Scientists See Potential In AI, But Hesitate To Embrace It, Survey Finds
A major international survey has revealed that while researchers acknowledge the potential of Artificial Intelligence (AI) in science, many have not yet fully incorporated it into their daily work routines.
According to a recent survey conducted by science publisher Wiley, involving nearly 5,000 researchers from over 70 countries, AI is expected to become a fundamental component of scientific work within the next two years. However, despite over half of the researchers recognizing the efficiency of AI for specific tasks, only 45% of the initial 1,043 respondents reported using AI tools in their research.
Current AI Usage
Among the respondents who do use AI, the dominant tool is ChatGPT by OpenAI, utilized by 81% of researchers. In comparison, tools like Google’s Gemini or Microsoft’s Copilot are known to only approximately a third of the surveyed individuals. AI applications are primarily employed for tasks such as translations, proofreading, and manuscript editing, with writing and text-related activities leading at 57%.
Future Acceptance and Applications
Over half of the participants anticipate a wider acceptance of various AI applications within the scientific community in the next two years. Young scientists especially leverage AI for tasks like drafting funding applications and identifying potential collaboration opportunities. Moreover, there is optimism about the integration of AI in the peer review process, with 51% believing that AI-based reviewer recommendation systems could become standard in the near future.
Barriers to Adoption
The survey highlighted ethical concerns as the primary barriers to AI adoption, with 54% of respondents expressing worry in this area. Other significant concerns included the potential for AI systems to generate inaccurate results or hallucinations (51%), data protection issues (47%), and the lack of transparency in AI tools (46%). Moreover, many researchers feel inadequately supported in using AI effectively, with 43% reporting insufficient training and 42% feeling overwhelmed by the myriad possibilities presented by these tools.
Challenges and Advantages
Safety remains a critical concern among participants, with 81% expressing apprehension regarding AI accuracy, biases, data protection, and the opaque development processes of these systems. Despite these concerns, AI is acknowledged to outperform humans in specific domains such as manuscript preparation, text enhancement, error detection, plagiarism checking, and formatting.
AI Strengths and Limitations
While AI excels in processing vast amounts of information (60%) and facilitating communication of scientific findings (58%), only 18% currently leverage its capability for automatically processing unstructured data. Conversely, human expertise is still favored in areas like creating educational materials, enhancing research methodologies, providing feedback through peer review, and driving collaborations.
It is evident that while AI holds great promise in revolutionizing scientific endeavors, researchers remain cautious about its ethical implications and the need for adequate support and training to harness its full potential.
Source link
