How AI Drives Scientific Research Breakthroughs

A nice review article "Transforming Science with Large Language Models: A Survey on AI-assisted Scientific Discovery, Experimentation, Content Generation, and Evaluation" covers the scope of tools and approaches for how AI can support science. Some of areas the paper covers: (link in comments) 🔎 Literature search and summarization. Traditional academic search engines rely on keyword-based retrieval, but AI-powered tools such as Elicit and SciSpace enhance search efficiency with semantic analysis, summarization, and citation graph-based recommendations. These tools help researchers sift through vast scientific literature quickly and extract key insights, reducing the time required to identify relevant studies. 💡 Hypothesis generation and idea formation. AI models are being used to analyze scientific literature, extract key themes, and generate novel research hypotheses. Some approaches integrate structured knowledge graphs to ground hypotheses in existing scientific knowledge, reducing the risk of hallucinations. AI-generated hypotheses are evaluated for novelty, relevance, significance, and verifiability, with mixed results depending on domain expertise. 🧪 Scientific experimentation. AI systems are increasingly used to design experiments, execute simulations, and analyze results. Multi-agent frameworks, tree search algorithms, and iterative refinement methods help automate complex workflows. Some AI tools assist in hyperparameter tuning, experiment planning, and even code execution, accelerating the research process. 📊 Data analysis and hypothesis validation. AI-driven tools process vast datasets, identify patterns, and validate hypotheses across disciplines. Benchmarks like SciMON (NLP), TOMATO-Chem (chemistry), and LLM4BioHypoGen (medicine) provide structured datasets for AI-assisted discovery. However, issues like data biases, incomplete records, and privacy concerns remain key challenges. ✍️ Scientific content generation. LLMs help draft papers, generate abstracts, suggest citations, and create scientific figures. Tools like AutomaTikZ convert equations into LaTeX, while AI writing assistants improve clarity. Despite these benefits, risks of AI-generated misinformation, plagiarism, and loss of human creativity raise ethical concerns. 📝 Peer review process. Automated review tools analyze papers, flag inconsistencies, and verify claims. AI-based meta-review generators assist in assessing manuscript quality, potentially reducing bias and improving efficiency. However, AI struggles with nuanced judgment and may reinforce biases in training data. ⚖️ Ethical concerns. AI-assisted scientific workflows pose risks, such as bias in hypothesis generation, lack of transparency in automated experiments, and potential reinforcement of dominant research paradigms while neglecting novel ideas. There are also concerns about the overreliance on AI for critical scientific tasks, potentially compromising research integrity and human oversight.

🔬 𝐓𝐡𝐞 𝐀𝐈-𝐏𝐨𝐰𝐞𝐫𝐞𝐝 𝐑𝐞𝐬𝐞𝐚𝐫𝐜𝐡 𝐑𝐞𝐯𝐨𝐥𝐮𝐭𝐢𝐨𝐧 — 𝐀𝐧𝐝 𝐖𝐡𝐚𝐭 𝐈𝐭 𝐌𝐞𝐚𝐧𝐬 𝐟𝐨𝐫 𝐭𝐡𝐞 𝐅𝐮𝐭𝐮𝐫𝐞 𝐨𝐟 𝐌𝐚𝐧𝐚𝐠𝐞𝐦𝐞𝐧𝐭 𝐒𝐜𝐢𝐞𝐧𝐜𝐞 We're standing at an inflection point in academic management research. Generative AI and Agentic AI are no longer just productivity tools — they're fundamentally reshaping how scientific knowledge is created, validated, and disseminated. For decades, the bottleneck in research has been tacit knowledge — the hard-won, deeply personal expertise that allowed only a handful of skilled researchers to produce high-quality scientific papers. That bottleneck is dissolving. As AI-assisted writing, synthesis, and agentic research workflows become mainstream, paper production is being decoupled from individual researcher expertise. The implications are staggering: journals like Technovation are already projecting a 3× increase in submissions within just a few years. This creates an immediate second-order effect: journals will have no choice but to deploy AI-driven pre-screening and peer review assistance to manage the flood. Human editors simply cannot scale fast enough. But here's the most profound shift — and one we're not talking about enough: When production is automated, the scarce resource becomes ideation. The competitive advantage will no longer lie in writing a rigorous paper. It will lie in asking the right question — identifying the novel insight, the overlooked gap, the counterintuitive hypothesis that is genuinely worth investigating. Research impact will be determined upstream, in the discovery phase, not the production phase. This has enormous consequences for PhD curricula. We need to urgently rethink what we're training doctoral researchers to do: ✅ Less emphasis on methodological execution (AI handles much of this) ✅ More emphasis on intellectual curiosity, critical thinking, and research question formulation ✅ Training in AI literacy — knowing how to direct, interrogate, and validate agentic research systems ✅ Developing the judgment to distinguish publishable from impactful The researchers who will thrive are not those who produce the most — but those who notice what others haven't noticed yet, for instance by networking intensively with managers. The age of Human-AI collaborative discovery is here. Are our PhD programs ready for it? 💬 I'd love to hear from researchers, supervisors, and journal editors — how is your institution adapting? Are we equipping the next generation for this new reality? #GenAI #AgenticAI #AcademicResearch #ResearchInnovation #ScientificPublishing #PhDEducation #FutureOfResearch #AIinScience #HigherEducation #Technovation #KnowledgeManagement #ResearchStrategy

