OpenAI’s launch of ChatGPT on March 14 heralds a new era of artificial intelligence that will have profound implications for society, including the life science and healthcare industries. As when any new technology appears on the horizon, a tremendous amount of overheated hyperbole has dominated coverage of the topic in the months since. There are, however, some truly alluring potential use cases for generative AI applications such as ChatGPT for the life sciences and healthcare industries – as well as some pressing limitations.
At Clarivate, we have been a longtime trailblazer in the implementation of AI to enhance our tools and solutions. For example, we are drawing on our connected data lakes in Cortellis™ and using machine learning to predict clinical trials progression, regulatory approvals and even valuations on M&A candidates.
We’re mindful of some challenges that life science and healthcare organizations must work through before this technology is mature enough for use in critical business decisions that may impact patient health. These include:
- Ensuring quality input data: AI applications can only ever be as good as the data fueling them. At Clarivate, we curate billions of proprietary best-in-class data assets which feed our machine, deep learning and large language models to power our insights, services and workflow solutions. Standardizing disparate data sets and processes represents a major roadblock to effective use of AI generally, including generative AI.
- Vetting sometimes-spotty output: Generative AI’s “hallucination” problem, wherein large language models produce responses that may be syntactically and semantically correct but factually incorrect, is well-documented. Responsible use of AI demands that outputs receive stringent human oversight to identify and eliminate machine-introduced errors. Our customers entrust our products and services to help them improve patient health, and we will not jeopardize that mission.
- Regulatory asymmetry: Laws governing generative AI vary widely across markets and are evolving rapidly as regulators scramble to address this emerging technology. Italy briefly banned ChatGPT before restoring it, and has vowed to review its competitors. Google and Meta have refrained from launching generative AI products (Bard and BlenderBot, respectively) in Europe, moves interpreted in the press as being motivated either by concern for that market’s stringent privacy laws or in protest of them. The sweeping new AI Act approved by the European Parliament in June requires disclosure of content generated by AI, designs that prevent generation of illegal content and publication of summaries of copyrighted training data used. Other countries are weighing drastic actions over data privacy concerns.
- The intersection of AI and IP: Who owns the data? Who owns the models? How can companies ensure that their data doesn’t fall into the hands of competitors through large language models? Here, again, we see differing approaches by regulators internationally – Japan, for example, has declared training data exempt from copyright protections, and Israel’s Ministry of Justice has staked out a similar position. Regulators in other markets are taking a more cautious approach. To protect the data of customers and our own, we have adopted stringent company-wide guidelines on the use of generative AI applications and tools.
At the same time, there are some obvious potential use cases that could significantly speed up drug development and better ensure that the right medicines reach the right patients, improving outcomes. These include:
- Assisting in molecule design with desired properties using predictive analytics, a hotbed of pharma dealmaking in recent years, including last year’s Sanofi-Exscientia collaboration, which featured a potential value of up to $5.2 billion (read our recent report on biopharma dealmaking to learn more about activity in this space).
- Predicting safety and efficacy by using large language models to identify relevant documents and ways to optimize existing solutions. As an example, Clarivate recently partnered with VeriSIM Life to use its BIOiSIM® platform, which uses AI and machine learning to predict compound safety and efficacy and help inform go-no go decision making, in tandem with Cortellis Drug Discovery Intelligence™ data. In addition, large language models may inform the use of machine learning to identify safety-relevant documents.
- Production of synthetic real-world data to augment and improve machine learning models, while maintaining patient privacy.
- Accelerating semantic search across medical and scientific literature to enable real-time natural language searches and curation of vast datasets (e.g., policy trackers) across geographic and language barriers. Clarivate is exploring the use of generative AI to augment the content curation process and to allow advanced search functionality across our interconnected data sets. We recently partnered with Nyqyist Data, Inc. to offer our medtech and research center customers access to clinical and regulatory intelligence from over 500,000 devices and three million clinical studies across major markets using the Nyquist Data platform, which uses proprietary AI-based algorithms to reveal insights previously hidden in unstructured data.
- Making processes more efficient throughout life science organizations and driving costs out of repetitive activities that can be accelerated exponentially.
Jonathan Gear, Chief Executive of Clarivate said: “AI and machine learning are poised to revolutionize how life sciences companies deliver treatments that can transform patient lives. Clarivate was an early adopter of AI technology that enables researchers to optimize treatment development from early-stage drug discovery through commercialization. We are committed to investing in innovative technologies that will support our customers efforts to solve healthcare’s biggest challenges across the entire drug, device and medical technology lifecycle.”
For more information on Clarivate and artificial intelligence, read our announcement regarding the launch of our new artificial intelligence tools.
Podcast: How science and big data are helping in the search for an effective treatment of a rare disease that causes blindness. This podcast demonstrates how AI can make a positive difference in the lives of individual patients.