Generative AI and Foundational Models in the Life Sciences: Landscape, Limits, Opportunities and Concerns

Description

The production of synthetic but realistic data (for example, text, images or biomolecular
structures) by generative Artificial Intelligence methods has made extraordinary progress in
recent years. Applications of generative models in biomedical research are numerous, from
assisted writing when composing a manuscript to designing new antibiotics or understanding
single-cell genomics datasets.
While the extent of generative AI's capabilities is not yet fully understood, it is already clear that
the potential of these approaches in accelerating scientific discovery is considerable.
The advances are so rapid that the ability of computer models to outperform humans in specific
tasks is becoming a realistic possibility. At the same time, the capabilities of generative models
are raising concerns about potential misuse, including the production of erroneous data or
misleading results, the development of biological agents of concern, and the loss of human
oversight.
EMBO is organizing a closed workshop that will bring together key stakeholders, including
biologists, AI developers, funders, experts in the governance of emerging technologies,
publishers, and decision-makers. The discussions will focus on generative AI in the context of
biomedical basic research. Applications in clinical sciences and medicine are beyond the scope
of this workshop. In the area of science communication, the focus will be on scientific
information access and retrieval, while other issues such as authorship and copyright will not be
addressed.
The aims of this workshop include the following:
1. Identify the areas of the life sciences where AI is used.
2. Highlight the opportunities that arise from developments and applications in generative
AI in the life sciences.
3. Explore the current limits of AI applications and the potential consequences of
overcoming them.
4. Identify categories of potential risks and concerns.
5. Draft recommendations on how individuals, labs, institutions, publishers and funders
may develop policies on the responsible use of generative AI in life science research.
6. Identify points of consensus about the required level of human oversight over
AI-generated data.

Period	7 Mar 2024 → 8 Mar 2024
Held at	EUROPEAN MOLECULAR BIOLOGY ORGANISATION
Degree of Recognition	International