Racing Against Time: Why Scientists May Use AI ‘Research Assembly Lines’

November 15, 2023 By Pete Wilkins

Artificial Intelligence (AI) is rapidly outpacing scientific research due to the latter’s inherent limitations. Addressing this issue is crucial to seize significant opportunities. Failure to do so may result in the United States falling behind scientists from other countries.

America’s scientists and researchers must harness the power of AI. The technology’s ability to collect and synthesize vast information surpasses human capacities. This capability makes AI particularly valuable in specific scientific research domains. For instance, AI excels when faced with the daunting task of identifying new drugs from millions of chemical compounds or discerning planets with potential alien life from countless celestial bodies. AI propels research by analyzing extensive datasets and suggesting promising leads. Moreover, AI combined with robotics can automate experiments, enabling scientists to iterate through experimental conditions at a speed beyond human capacity.

This innovation is critical; the existing scientific research model must keep pace. Even within narrow research domains, new findings are published at a speed exceeding the limits of human absorption. Traditional discovery timelines often fail to address urgent needs, as evidenced by the 12-year average for a new drug to be discovered, clinically tested, and approved. Individuals whose lives depend on breakthrough medications cannot afford prolonged waiting periods.

Reshaping the Future of Scientific Research

AI has the potential to revolutionize scientific discovery. What is this transformative revolution, and are scientists equipped to embrace it? To delve into these questions, I interviewed Dr. Jing Liu, a distinguished Ph.D. with extensive expertise. Despite her busy schedule preparing for the U-M Annual Data Science & AI Summit scheduled for November 13-14, 2023, she generously shared her perspective, shedding light on the promising future of AI in scientific advancements.

Liu is the executive director of the Michigan Institute for Data Science (MIDAS) at the University of Michigan and co-director of the Eric and Wendy Schmidt AI in Science Postdoctoral Fellowship Program. MIDAS plays a crucial role in propelling data science and AI research across diverse disciplines. It enables the development of cutting-edge methodologies and facilitates their integration into various research fields such as physical sciences, biological sciences, engineering, earth and environmental sciences, social sciences, and humanities and the arts.

Notably one of the pioneering institutions in this field since its establishment in 2015, MIDAS also stands as one of the largest data science and AI institutes, proudly boasting a distinguished roster of approximately 530 esteemed faculty affiliates.

Liu’s valuable perspective stems from her strong commitment to advocating for scientists and empowering them to achieve groundbreaking discoveries across diverse domains and disciplines. Through tapping the power of AI, she is among those leading the charge to revolutionize the field of science. Her work is paving the way for embracing new technologies and the future of scientific advancements.

Building the Research Assembly Line

Liu asserts that AI has the potential to reshape the research landscape. According to her, “By establishing efficient ‘research assembly lines,’ scientists can liberate their time to think creatively about new research questions that have so far been impossible while AI handles the entire process of designing and executing experiments with unparalleled speed and precision. This development holds great promise for the future of research.”

She explained “the ‘implementation gap’: organizations and scientific research groups must determine how an AI system aligns with their goal, integrate it into their work, assess and validate the system’s output, and be prepared to mitigate any potential issues that may arise. Many companies and academic institutions are still grappling with this process, which mirrors most researchers’ early stages of AI adoption.”

MIDAS is a founding member of the nationwide Academic Data Science Alliance and the Midwest Big Data Innovation Hub. The institute collaborates with academic institutions, industry, government, and community organizations to advance the use of data science and AI research and support data-intensive and AI-enabled decision-making.

As Liu contributes to collaboration across the MIDAS network and affiliated organizations, the prevalence of the implementation gap becomes increasingly apparent. Most researchers function within a small research group with limited resources and a narrow range of expertise, which makes it difficult for each of them to develop their own AI strategy. Liu finds herself wondering: How can we accelerate and expand the trajectory of scientific research? She asserts that the key lies in scaling. This viewpoint finds resonance among science and business leaders, including in the insightful discourse held during a recent Schmidt Futures convening, where co-founder Eric Schmidt presented his visionary perspective on enabling AI-driven research by creating scalable platforms.

Liu went on to elaborate that there are different approaches to scaling. “While one approach is to allocate additional financial resources to more scientists so that more of them can independently develop advanced tools and skills to use such tools, I firmly believe that a fundamental transformation is required. We should aspire to establish scientific ‘assembly lines,’ empowering researchers to produce knowledge at a larger scale and tackle societal issues. Universities and research institutes, such as MIDAS, play a crucial role in driving this institutional shift for the betterment of our society.”

