Artificial intelligence (AI) is widely considered the technology that will bring the most benefits to clinical trials. In the annual GlobalData State of the Biopharmaceutical Industry survey of pharmaceutical professionals, AI came out on top for technology with the greatest impact on the industry in consecutive years 2020-2023. But which type of AI is the most promising?
While the potential applications of AI are numerous, it is paramount that pharmaceutical companies looking to AI adopt a proactive, value-driven strategy rather than implementing technology in isolation. Major Language Models (LLMs) like ChatGPT have amazed the world with their ability to quickly take input from the Internet and generate large amounts of text for a given prompt.
This capacity for limitless creativity is exciting but also potentially troublesome, as LLMs often generate unreliable responses. Supervised machine learning (ML) often proves to be a better way to leverage the benefits of AI in the clinical research space while mitigating some of the drawbacks of LLMs, such as transparency, cost, and consistency.
LLMs are not reliable information generators
LLMs are able to understand and generate coherent human language text over long passages and can power translation models, content creation, and automated customer support bots. Their capacity for traditional natural language processing (NLP) also surpasses traditional ML methods, facilitating tasks such as classification, entity extraction, and relationship extraction.
But even at this preliminary stage, inevitable problems arise with LLMs. Computing costs remain high, as do risks, including manufacturing inconsistencies and regulatory hurdles. Particularly problematic are “hallucinations,” AI-generated responses that seem sincere and convincing, but are factually incorrect.
Consider the example below, where ChatGPT was asked to provide the MedDRA code for “heart attack.” The response was impressively fluid, clear and authoritative – and the code it returned was a real MedDRA code. The problem was that it wasn’t code for a heart attack.
The complete response gives the feeling that ChatGPT has understood the question and has in-depth knowledge of MedDRA. By using appropriate language from MedDRA, for example the phrase “preferred term,” the LLM conveys a seemingly in-depth knowledge of the terminology of medical conditions. However, the medical code she provided in this case is actually the MedDRA code for mesothelioma, and not for myocardial infarction.
ChatGPT is not deliberately trying to mislead the user here. Rather, this hallucination is the result of the LLM choosing a highly likely next word at each position in the text as it generates the answer. Anything that looks like a valid MedDRA code can be run, regardless of whether it is correct.
Therefore, careful consideration is required when constructing a solution around generative AI. Even adding a human validation process to AI-generated answers can be a wasted effort because the original generated text is so authoritative and compelling.
However, supervised machine learning models therefore have an inherent advantage. Because they are considered unreliable, they have no ability to ‘convince’ the user through a convincing story. Instead, the user gains a level of familiarity and trust with the AI model.
The benefits of supervised machine learning
Given concerns around data privacy and transparency, depending on the use case, LLMs may not be the ideal choice in certain clinical trials. For many use cases, supervised ML models are a fraction of the cost of LLMs and are typically more effective, scalable, and maintainable.
Industry leaders like Merative’s Zelta are expanding their philosophy of the right tool for the right job and putting customer value first in AI implementations in clinical trials. It’s easy to get sucked into the hype of LLMs, but investing in them – and focusing your AI strategy solely on LLMs – may be unnecessary when supervised ML models deliver a more efficient and concrete return on investment. Zelta delivers clear, measurable value statements supported by pragmatic solutions, while avoiding buzzwords and overpromising in deliverables.
LLMs such as ChatGPT, Llama and Gemini have made impressive progress in language understanding and generation, offering valuable applications in various domains. However, Zelta has discovered more practical benefits in supervised machine learning to help sponsors and CROs achieve their clinical trial goals. Supervised ML provides greater control over the data and ensures that it meets the necessary standards for accuracy, compliance, and relevance. This approach reduces the risk of unpredictable results and allows them to refine results for more reliable and actionable insights. While LLMs offer exciting potential, the precision and overview of supervised learning better meet their high-impact needs at this stage.
For many clinical trials, supervised ML is ideal for meeting specific data requirements, unlike other one-size-fits-all AI solutions. These more specific tasks also require less data, reducing the burden on managers training the models, and they can be built into the workflow to create a strong feedback loop with human users.
While it can be difficult and expensive to obtain training data for supervised models, the long-term cost savings are significant, especially considering the speed at which supervised ML models can run. When implemented properly, they can make a difference in transforming long-standing data challenges into new strategic advantages.
How Zelta uses AI to improve clinical trial services
Market leader Zelta has been integrating AI and advanced automation for more than a decade. The cloud-based eClinical platform makes managing clinical trials simple and easy, powered by AI for medical coding and CDASH annotation. This machine learning software has saved hundreds of hours of human labor and improved accuracy and efficiency in clinical trials.
With AI poised to continue to have an outsized impact in clinical research studies, Zelta aims to expand its AI-enhanced features in the future – but only in accordance with tested use cases, rather than blindly following the crowd on the most exciting trends.
Download the report below to learn more about Zelta’s approach to deploying supervised machine learning for clinical trials, and their decade-long track record of deploying AI and automation in the clinical space to deliver measurable results.