Delighted to share a new paper in Cell from our Google DeepMind Google Research collaboration with Profs José R Penadés and Tiago Costa at Imperial College London, demonstrating how AI can significantly accelerate biomedical discovery. When our teams first met, we found ourselves at a unique moment. Both our groups were on the cusp of breakthroughs unknown to the outside world. Our team Vivek Natarajan/ Juraj Gottweis had developed an AI co-scientist system designed to assist researchers by generating novel, impactful hypotheses. We wanted to test its potential on grand challenges for medicine like tackling antimicrobial resistance. Putting it into the hands of great scientists like Jose and Tiago was a perfect opportunity for this. Their group had a paper under peer review marking a pivotal milestone after many years of work. They reported proof of a novel mechanism for gene transfer through 'microbial piracy'. Their research revealed how mobile genetic elements "cf-PICIs" can hijack the tails from viruses to create hybrid particles, allowing them to inject DNA -- which can carry genes for antibiotic resistance -- into a wide range of bacteria, explaining how these dangerous traits spread so efficiently. This was the result of years of meticulous experimental work but the breakthrough insight had never been made public and was under confidential peer review. Since their ingenious discovery was unknown to the wider world, Jose and Tiago posed this original research question to the system. The results were astounding to all of us. After several days of work, not only did the AI co-scientist system recapitulate their core finding, but in the Cell paper, Jose and Tiago go into the details of four other ideas proposed by the system that they are now excited to explore further. I've had many amazing moments at Google DeepMind, but a great highlight was hearing Jose's response to these outputs. He asked if we had somehow managed to cheat the test and read his original manuscript—the ideas generated by the AI so closely matched the original and real discovery. It's an extraordinary privilege to work on AI systems with such a profound potential to super-charge science, not by replacing human insight, but by accelerating it. The true impact comes from collaborating with experts like Jose and Tiago to tackle grand scientific challenges. We're delighted to share authorship on our paper in Cell and am excited for the many more discoveries this technology will help enable. You can read our paper, "AI mirrors experimental science to uncover a novel mechanism of gene transfer crucial to bacterial evolution" , and the original experimental discovery from José R Penadés Tiago Costa in the same issue of Cell here: https://lnkd.in/edDed3Gv

Exciting to see AI agents highlighted as a “Method to Watch” in Nature Methods. https://lnkd.in/dipBKDPn Lin Tang's piece points to a future where AI agents do far more than automate lab tasks; they actively help generate hypotheses, design experiments, and accelerate the search for new scientific theories in biology. This resonates strongly with our own work. In the last years, we’ve begun developing active learning loops powered by perturbation models, where AI proposes experiments, wet labs execute them, and models improve iteratively. This hands-on experience makes me particularly excited about a future where robust multimodal foundation models can be supplied directly to AI agents, giving them a rich prior about biological systems. What inspires me most is reframing experimental design as a search through hypothesis space. AI doesn’t replace scientific intuition — it amplifies it. It helps us navigate complexity, connect molecular and cellular states to patient-level phenomena, and ultimately push biological understanding toward deeper causal theories. We’re only at the beginning, but I believe that providing reliable perturbation and system models to AI agents will meaningfully accelerate discovery across biomedicine and beyond.

