AI-human collaborations in cancer

Cancer research continues to become more complex. Research questions increasingly require a wide range of expertise to address. Additionally, there is a growing volume and variety of data that spans scales from molecular data to epidemiological data. Furthermore, publications are increasing at an exponential rate making it more challenging to identify the most important new findings. However, recent advances in AI can help overcome these challenges. Large language models (LLMs) have been trained on vast quantities of text data, including scientific literature, and have shown high accuracy in answering scientific questions. Early work has begun to explore the use of LLMs to conduct scientific research.

This challenge aims to go beyond the ability of AI to extract information, make predictions, or conduct standard research tasks, making the leap to AI agents that can collaborate with humans to generate paradigm-shifting discoveries that advance cancer research. Agents are layers on top of LLMs that observe and collect information, provide input to the model, can conduct complex tasks, can have specific expertise, and can work together with other agents and humans. 

This challenge will require technological innovation to develop AI agents that can generate novel hypotheses that address important challenges in cancer research. The AI agents will also need to develop research plans to test these hypotheses that can be experimentally validated. Iterative cycles of experimental design and validation may be required. It is likely that multiple AI agents with different expertise and purposes will be required. It will also be important for there to be a human-in-the-loop to provide high level guidance and context that AI agents may lack.

The cancer research question that the AI agents are addressing should be complex and transdisciplinary and go beyond straightforward tasks and predictions. Of particular interest is cancer initiation and progression and the development of AI agents that can investigate the malignant conversion of precancerous lesions across tissues and identify novel therapeutics to prevent cancer initiation or progression in patients at risk.

This challenge will require close collaboration between AI researchers and cancer researchers and may also include machine learning specialists, data scientists, biologists, clinical researchers, and ethical advisors. Strategic academic-private partnerships are encouraged, with private partners contributing funds. 

Cancer research is becoming more complex, requiring scientists with diverse expertise to make sense of vast amounts of data, from genetic mutations to patient health records. At the same time, scientific discoveries are being published at an overwhelming pace. Recent advances in artificial intelligence (AI), particularly large language models such as ChatGPT, offer a powerful new way to accelerate research. AI has already shown promise in analysing scientific literature and making predictions, but the next big step is developing advanced AI tools that can work interactively with scientists. These AI-powered assistants, known as AI agents, could help researchers explore complex problems and guide research in new directions.

This Cancer Grand Challenge aims to develop AI tools that do more than process information. Rather than simply analysing existing data, these AI agents will work alongside researchers to propose new theories about cancer and help design experiments to test these ideas. By combining human expertise with AI, these AI agents could help uncover new ways to prevent, detect, or treat cancer. If successful, this research could transform how we study cancer, leading to groundbreaking new therapies. Ultimately, this could help more people receive effective treatments and improve survival rates.