The complementary contributions of academia and industry to AI research

Lizhen Liang,Han Zhuang,James Zou,Daniel E. Acuna

2024-09-19

Abstract:Artificial intelligence (AI) has seen fast paced development in industry and academia. However, striking recent advances by industry have stunned the field, inviting a fresh perspective on the role of academic research on this progress. Here, we characterize the impact and type of AI produced by both environments over the last 25 years and establish several patterns. We find that articles published by teams consisting exclusively of industry researchers tend to get greater attention, with a higher chance of being highly cited and citation-disruptive, and several times more likely to produce state-of-the-art models. In contrast, we find that exclusively academic teams publish the bulk of AI research and tend to produce higher novelty work, with single papers having several times higher likelihood of being unconventional and atypical. The respective impact-novelty advantages of industry and academia are robust to controls for subfield, team size, seniority, and prestige. We find that academic-industry collaborations produce the most impactful work overall but do not have the novelty level of academic teams. Together, our findings identify the unique and nearly irreplaceable contributions that both academia and industry make toward the progress of AI.

Computers and Society,Artificial Intelligence,Social and Information Networks

What problem does this paper attempt to address?

The paper aims to explore the different contributions and mutual influences of academia and industry in artificial intelligence (AI) research. Specifically, the paper reveals the following key issues by analyzing a large number of AI-related literature and models over the past 25 years:

1. **Impact Comparison**: The paper finds that articles published by purely industrial researchers are more likely to receive high citation counts and are more likely to exhibit citation-disruptive patterns. Additionally, these articles are more likely to propose state-of-the-art models.

2. **Novelty Comparison**: In contrast, purely academic teams have published more AI research papers, and these papers tend to be more novel, meaning that individual papers are more likely to be unconventional and unique.

3. **Collaboration Effect**: Joint research teams from academia and industry, while showing the strongest performance in terms of impact, are less innovative compared to purely academic teams. The paper further explores different types of collaboration models, such as academia-led or industry-led projects, and analyzes the impact of these models on research impact and novelty.

4. **Control Variable Analysis**: To ensure the reliability of the conclusions, the authors also controlled for multiple variables through mixed-effects models, including research subfields, team size, and researcher seniority, to eliminate potential confounding factors. The results show that the above conclusions still hold after controlling for these variables.

In summary, the core objective of the paper is to reveal the complementary advantages of academia and industry in AI research through systematic data analysis and to emphasize the importance of both for AI advancement.