Applications of AI in Research – Part 2 (An Interview with Lucas Brower)

Recently, The Acronym interviewed Dr. Dong to delve into the applications of AI in scientific research; Dr. Dong provided an insightful perspective on the potential evolution of the ecosystem and offered invaluable advice to young researchers as they try to navigate this rapidly changing world. To complement that interview, we decided to assess the impact from a student’s perspective. Subsequently, we spoke with one of our Math whiz seniors, Lucas Brower, to gather his insights on the applications of AI in research. Similar to Dr. Dong, Lucas generously shared his time and thoughts on the topic.

The interview explored the significant role of artificial intelligence (AI) and computational tools in research, drawing on examples from his summer internship at a mining equipment manufacturer where AI was used to predict temperatures in clutch plates. The discussion highlighted AI’s potential to expedite research processes and replace traditional computational methods with neural networks for faster simulations. It also addressed the application of AI beyond data analysis, including literature reviews and report writing, while acknowledging challenges such as data security and legal concerns regarding AI-generated content. See the interview below to learn more about essential computational skills for researchers, ethical considerations in AI use, and the future evolution of AI in research, particularly in peer review systems. 

Have you used AI or other computational tools in research?  Does it primarily expedite the process, or can it help identify something new that didn’t exist before?

I have a summer internship with a large mining equipment manufacturer where I implemented neural networks within simulations for predicting temperatures of clutch plates within track-type tractors. The main incentive for utilizing neural networks is that they significantly, and primarily, expedite the computation of simulation. AI has significant potential in replacing approximative equations in computation simulation for a significant increase in speed and minimal, if any, decreases in quality and accuracy.

Apart from data analysis, which steps of research should benefit the most from AI (for example in lit review, report writing)? 

If a researcher does not fall into dependency on AI, it can prove to be an extremely useful tool. Literature review, report writing and documentation, and peer review steps are the most prominent, as natural language generation and natural language processing algorithms can assist researchers in drafting, reviewing, and finalizing their reports. 

What are the top concerns with using AI in research?  

The top concern for generative AI is the legal ramifications of using the generated solutions. For example, if ChatGPT generates code for proprietary research, who legally owns that code? This is a major bottleneck for the usage of AI in research, in addition to ensuring the reliability of AI-based results. Ensuring reliability entails rigorous testing, which in some cases could be less time-efficient than not using AI. 

Which computational tools and skills are absolutely critical for any researcher to learn?

Version Control, Data Visualization, and High-Performance Computing. Familiarity with version control systems like Git enables researchers to track changes in their code and collaborate effectively with team members. Being able to create clear and informative data visualizations is essential for communicating research findings to a wider audience. For researchers working with large datasets or computationally intensive tasks, knowledge of HPC techniques and tools can help optimize code performance and accelerate simulations or analyses.

What ethical considerations should young researchers consider while using AI?

Young researchers should always consider inherent bias and unfairness from the data they are trained on. Researchers should carefully evaluate and mitigate biases in their data, algorithms, and models to ensure fairness and equity.

In a similar vein, researchers should strive to make AI systems transparent and understandable, providing explanations for their decisions and disclosing limitations or uncertainties. Transparent reporting of research methods and results promotes accountability and facilitates reviewing the training data for potential biases.

How do you expect AI applications in research to evolve over the next 5-10 years?

I expect AI-powered peer review systems will become more prevalent, automating aspects of the peer review process, such as manuscript screening, quality assessment, and reviewer assignment. AI-driven peer review platforms will streamline the peer review process, potentially improving the quality and reproducibility of scientific research.

Conclusion 

This interview underscores the transformative impact of AI on research methodologies, highlighting both its potential to revolutionize traditional processes and the challenges it presents. As we navigate the complexities of integrating AI into research, it’s clear that the technology not only expedites existing methods but also opens avenues for discovering novel insights. With careful consideration of ethical standards and legal implications, the integration of AI into research promises to enhance efficiency, accuracy, and innovation across various domains, shaping the future of scientific inquiry and exploration.