Global accord on the integration of artificial intelligence in medical science publishing: implications of the Bletchley Declaration
Cardiovascular Research · 30.12.2023 · 119 (17) · 2681-2682
DOI: 10.1093/cvr/cvad170 ↗Kurz gefasst
Editorial in der Zeitschrift Cardiovascular Research, online erschienen am 23.11.2023. Verfasst von Tomasz J Guzik und 1 weiteren Autor:innen. Beteiligte Einrichtung: Centre for Cardiovascular Sciences, The Queens Medical Research Institute, University of Edinburgh, 47 Little France Crescent, EH16 4TJ Edinburgh, UK. Hauptthemen laut MEDLINE: Artificial Intelligence, Publishing. Die Arbeit wurde bisher 6-mal zitiert (Crossref) und nennt 11 Literaturangaben.
Abstract (Originalsprache)
This manuscript was handled by Deputy Editor, Charalambos Antoniades. Paying homage to the ground-breaking work of World War II’s cryptanalysts, Britain coordinated the ‘Bletchley Declaration’, a pivotal document designed to strengthen global collaboration on artificial intelligence (AI) safety.1 This declaration, endorsed by 28 countries and the European Union, was revealed as a culprit of the AI Safety Summit held at Bletchley Park, the very site where Alan Turing and his contemporaries pioneered the field of modern computing. Just as for the current AI Safety Summit, international cooperation was at the heart of Turing’s success, which included breaking the Enigma code by developing ground-breaking contributions of less visible, but hugely influential, polish mathematicians from Poznan—Rejewski, Różycki, and Zygalski.2 The venue stands as a monument to the ingenuity of mathematical breakthroughs that have continually propelled advancements in various scientific fields, including modern medicine. The Bletchley Declaration itself is a testament to the enduring importance of collaborative scientific endeavour, seeking to harness the full potential of AI while navigating its risks with international unity and foresight.1 The integration of AI into scientific publishing is not only imminent but is already in motion, putting medical publishing at a crucial crossroads. Concerns voiced during the summit included risks of AI tools replacing human contributions in the workforce as faster, more accurate, and efficient. Is that the future of science? At least at present, however, the flexibility, creativity, and uniqueness of human mind make us believe that human-driven scientific discovery, while hugely aided by AI tools, will continue to be at the centre of medical progress. The growth of AI tools, however, can greatly revolutionize medical science publishing. It brings promise to enhance data analysis making us quickly uncover multiple layers of complexity in medical data, accelerate research dissemination, improve content curation, and optimize accessibility.3 We experienced the efficacy of such approaches during an unprecedented discovery path to develop COVID-19 vaccines and understand health impacts of this pandemic.4 AI tools for in-depth and advanced discovery of biomarkers but also novel mechanisms of disease, may provide, at least in part, an alternative to the use of animal models and other complex biological systems.5–8 However, it becomes imperative to navigate the challenges that come with these great promises responsibly. To harness AI’s potential ethically and effectively, we must commit to principles that prioritize safety, human-centric design, trustworthiness, and responsible sharing of the outcomes. Our collective aim should be to enable utilization of AI in ways that bolster discovery, driving sustainable development and innovation. These overarching aims were clearly emphasized in the Bletchley Declaration1 and align perfectly with aims of medical research. At the same time, the ‘AI Safety Summit’ emphasized that the proliferation of AI systems in everyday life underscores the urgency of our commitment to safe AI development.1 This is especially relevant in areas that directly affect human health and wellbeing. One of the most important risks, particularly in medical publishing where accuracy is paramount, is potential for misinformation or inaccuracy that can have dire consequences.9 The call for international collaboration to address these challenges is clear in the Bletchley Declaration. We must work towards the development of transparent AI systems specifically designed for medical research and publishing purposes, whose processes and decisions can be easily explained and understood, ensuring fairness, alignment with ethical standards, and accountability. The advent of powerful general-purpose AI models termed by the Declaration as ‘frontier AI’, capable of performing a wide array of tasks, poses additional safety risks.1 The medical publishing sector must be vigilant against the misuse of AI, intentional or otherwise, and strive for alignment with human intent. Awareness of potential utility of these tools by already existing ‘paper mills’ and other groups unethically misusing available data is of paramount importance. This includes potential for AI to inadvertently contribute to issues like disinformation within the medical domain—a risk we cannot overlook. Given the inherently international nature of medical science publishing and the risks posed by AI, it is essential to commit to global cooperation to promote safety and trust in AI applications. While every journal is responsible for ensuring transparency and accuracy of their content, in the case of AI-related policies, a flexible cooperative approach, allowing for innovation while considering the