For centuries, we used willow bark to treat pain long before understanding salicylic acid. We observed lithium’s calming effects decades before grasping its neurochemistry. Medicine’s greatest leaps often begin with a simple truth:
It works – even if we don’t yet know why.
This tradition now continues through ‘black box’ AI, revealing connections across petabytes of biological data that would remain invisible to even the most trained human eye working in isolation. Like a master chess player who intuitively knows the right move but can’t fully articulate their strategy, these algorithms leverage deep neural networks to uncover biological insights no reductionist approach could replicate.
Today, NVIDIA Clara™ – spanning Biopharma, Medpharma and Genomics – with its NVIDIA BioNeMo™ framework for generative biology, is turning this principle into a computational superpower, enabling accelerated AI-driven discovery by training on vast biological datasets beyond human capacity.
Clara’s models uncover hidden patterns, propose novel drug candidates and power AI Factories that create end-to-end discovery ecosystems – custom pipelines that integrate NVIDIA’s accelerated computing with proprietary biological AI to generate, simulate, validate and learn across previously inaccessible chemical spaces. These platforms support rapid biomolecular AI model development, including virtual screening, protein binder design, de novo drug design, molecular property prediction, optimised molecule generation, model building and training.
This isn’t a rejection of human intelligence – it’s an augmentation beyond our natural limits.
The most visionary biotech leaders recognise a paradigm shift: Requiring AI to conform to human-interpretable logic might exclude its most transformative discoveries. The unexplainable is where breakthroughs live. The question is no longer – Can we rationalise it? But – Can we harness it?
Through this article let us explore evolving regulatory frameworks, the surrounding discussion and the emergence of a new paradigm:
It has never been a strict requirement that a drug’s mechanism of action be fully explained prior to approval, so long as robust clinical evidence shows the drug is effective and safe. Analyses have found that an estimated 10–20% of FDA-approved drugs have no clearly identified mechanism or known molecular target at the time of approval (1).
In recent years, there has been greater use of expedited approval pathways that rely on preliminary evidence, which align with an “effectiveness-first” ethos (approve based on observed effect, then explain/confirm later). For instance, the FDA’s Accelerated Approval program allows drugs for serious conditions to be approved based on a surrogate endpoint that is reasonably likely to predict clinical benefit (2). EMA in Europe and PMDA in Japan also prioritise patient outcomes over theoretical understanding in approval decisions. It’s generally accepted that demonstrated clinical efficacy is the gold standard.
South Africa became the first country in the world to formally grant a patent that listed an AI as the inventor. In July 2021, the South African Companies and Intellectual Property Commission issued a patent for a food container invention, crediting “DABUS” (an AI system created by Dr. Stephen Thaler) as the inventor and naming Thaler as the patent owner. This was a landmark formality as it stands as the first patent acknowledging a non-human inventor (3).
Policy bodies globally are examining whether laws should be updated to accommodate AI-generated inventions. For example, the U.S. Executive Order on AI (Oct 2023) directed the USPTO to issue guidance on “inventorship and the use of AI in the inventive process,” acknowledging the need to clarify how inventions involving AI are handled. There’s also an ongoing debate about whether inventions autonomously created by AI (with no human inventor) should receive some kind of protection (if not patents, perhaps a new IP right) (4).
The idea of AI systems being named as inventors on patent applications is a highly debated, emerging issue in intellectual property law.
As regulatory bodies grow more comfortable approving therapies based on demonstrable outcomes rather than mechanistic clarity and as legal frameworks develop acknowledging AI’s role in invention, a broader shift is taking place: the locus of innovation is laying the foundation of machine-generated insights. This shift doesn’t just invite us to accept the unfamiliar – it compels us to build infrastructure that can harness it. The convergence of regulatory openness and technological capability is creating fertile ground for entirely new modes of discovery.
What’s needed now is a way to operationalise this potential at scale – translating black-box insights into real-world therapies, quickly, reliably and with minimal bottlenecks.
This is where AI Factories come into focus. They sit at the heart of this evolving landscape – where deep learning doesn’t just aid discovery but becomes a co-discoverer. Unlike traditional labs these self-improving, computational ecosystems operate at a scale and abstraction level that challenges our regulatory and intellectual property frameworks. AI Factories don’t merely accelerate research – they industrialise it, turning biological discovery into a process of continuous generation, evaluation and iteration. They blur the line between hypothesis-driven science and hypothesis-generating computation, producing candidate molecules, biological models and therapeutic strategies faster than conventional R&D timelines could imagine.
In this new paradigm, the source of innovation may not be a lone scientist with a single theory, but a distributed architecture of AI models refining each other's outputs in real time. As this shift unfolds, AI Factories position forward-looking companies to capitalise on this moment. Startups and established firms alike can now explore vast, previously unreachable chemical and genetic terrains – guided by models that don’t need to ‘understand’ in the human sense but demonstrably work. This opens the door to personalised treatments, ultra-rare disease therapies and mechanisms of action that would never have emerged from hypothesis-driven research alone. In a world where the most disruptive breakthroughs may come from systems that defy classical interpretability, AI Factories become the crucible where human insight and machine intuition co-create the future of medicine.
As we move toward this future, the infrastructure behind that intelligence becomes as vital as the models themselves. Boston Limited, as a long-standing NVIDIA Elite Partner, plays a crucial role in this evolution, delivering the high-performance computing systems, AI-ready architecture and bespoke deployment expertise that turn possibility into practice. Whether it's enabling AI factories for generative chemistry or supporting real-time molecular simulation pipelines, Boston empowers researchers, biotech firms and pharmaceutical giants to harness their full potential.
In this age of accelerated discovery, trust in AI begins with trust in the systems that power it – and Boston Limited stands at the foundation of that trust, helping scientists transform molecules into medicines and ambition into action.
Senior Campaign Executive
Boston Limited
References:
First-in-Class Drugs Get Faster FDA Review Than EMA | Technology Networks
South Africa issues world's first patent listing AI as inventor - The Global Legal Post
UK Supreme Court Rules Against AI Inventorship of Patents | White & Case LLP
To help our clients make informed decisions about new technologies, we have opened up our research & development facilities and actively encourage customers to try the latest platforms using their own tools and if necessary together with their existing hardware. Remote access is also available
The International Conference for High Performance Computing, Networking, Storage, and Analysis