced anytime soon, simply because the natural world is so complex.” 

Other researchers agree: Domain knowledge, critical thinking skills, and human creativity are key to scientific research and cannot be replaced with AI. “You can’t just blindly swing the latest computational method at a problem, out of the box. It doesn’t work. You have to model the problem based on the right assumptions. And for that, biological expertise is indispensable,” said Dr. Dana Per’er, chair of the computational and systems biology program at Memorial Sloan Kettering.

AI cannot contextualize and interpret data as well as a human, but it can serve to streamline the research process and free up more time for critical thinking and decision making. “Individuals with data science expertise will have more time to understand and implement other strategic decisions, as AI improves efficiency and reduces errors by minimizing human intervention,” says Prashant Mishra, a finance and technology expert5. Repetitive tasks, like pipetting or DNA extractions, are already automated by technologies developed by companies such as ThermoFisher or Opentrons. But these processes can be complemented with AI to uncover patterns within large datasets or provide advanced analysis. AI can also be used to predict  genome-wide variants, functions of cis-regulatory elements, or the 3D arrangement of DNA. “These tools can be used to enhance productivity, but with expert oversight,” recommends Dr. Ulysses Balis, Professor of Pathology Informatics and Associate Chief Medical Information Officer at the University of Michigan, in an interview with the Critical Values magazine. “[AI models] are good at recognizing patterns that have already been seen, but in terms of carrying out the scientific method of hypothesis generation and further investigation to come to a real answer, we’re not there yet.” 

Based on this, we can expect that AI will augment and accelerate the rate of discovery without replacing human researchers. But what are the areas where we expect to see significant AI advancements in 2024?

Predictions for 2024:

According to Google, 2024 will be the year of optimizing administrative work in healthcare. AI technologies are already moving from trials to real world applications in administrative work to assist clinicians. AI is also expected to continue to advance personalized and precision medicine and gene therapies. In many fields of medicine, AI will be used to analyze patient data, improve surgical precision, and enhance post-operative monitoring. 

Generative models are already being used for habitat and species conservation. For example, scientists are using AI to track wildlife populations and understand social dynamics. In addition, AI is being used  to integrate different data types such as sequence data, imaging, and metadata. An example of this is the recent development of facial recognition models created to discriminate between geese with the hopes to study migration patterns: “Birdwatchers will someday be able to snap a picture of a goose, ID it, and share its location with scientists,” says Sonia Kleindorfer, director of the Konrad Lorenz Research Center for Behavior and Cognition in Vienna, Austria. 

Likewise, Krista Ingram, a biologist at Colgate University in New York, developed the AI tool SealNet to identify individual harbor seals. Prior to this technology, “the only way to identify individual seals was by tagging them, but that was difficult.” Now with SealNet, scientists just need a photo to ID seals with high accuracy, making it faster, easier, cheaper, and less stressful for the seals.

New language models like GPT-5 by OpenAI and Gemini by Google can further enable enhanced data curation by simultaneously filtering and reviewing thousands of research articles. AlphaFold by Google DeepMind is also anticipated to release their newest version in 2024, which is expected to more accurately predict structures for proteins, nucleic acids, small molecules, ions, and modified residues.

I believe that AI should be seen as a powerful tool capable of enhancing work, complementing and empowering research, and sparking new hypotheses and scientific discoveries. However, machines lack a fundamental quality that defines us: human ingenuity.

 

 

References