risks, is necessary. It should foster common principles and codes of conduct that are sensitive to the varying legal and regulatory landscapes across different countries. Following valuable insights from the AI Safety Summit, we can clearly see that ensuring AI’s safety involves multiple stakeholders—governments, companies, and civil society—that all need to work closely with academia and AI industry. This may be particularly challenging when we consider inequalities in healthcare provision across the globe. Gaps in digitization are similarly huge between nations, making some areas of the world potentially particularly vulnerable.1 The list of absentees in the AI Safety Summit, as well as the Bletchley Declaration, is very visible and should be remedied by future proceedings. The digital divide must be bridged, and international collaboration should include a broad spectrum of partners, with policies that empower developing countries to build AI capacity.1 Frontier AI developers bear a particular responsibility for the safety of their systems. One of the most important practical conclusions of the AI Safety Summit was a call for rigorous safety testing, evaluations, and transparency in their operations, especially as they relate to potentially harmful effects. While we believe that multiple layers of safety approvals in medical sciences can ensure enforcement of this, difficulty in understanding complexity and details of some of the systems by the regulators poses challenges. The classic tools we use to bring medications and medical devices to clinical practice, including randomized clinical trials and layers of approval by regulatory authorities, may require adjustments and developments.1 Based on the Bletchley Declaration, one can propose some key general areas of development (Table 1). To achieve this, an internationally inclusive network for frontier AI safety research is needed. Such network can complement existing collaborations and serve as a resource for policymakers and the public. Key areas identified as focus points for safe and efficient use of AI in medical publishing based on the Bletchley Declaration1 Identifying AI safety risks pertinent to medical publishing. Establishing a shared, evidence-based understanding of these risks. Building international, risk-based policies to ensure safety. Fostering transparency in the development of ‘frontier AI’ capabilities. Developing tools for safety testing and relevant public sector capabilities. Identifying AI safety risks pertinent to medical publishing. Establishing a shared, evidence-based understanding of these risks. Building international, risk-based policies to ensure safety. Fostering transparency in the development of ‘frontier AI’ capabilities. Developing tools for safety testing and relevant public sector capabilities. Key areas identified as focus points for safe and efficient use of AI in medical publishing based on the Bletchley Declaration1 Identifying AI safety risks pertinent to medical publishing. Establishing a shared, evidence-based understanding of these risks. Building international, risk-based policies to ensure safety. Fostering transparency in the development of ‘frontier AI’ capabilities. Developing tools for safety testing and relevant public sector capabilities. Identifying AI safety risks pertinent to medical publishing. Establishing a shared, evidence-based understanding of these risks. Building international, risk-based policies to ensure safety. Fostering transparency in the development of ‘frontier AI’ capabilities. Developing tools for safety testing and relevant public sector capabilities. In conclusion, AI’s transformative potential in medical science publishing is unquestionable, and we need to commit to a continuous global scientific discussion and research into AI safety.10 This way, our society will be able to benefit from AI’s strengths for the advancement of medical science and the improvement of global health outcomes, allowing us to develop better personalized approaches in conditions causing most global deaths such as coronary artery disease, heart failure, or hypertension.11 As authors, reviewers, and editors, we must all commit to these efforts to ensure responsible growth of this potential.
Schlagwörter
MeSH-Schlagwörter (MEDLINE)
Autor:innen und Institutionen
- Centre for Cardiovascular Sciences, The Queens Medical Research Institute, University of Edinburgh, 47 Little France Crescent, EH16 4TJ Edinburgh, UK.
- Department of Medicine and Omicron Medical Genomics Laboratory, Jagiellonian University, Collegium Medicum, Kraków, Poland.
- Massachussets General Hospital, Harvard Medical School, Boston, USA.
ORCID-Profile: Tomasz J Guzik
Zitieren
APA
Guzik, T. J., & Sitek, A. (2023). Global accord on the integration of artificial intelligence in medical science publishing: implications of the Bletchley Declaration. Cardiovascular Research, 119(17), 2681-2682. https://doi.org/10.1093/cvr/cvad170
Vancouver
Guzik TJ, Sitek A. Global accord on the integration of artificial intelligence in medical science publishing: implications of the Bletchley Declaration. Cardiovascular Research. 2023;119(17):2681-2682. doi:10.1093/cvr/cvad170
BibTeX
@article{Guzik2023,
author = {Guzik, Tomasz J and Sitek, Arkadiusz},
title = {Global accord on the integration of artificial intelligence in medical science publishing: implications of the Bletchley Declaration},
journal = {Cardiovascular Research},
year = {2023},
volume = {119},
number = {17},
pages = {2681--2682},
doi = {10.1093/cvr/cvad170}
}
EUVIMED