“Take, for instance, the University of Michigan,” Liu continued. “Here, we have a remarkable cohort of researchers in the AI Lab within the College of Engineering—a team that boasts a number of exceptional AI experts. Furthermore, as one of the world’s largest research universities, expertise extends across diverse fields, presenting extensive opportunities to leverage robust AI systems for accelerated research. Encouragingly, an AI-enabled research infrastructure is on the horizon. University of Michigan IT, in collaboration with Microsoft, developed a fine-tuned LLM model for the university community, safeguarding their data and offering tailored customization options. With the capacity to construct an assembly line and an apparent demand for it, what remains critical is the presence of individuals who can actualize this vision.”

In this context, institutions like MIDAS are ideally positioned as hubs to assist scientists in embracing AI in their research endeavors. Just as universities furnish researchers with internet access, computing resources, and experimental equipment, it is also imperative that universities supply them with AI tools to accelerate research and enable inquiry into previously impossible questions. Rather than leaving scientists to navigate the implementation of AI technologies independently, MIDAS has already begun extending its support. Recently, it has organized workshops to demonstrate the utilization of Generative AI in research, alongside developing comprehensive guidelines for its appropriate application. Liu’s optimistic outlook harbors the hope that we will swiftly advance to creating an array of AI tools catering to diverse researchers. Though each project operates with unique data and yields distinctive insights, many facets of the research workflow are shared, enabling researchers to select and personalize their assembly lines, heralding a new era of knowledge production.

Collaboration and Innovation

I asked Liu about the allies involved in creating a research assembly line. Her response emphasized the significance of collaborative efforts when incorporating new and powerful technologies like AI. She explained, “Obviously, scientists who develop new AI concepts and systems—for example, the wonderful AI experts in the University of Michigan AI Lab—are our allies because we can’t develop AI tools for scientists without their work. Additionally, we find allies among the scientists who will utilize the AI tools we develop. Their input and needs are at the core of our mission. We should also learn from like-minded organizations, such as Argonne National Lab, that contribute to advancing scientific experimentation by spearheading the development of ‘self-driving labs.’ We also recognize the importance of collaboration with industry to achieve these goals. By leveraging their expertise and tools, we can further harness the potential of scientific research. In this pursuit, MIDAS is actively exploring partnerships with notable companies like Microsoft and AWS to augment the power of AI in enabling significant scientific discoveries. By fostering collaboration, embracing knowledge exchange, and engaging industry, we strive to unlock new frontiers in scientific research.”

Liu emphasized the importance of ensuring the trustworthiness of scientific discoveries, highlighting MIDAS’ dedication to guaranteeing rigor and reproducibility in scientific research. By selecting the best methods for experimentation and facilitating consistent results through replication, scientists can earn the trust of their peers. Recognizing the significance of AI-enabled research, MIDAS is actively developing initiatives to support rigor and reproducibility in this field. Collaboration between AI developers, industry and academic researchers, AI users, and AI regulators is imperative to this endeavor. MIDAS has established a fruitful partnership with the Ethical AI team at Rocket Companies, fostering reciprocal learning and technical approaches.

Essential stakeholder collaboration must be viewed in a broader context as multiple arms races unfold. Nationally and internationally, there is a race to develop increasingly powerful AI systems while implementing safeguards to determine access rights. At more local levels, tension exists between harnessing the benefits of AI for individuals and organizations while safeguarding against potential negative consequences. Furthermore, organizations face dilemmas such as balancing the desire to acquire the most powerful AI systems to gain an advantage in business or research with the responsibility of validating and justifying the AI system’s output. Thoughtful AI evangelists aim to establish an effective and collaborative model involving government, industry, academia, and the community to ensure that AI technology promptly aligns with society’s fundamental values.

The future of scientific research is poised for a revolution driven by AI. Through collaboration, embracing knowledge exchange, and engaging industry, scientists strive to unlock new frontiers in scientific research. The optimism and momentum surrounding AI in scientific research are palpable, and the future holds immense potential for breakthrough discoveries and innovations. As we navigate this exciting journey, the role of AI in propelling scientific research to the next level is undeniable. Together, we can harness this power and create a future where AI and scientific research move forward hand in hand.

Originally featured in Forbes.