The chemistry community is being urged to ban the use of generative artificial intelligence (GAI) tools to draw chemical structures due to serious concerns about accuracy and reliability. Researchers Audrey Moores and Vânia Zuin Zeidler argue that GAI tools are not yet capable of producing accurate representations of chemistry, which can lead to misinformation and undermine scientific rigor. They claim that these tools can generate structures with incorrect bond lengths, angles, and even missing or extra atoms.
The lack of chemical understanding in GAI models is a major concern. Unlike humans, these models don't truly comprehend chemical principles, leading to flawed representations that can be misleading. This can have serious consequences, particularly in fields like medicinal chemistry where accurate structures are crucial for drug development. The potential for misinformation is also a worry, as inaccurate structures can spread quickly and mislead learners.
In response to these concerns, some publishers are taking steps to ensure the accuracy of chemical structures in scientific publications. For example, Nature doesn't allow GAI-generated images or videos, except for articles specifically about AI. The Royal Society of Chemistry requires authors to confirm that AI tools were trained on licensed datasets and allow commercial reuse. Elsevier also doesn't permit GAI-generated images that support scientific claims.
Given the potential risks, Moores and Zuin Zeidler's call for a ban on using GAI tools to draw chemical structures seems prudent. As the field of chemistry continues to evolve, it's essential to prioritize accuracy and reliability in scientific representations. By doing so, we can ensure that scientific research is built on a foundation of trust and integrity.