Scientific discovery and clinical medicine are often treated as distinct phases. But for patients with rare, complex, and undiagnosed diseases, this separation is a luxury they cannot afford. The timeline from understanding a genetic mechanism to accessing subspecialist care is often too long and too fragmented. Two new Google DeepMind Google Research collaborations with Stanford University School of Medicine, published in Advanced Science and Nature Medicine respectively last week, demonstrate how AI can bridge this gap. 1. Accelerating discovery (the science) In Advanced Science, we present one of the first wet-lab validated examples of AI-assisted genetic discovery . Our AI identified a novel genetic factor for hearing loss (Crym) in mice, which Dr Gary Peltz and team validated using CRISPR knock-in experiments to restore the wild-type gene and rescue the phenotype. We applied this agentic AI scaffold to human patients with complex, undiagnosed conditions in a retrospective manner. The system analyzed genomic data for rare diseases, such as IRAK4 deficiency and ODC1 mutations, successfully identifying causative variants that matched expert clinical assessments. 2. Scaling Expertise (the medicine) Discovery is only the first step; patients then need access to specialized care. As we note in our Nature Medicine paper, hypertrophic cardiomyopathy (HCM) is a leading cause of sudden cardiac death, yet ~60% of patients remain undiagnosed due to a lack of specialist centers . In our RCT using our research AI system AMIE, we showed AI could help bridge this gap. General cardiologists using AMIE reported the system helped their assessments in 57.0% of cases, missed no clinically significant findings in 93.5% of cases and reduced assessment time in 50.5% of cases. Crucially, these studies used models like Med-PaLM 2, Gemini 2.0 Flash, and Gemini 2.5 Pro with simple agentic scaffolds. If we can achieve this with previous generations, the potential for Gemini 3 and AI co-scientist to accelerate both the biology of discovery and the delivery of care is profound. Its a true privilege to collaborate with Euan Ashley, Jack W O'Sullivan MD, PhD, Dr Gary Peltz and their teams at Stanford Medicine. With incredible team mates at Google including Tao Tu, Anil Palepu, Alan Karthikesalingam MD PhD, Juro Gottweis and many more. Advanced Science paper - https://lnkd.in/dggduzka Nature Medicine paper - https://lnkd.in/dPEZQ4bz AI co-scientist blog - https://lnkd.in/gEDeaRfu AMIE blog - https://lnkd.in/gzkn2ywe

2025 could be the year we transition from AI systems that answer questions to autonomous AI agents capable of performing complex, real-world tasks independently. Last week, I explored the groundbreaking work being done by Google's AI Co-Scientists and Stanford and Chan Zuckerberg BioHub's Virtual Lab, highlighting how autonomous AI agents are already transforming complex research processes. Now, two additional studies further showcase the remarkable capabilities of advanced AI systems working to accomplish tasks: Researchers from Harvard and MIT introduced TxAgent, an AI agent leveraging an extensive toolkit of 211 specialized tools. TxAgent analyzes drug interactions, contraindications, and patient-specific health data to suggest personalized medical treatments in real-time. It thoroughly evaluates medications at molecular, pharmacokinetic, and clinical levels, factoring in individual patient risks such as comorbidities, existing medications, age, and genetic predispositions. By synthesizing vast biomedical evidence, TxAgent rapidly generates precise and tailored recommendations, dramatically optimizing healthcare delivery, which is particularly beneficial for resource-limited settings. Meanwhile, Sakana AI introduced "AI Scientist-v2," a remarkable autonomous AI researcher that generated the first-ever fully AI-written scientific paper to pass peer review at an ICLR 2025 workshop. This achievement marks a milestone in AI-driven research, demonstrating AI’s capability to independently execute the full scientific research cycle, systematically generate hypotheses, perform computational experiments using advanced machine learning models, rigorously analyze results, iteratively refine methodologies, and draft comprehensive manuscripts that meet the rigorous standards of peer review. LinkedIn: Why Your Next Coworker Might Be an AI Agent  https://lnkd.in/eAznknyh TxAgent: An AI agent for therapeutic reasoning across a universe of tools: https://lnkd.in/e7HW7j7t The AI Scientist Generates its First Peer-Reviewed Scientific Publication: https://lnkd.in/eYWmQs7m American Enterprise Institute Sakana AI Harvard Medical School Massachusetts Institute of Technology Harvard Data Science Initiative Coalition for Health AI (CHAI)

