Hype, speculation and alarming forecasts aside, few specialists doubt that artificial intelligence will change the world. Who gains and at what cost remains unsettled.<\/p>\n
History shows that breakthroughs bring crises as well as opportunities, forcing societies to rebalance. One field, though, has long shown almost unambiguous gains from technological progress: medicine.<\/p>\n
ForkLog examines how, even today, AI accelerates the creation of new drugs, optimises lab workflows, sharpens diagnosis and reshapes approaches to treatment.<\/p>\n
Most medicines act by binding to protein receptors\u2014molecular structures that regulate cellular function and feature in almost all bodily processes.<\/p>\n
AI systems can analyse receptor structures and predict which compounds will bind most effectively with minimal side-effects. That is why tasks that once took years of lab work are increasingly being solved<\/a> in months.<\/p>\n According to estimates<\/a> by the World Health Organization (WHO), most new pharmaceuticals in the coming years will in some way be developed with AI.<\/p>\n In 2024 the Nobel Prize in Chemistry went to<\/a> David Baker, Demis Hassabis and John Jumper. The latter two work at Google DeepMind and were recognised for protein-structure prediction methods, including AlphaFold<\/a>, built on machine learning.<\/p>\n In 2018 AlphaFold took first place in the protein-structure \u201ccontest\u201d Critical Assessment of Structure Prediction<\/a> (CASP), excelling in the hardest categories. Two years later, at the next CASP, a new version\u2014AlphaFold 2\u2014won.<\/p>\n In 2021 Google DeepMind released<\/a> the AlphaFold2 code and a database of predicted protein structures. Around the same time Hassabis founded Isomorphic Labs<\/a>, an Alphabet subsidiary focused on AI for drug discovery.<\/p>\n In 2024 Isomorphic Labs struck<\/a> partnerships with Eli Lilly and Novartis. The deals envisaged funding of up to $1.7bn and $1.2bn, respectively, for the company\u2019s AI research. In 2026 Isomorphic Labs also announced<\/a> a partnership with Johnson & Johnson.<\/p>\n In February 2026 Isomorphic Labs unveiled a universal drug-design environment, Drug Design Engine<\/a> (IsoDDE), built on AlphaFold technologies.<\/p>\n The firm is working on oncology and immunology. Despite AI-driven acceleration, projects remain at the preclinical<\/a> stage. The company expects<\/a> to begin first-in-human trials in the coming years.<\/p>\n Founded in 2012, Exscientia was among the first companies to apply machine learning systematically to drug design.<\/p>\n In 2020 the drug DSP-1181 for therapy in OCD<\/span> became<\/a> the first AI-designed product to reach clinical trials. The project was conducted with Japan\u2019s Sumitomo Dainippon Pharma, which handled synthesis and lab tests, guided by Exscientia\u2019s theoretical results.<\/p>\n By 2023 the company had<\/a> eight candidate molecules ready, developed \u201csubstantially faster\u201d than the industry average.<\/p>\n In 2024 Recursion Pharmaceuticals acquired<\/a> Exscientia in a $688m deal. Some research programmes were shut.<\/p>\n By then several drugs had reached<\/a> phase II\u2014testing efficacy and side-effects in a cohort of 100\u2013300 patients.<\/p>\n The tie-up with Recursion Pharmaceuticals combined Exscientia\u2019s AI systems with an automated lab-testing complex. Recursion also built<\/a> its own AI supercomputer, BioHive-2, on NVIDIA H100s to train specialised models.<\/p>\n The company also contributed<\/a> to the open, generative model Boltz-2 for predicting proteins\u2019 three-dimensional structures.<\/p>\n By 2025 Recursion Pharmaceuticals focused<\/a> on four oncology programmes and two related to rare diseases. Several drugs sit between phase I and II:<\/p>\n REC-3565, designed for treating chronic lymphocytic leukaemia<\/a>, is in phase I trials.<\/p>\n Founded in 2014, Insilico Medicine<\/a> is another significant player in AI-driven drug development.<\/p>\n In 2017 Insilico Medicine was named<\/a> to Nvidia\u2019s top five for social impact.<\/p>\n The company uses<\/a> AI across the development cycle:<\/p>\n One early AI milestone for Insilico Medicine is Rentosertib<\/a> (ISM001-055), linked to fibrosis treatment. Development took 18 months from an AI-identified target to a candidate molecule. As of 2025, Rentosertib is in<\/a> phase II trials.<\/p>\n Also in 2024, the AI-designed immunomodulatory ISM3312 for COVID-19 and other viral infections completed<\/a> phase I. ISM3091, related to cancer therapy, was admitted<\/a> to patient testing.<\/p>\n Specialists estimate that about 90% of all medical information is represented<\/a> by images such as X-rays and scans. These data are critical to diagnosis, but labour-intensive and tricky to read.<\/p>\n Machine-learning methods, especially convolutional neural networks, suit complex visual pattern recognition. Much like human vision, such systems can pick out contrasts, edges, shapes and textures. This allows<\/a> tumours, bleeds and other anomalies to be found with high confidence.<\/p>\n Model training benefits from high-quality datasets\u2014large archives of documented images with expert annotations.<\/p>\n In 2024 researchers at Harvard Medical School presented<\/a> an AI model, Chief, that can detect several cancers. According to the team, it correctly spotted signs of disease on digital images in 94% of cases.<\/p>\n In 2025 America\u2019s Food and Drug Administration (FDA) granted<\/a> \u201cbreakthrough device\u201d status to Damo Panda, a model from Alibaba\u2019s Damo Academy.