AI and scientific discovery: a collective challenge

• Data
• Intelligence artificielle
• Open innovation
• Prospective
• Sciences humaines

Share on Twitter : AI and scientific discovery: a collective challenge (New window) Share on Linkedin : AI and scientific discovery: a collective challenge (New window)

In Montreal, a novel compound designed entirely with AI entered Phase II clinical testing this past spring. At Yale University, researchers recently succeeded in identifying a potential new cancer therapy pathway using C2S-Scale, an AI developed by Deepmind. In France, TechBio start-up Aqemia is using an AI physics engine to develop future drugs. Determining the efficacy of a pharmaceutical compound with artificial intelligence takes just six months, as opposed to six years of traditional research. In laboratories at Berkeley, augmented research with DeepMind has resulted in the ‘creation’ of some 40 new materials – stimulating news in the light of the current race to secure supplies of rare metals. Could it be that the future of innovation is already being determined by AI algorithms, and not just by laboratory experiments?

AI-driven discovery will only progress if communities learn to work together with more interdisciplinary collaboration and education, and open and standardized data…

De la recherche au business

According to a report on the role of AI in scientific research from the European Commission’s Joint Research Centre (JRC), the phenomenon is destined to grow. In many laboratories, the technology has already assumed the status of “a creative partner, transforming every stage of the scientific process: from the generation of concepts and the testing of hypotheses, to the accelerated analysis of data and publications presented to the scientific community.”

Head of Research and Intellectual Property at Orange Lyse Brillouet argues that this is welcome news for companies: in the November 2025 Hello Future Let’s Talk Tech podcast, she pointed out that research should “give companies technological agency. Our strategy is based on dual timeframes: pure research and applied innovation.” In short, research also drives business.

However, if AI research is to deliver meaningful innovation, Hugging Face researchers Georgia Fanning and Avijit Ghosh warn that the spirit of collaboration among scientists needs to be reinforced to overcome obstacles that are largely social rather than technical: dysfunctional communities, siloed initiatives rather than interdisciplinary challenges, data that remains fragmented (insufficient standards and proprietary formats) and infrastructural inequality (a lack of equal access to high-performance computing, notably in Global South). The researchers insist that AI-driven discovery will only progress if communities learn to work together with more interdisciplinary collaboration and education, open and standardized data, and the creation of infrastructure that is accessible to all.

A team spirit

Several projects pursued at Orange illustrate how this team spirit applies in practice, notably the commercial implementation of post-quantum cryptography and quantum key distribution: the fruit of eight years of pure research and a partnership with Toshiba Europe, the Orange Quantum Defender network security service launched in mid-2025. “Research is at the heart of the innovation pipeline at Orange, where it provides the company with tools and the competitive advantage it needs to navigate a world of perpetual change,” points out Lyse Brillouet. Research undertaken by Orange system and cybersecurity engineer Kahina Lazri highlights a similar quest for interdisciplinarity. “Our work focuses on eBPF formal verification to ensure that no programme can compromise underlying infrastructure” explains the researcher. Here too, work on the project is being conducted in collaboration with a partner, Inria Paris — yet another example of the collective spirit that scientists must cultivate in the drive to fulfil the potential of AI-powered discovery.