The development of the AI-driven models took several months, but once established, they were capable of designing candidates for therapies in a very short time, reportedly within a day. Researchers focused on anti-COVID nanobodies—smaller and less common relatives of antibodies—after determining these molecules would be well-suited for the limited computational resources available. This decision proved effective, enabling the models to efficiently create nanobodies that demonstrated a novel scientific advancement, with most able to bind to the original COVID-19 variant. However, authors Pak and Zou noted that the primary significance of their work lies in the Virtual Lab itself as a tool; Yi Shi, a pharmacologist at the University of Pennsylvania, supports this view, emphasizing the importance of automation in the research process.
The study was accepted by Nature, which expedited its publication due to the emerging interest in the utilization of AI in scientific research. Zou, however, faced challenges in adequately crediting AI’s role in the research process during submission due to policies that restrict AI’s inclusion as coauthors. Many journals, including Nature, have cited concerns related to accountability and copyright as reasons for these restrictions. Zou expressed disappointment with these limitations, suggesting they discourage transparency regarding AI’s involvement.
In response, Zou initiated the Agents4Science conference, which uniquely stipulates that the primary author on all submissions must be an AI. Other AI bots will evaluate these submissions, while a select team of human experts, including a Nobel laureate in economics, will review the top contributions. Although Zou is uncertain about the conference’s outcomes, he anticipates a diverse array of submissions and acknowledged the potential for both significant discoveries and notable errors. While initial feedback has been favorable, some in the scientific community remain skeptical.
Source: https://www.technologyreview.com/2025/08/22/1122304/ai-scientist-research-autonomous-agents/