The intersection of artificial intelligence and the biosciences is rapidly evolving, bringing profound shifts to how modern research is conducted. In a recent live panel discussion at the Bridging AI & Biosciences one-day workshop, recorded for the Artificially Ever After podcast, experts across computer science and biology gathered to discuss the practical realities, collaborative hurdles, and future implications of this interdisciplinary frontier. Whether AI proves to be a permanent scientific fixture or just another technology hype cycle, its current applications are already unlocking capabilities that were previously impossible.
The panel was chaired by Kieren Sharma, PhD student and presenter of the Artificially Ever After podcast. The panellists were:
- Mark Basham – Science Director for Artificial Intelligence, Rosalind Franklin Institute
- Elisa Pedone – Stem Cell Operation Lead, CellVoyant
- James Thomas – Senior Data Scientist, Jean Golding Institute (University of Bristol)
- Andrew Paige – Cohost, Micro Binfie Podcast; CTO, Origin Science
You can listen to the recorded session above, or read on for some key points from the discussion.
- The consensus among the panelists is that while specific AI trends may fade, the foundational capabilities are here to stay.
- A shift in scale: Neural networks and deep learning have fundamentally transformed image processing since the mid-2010s. Researchers can now scale their analyses from single images to massive datasets
- Patterns, not explanations: Machine learning excels at identifying complex patterns in biology. It is less capable of explaining the underlying biological mechanisms behind those patterns
- Democratising Database Access: Biologists who lack SQL programming skills can use LLMs to ask plain English questions of complex data repositories, seamlessly translating those queries to retrieve vital information.
- Productivity Multipliers: Using AI agents like Copilot or Claude can make experienced coders five to ten times more productive. However, without a strong foundational understanding of computer science, relying on these tools can cause significant problems.
- Automated Research Pipelines: AI can generate publication-ready outputs with minimal effort. This is dramatic shift in the role of the researcher, and the value of publications in science
- Interdisciplinarity is difficult: Biologists often view AI models as “black boxes” and prefer empirical experimental validation over generated plots. Conversely, data scientists frequently underestimate the sheer complexity of biological problems.
- The Myth of Perfect Data: Computer scientists entering biology shouldn’t expect beautifully curated, labeled datasets ready for training. Raw data is often useless for AI without the direct, ongoing intervention of domain experts to translate it into useful information.
- Shared Vocabularies: Success requires both sides to align on terminology early. The word “data” means completely different things depending on the discipline, and data scientists must understand what metrics biologists actually value to build trust.
- The “Lethal Trifecta”: A major security concern arises when agentic generative AI tools combine untrusted input data, access to private data, and internet access. This combination creates a significant risk of private health or donor data leaking to bad actors via prompt injection.
- The Accuracy Imperative: AI-generated insights must be rigorously validated. Researchers cannot excuse incorrect biological outputs simply because an AI generated them quickly.
- Environmental Costs: The power consumption of large AI models is immense. To combat this, researchers must benchmark sophisticated models against traditional methods; often, a low-power random forest model performs nearly as well as an energy-intensive transformer model.
The path forward requires deep integration between data scientists and subject matter experts. By starting with the biological problem rather than the AI model, interdisciplinary teams can build the necessary trust to turn raw data into genuinely meaningful scientific breakthroughs.