<\/p>\n According to its developers, the system can flag pancreatic cancer on scans before symptoms appear\u2014vital for this particularly insidious disease.<\/p>\n In 2026 a significant breakthrough<\/a> in AI diagnostics was REDMOD, developed by the Mayo Clinic, a US non-profit.<\/p>\n The model, also aimed at detecting pancreatic cancer, outperformed<\/a> specialists at early-stage diagnosis. The researchers said it found pathological changes on scans a median 475 days before diagnosis.<\/p>\n Google is a key provider of AI for medical diagnostics and research.<\/p>\n The company offers MedGemma, an open family of models for medical text, image and audio analysis MedGemma<\/a> based on Gemma 3.<\/p>\n Through Health AI Developer Foundations<\/a>, developers can access open weights and tools.<\/p>\n Google collaborates with clinics and research organisations, focusing on foundational technologies.<\/p>\n In 2019 the company introduced<\/a> a model to detect and forecast lung cancer. It matched or beat a panel of six certified radiologists.<\/p>\n In 2020, with Northwestern Medicine, researchers demonstrated<\/a> a system for mammogram analysis that could detect cancer at a specialist\u2019s level.<\/p>\n In 2024 Google Cloud and Germany\u2019s Bayer announced<\/a> a platform for X-ray screening. It reviews imaging histories and medical records to suggest possible pathologies.<\/p>\n Tech giant Nvidia and American medtech firm GE HealthCare, a maker of X-ray equipment, are developing<\/a> an AI system for autonomous image acquisition.<\/p>\n Unlike models that analyse images after the fact, this system aims to reduce clinicians\u2019 routine workload and standardise diagnostics.<\/p>\n The first phase will focus on X-rays and ultrasound.<\/p>\n GE HealthCare also plans to use NVIDIA Isaac for Healthcare<\/a>, a platform for building autonomous medical systems, including surgical robots.<\/p>\n Founded in 2016, PathAI has built a \u201cdigital pathology platform\u201d, AISight Dx<\/a>, for primary diagnosis in clinical settings.<\/p>\n The system provides a workspace for medical images with the option to plug in third-party algorithms.<\/p>\n The platform supports a set of CE\u2011IVD-certified, AI-based tools\u2014specifically, oncology \u201cplug-ins\u201d:<\/p>\n The platform also includes native functions for automated image analysis, diagnostic assistance and report drafting, but these are \u201cfor research use only\u201d and not permitted for clinical application.<\/p>\n AISight Dx also offers built-in assistive AI tools:<\/p>\n In 2022 the platform received<\/a> US FDA clearance via 510(k)<\/span> and Europe\u2019s CE mark<\/a>, attesting to safety for consumers and the environment.<\/p>\n In 2025 PathAI announced<\/a> a partnership with the Moffitt Cancer Center in Florida, USA, to deploy AISight Dx in diagnostics. In 2026 the company signed<\/a> a similar agreement with University Hospital Zurich.<\/p>\n In May 2026 Swiss drugmaker Roche said<\/a> it would acquire PathAI in a deal worth over $750m.<\/p>\n As elsewhere, AI in medicine exposes old problems and creates new ones.<\/p>\n AI assistants, especially those based on LLM<\/span>s, are prone to hallucinations.<\/p>\n In a research paper<\/a> on Med-Gemini by Google, an error surfaced: the model \u201cinvented\u201d a non-existent brain region called the basilar ganglia.<\/p>\n The hallucination blended two real anatomical terms: the basal ganglia<\/a> and the basilar artery<\/a>. The developers blamed a typo, but several specialists called<\/a> the incident a worrying example of the risks of deploying AI assistants in medicine.<\/p>\n Researchers at Stanford University found<\/a> that some models could convincingly diagnose from medical images without access to the images themselves.<\/p>\n One system scored highly \u201cblind\u201d on a radiology test. GPT-5, Gemini 3 Pro and Claude Opus 4.5 \u201cconfidently described visual details\u201d on non-existent images.<\/p>\n According to a June study<\/a>, in a medical context 7.1% of GPT-4\u2019s replies to patient questions were incorrect and could have caused significant harm. One in 156 posed a life-threatening risk.<\/p>\n As of 2025, tools that auto-generate documentation from doctor\u2013patient conversations introduced errors<\/a> into 70% of clinical notes\u2014adding false facts, omitting points and muddling concepts.<\/p>\n Beyond inventing organs, LLMs are notoriously<\/a> opaque, making it hard for humans to interrogate their reasoning.<\/p>\n Poorly representative datasets can bake in<\/a> biases and spurious correlations.<\/p>\n Meanwhile, familiar issues for AI assistants\u2014user over-reliance and data privacy\u2014are only sharper in healthcare.<\/p>\n WHO experts class<\/a> medical AI as high risk.<\/p>\n Under Europe\u2019s AI Act<\/a>, from August 2026 such systems must meet special requirements for risk management, reporting and human oversight.<\/p>\n Despite the challenges and risks, WHO sees promise in medical AI\u2014given proper rules and government oversight.<\/p>\n The US FDA is also optimistic about AI\u2019s prospects<\/a>, while acknowledging that current regulation is dated. In the US these systems are formally classed as Software as a Medical Device<\/span>.<\/p>\nAlphaFold and Isomorphic Labs<\/h3>\n
Exscientia and Recursion Pharmaceuticals<\/h3>\n
\n
Insilico Medicine<\/h3>\n
\n
Diagnostics and research<\/h2>\n
Google\u2019s initiatives<\/h3>\n
NVIDIA and GE HealthCare\u2019s autonomous imaging<\/h3>\n
PathAI\u2019s diagnostic platform<\/h3>\n
\n
\n
Pitfalls and limits<\/h2>\n