Artificial Intelligence in Scientific Research

Since the launch of Chatgpt nearly two years ago, artificial intelligence made a huge break into the mainstream. Everywhere, companies are turning to AI, integrating it into their automated services and consumer products. Likewise, AI has made inroads in scientific research, as researchers increasingly find innovative ways to integrate artificial intelligence into their work, in fields like biology, chemistry, and engineering.  

Artificial intelligence opens up possibilities for humans to speed up time-consuming processes, leading to more innovation and discovery. Eric Schmidt, former CEO of Google, claims that large language models (LLM) are uniquely suited for scientific research due to their function, predicting the next word in a series of words and generating whole sentences and paragraphs. Science works similarly, with new discoveries being founded on existing knowledge. Therefore, LLMs may be able to predict new discoveries in biology or physics. Schmidt also argues that AI will be able to analyze potential hypotheses and narrow down candidates quicker, allowing for stronger hypotheses to be chosen and tested. 

For example, scientists at MIT and McMaster University used a neural network in their discovery of abaucin, an antibiotic that can target A. baumannii, an antibiotic-resistant bacterium. The neural network analyzed and predicted the effects of around 7,500 molecules, and gave researchers a promising molecule that could act as an effective antibiotic against a challenging, antibiotic-resistant bacterium. In addition, AI tools could promote interdisciplinary communication and collaboration, generating rough drafts of papers and proposals, as well as offering peer reviews alongside their human counterparts. 

Yet, with technology comes potential risks that may overrule the benefits. Although there are obvious drawbacks related to the current limitations of AI, such as hallucinations and errors, it is possible that, in the future, these limitations may be overcome. Even then, artificial intelligence still poses a danger to the current form of academics. According to Lisa Messeri of Yale and M.J. Crockett from Princeton University, the risks the AI introduces involve epistemic dangers, epistemic referring to ways that knowledge is produced. The article assumes that all shortcomings of AI will be fixed and the models will function as their overseers intend. They describe AI as becoming integrated across various disciplines and in every step of the research process, with their human overseers only needing to check in occasionally. While the usage of AI may bring more speed and efficiency to academic research, it may bring a monoculture in which scientific projects are solely based on what the AI is good at. Messeri explains, “If AI becomes this tool that everything passes through, you risk narrowing the kinds of questions asked and limiting the kind of perspectives that are brought to bear on a scientific problem” (Damiani, 2024). Messeri and Crockett advocate for the responsible and cautious use of AI in research and reiterate that they use AI tools in their own work. They simply state that while AI can be a beneficial tool in many aspects, relying on it too much may cause problems. 

In the end, the introduction of artificial intelligence into science has the potential to encourage new experimentation and discoveries but also brings with it the danger of overreliance. Therefore, it is imperative that we understand the possible consequences and outcomes of the use of AI in order to take full advantage of its capabilities while avoiding pitfalls. 

Schmidt, E. (2023, August 31). Eric Schmidt: This is how AI will transform the way science gets done. MIT Technology Review. https://www.technologyreview.com/2023/07/05/1075865/eric-schmidt-ai-will-transform-science/ 

Liu, G., Catacutan, D. B., Rathod, K., Swanson, K., Jin, W., Mohammed, J. C., Chiappino-Pepe, A., Syed, S. A., Fragis, M., Rachwalski, K., Magolan, J., Surette, M. G., Coombes, B. K., Jaakkola, T., Barzilay, R., Collins, J. J., & Stokes, J. M. (2023). Deep learning-guided discovery of an antibiotic targeting Acinetobacter baumannii. Nature Chemical Biology, 19(11), 1342–1350. https://doi.org/10.1038/s41589-023-01349-8 

Damiani, J. (2024, July 2). The Risks of AI in Science, per Princeton, Yale Professors. Forbes. https://www.forbes.com/sites/jessedamiani/2024/05/31/will-ai-change-scientific-research-for-the-better-or-worse/