📄 A noteworthy paper in Nature Medicine on Co‑Scientist A compelling read in Nature Medicine exploring the concept of AI as a “co‑scientist” not as a replacement for human expertise, but as a powerful partner in scientific discovery. What stood out to me is the framing of AI as an augmentation of human creativity and reasoning: helping generate hypotheses, interrogate complex datasets, connect disparate knowledge domains, and accelerate insight generation while keeping scientists firmly in the driver’s seat. This work illustrates how thoughtfully designed AI systems can: - Support hypothesis generation and experimental design - Scale scientific reasoning across vast and heterogeneous data - Reduce cycle times from question to insight - Enhance, rather than dilute, scientific rigor For those of us working at the intersection of biomedical research, drug discovery, and AI, this paper reinforces an important message: the future is not AI versus scientists, but AI with scientists. Well worth reading for anyone thinking seriously about how AI can be embedded responsibly and effectively into R&D teams. https://lnkd.in/e7sDv5bz #ArtificialIntelligence #CoScientist #NatureMedicine #DrugDiscovery #BiomedicalResearch #RAndD #DigitalTransformation #AIinScience

𝘏𝘶𝘮𝘢𝘯 𝘪𝘯𝘵𝘶𝘪𝘵𝘪𝘰𝘯 𝘩𝘢𝘴 𝘭𝘰𝘯𝘨 𝘣𝘦𝘦𝘯 𝘵𝘩𝘦 𝘣𝘢𝘤𝘬𝘣𝘰𝘯𝘦 𝘰𝘧 𝘴𝘤𝘪𝘦𝘯𝘵𝘪𝘧𝘪𝘤 𝘴𝘰𝘧𝘵𝘸𝘢𝘳𝘦 𝘥𝘦𝘷𝘦𝘭𝘰𝘱𝘮𝘦𝘯𝘵. 𝘉𝘶𝘵 𝘸𝘩𝘢𝘵 𝘪𝘧 𝘵𝘩𝘪𝘴 𝘢𝘱𝘱𝘳𝘰𝘢𝘤𝘩 𝘭𝘦𝘢𝘷𝘦𝘴 𝘵𝘩𝘦 𝘣𝘦𝘴𝘵 𝘴𝘰𝘭𝘶𝘵𝘪𝘰𝘯𝘴 𝘶𝘯𝘥𝘪𝘴𝘤𝘰𝘷𝘦𝘳𝘦𝘥? A new paper from Google Research and Google DeepMind demonstrates an AI system that not only automates this process but achieves superhuman performance. This is crucial because the slow, manual creation of code for computational experiments severely limits the hypotheses scientists can explore, creating a major bottleneck in the cycle of discovery. The paper, "𝐀𝐧 𝐀𝐈 𝐬𝐲𝐬𝐭𝐞𝐦 𝐭𝐨 𝐡𝐞𝐥𝐩 𝐬𝐜𝐢𝐞𝐧𝐭𝐢𝐬𝐭𝐬 𝐰𝐫𝐢𝐭𝐞 𝐞𝐱𝐩𝐞𝐫𝐭-𝐥𝐞𝐯𝐞𝐥 𝐞𝐦𝐩𝐢𝐫𝐢𝐜𝐚𝐥 𝐬𝐨𝐟𝐭𝐰𝐚𝐫𝐞," introduces a system that tackles this challenge. It reframes software development as a "scorable task." The core methodology combines a Large Language Model (LLM) for intelligent code rewriting with a Tree Search (TS) algorithm. The TS intelligently navigates the vast space of possible solutions, guiding the LLM to iteratively refine code to maximize a quality score. It's not just generating code; it's evolving it. The results are stunning: - In bioinformatics, it discovered 40 novel methods for single-cell data analysis that outperformed all top human-developed methods on a public leaderboard. - For epidemiology, it generated 14 forecasting models that were more accurate than the official CDC ensemble for predicting COVID-19 hospitalizations. This represents a fundamental shift from scientists manually coding solutions to defining "scorable problems" and letting an AI discovery engine find the optimal software. By systematically exploring and even recombining complex research ideas, this system can uncover novel "needle-in-a-haystack" solutions that humans might never find. It could accelerate progress in fields from genomics to climate science by automating one of the most tedious parts of research. #AI #MachineLearning #ScientificDiscovery #GenerativeAI #Research

Researchers at #MIT have developed #SPARKS, an AI system built not just to analyze data, but to mimic the entire scientific process. It works as a team of AI agents where one proposes a hypothesis, and another immediately critiques it. 👉 https://lnkd.in/eD5KtuJY This continuous loop of generation and reflection pushes the system to explore ideas beyond its initial training. When turned loose on the complex world of protein science, SPARKS operated autonomously. It generated its own hypothesis about protein stability, designed the computational experiments to test it, refined the process as the data came in, and wrote a report on its findings. In doing so, it uncovered a previously unknown "frustration zone" in certain protein structures. This work demonstrates a new potential for AI in science, where the system acts less like an assistant and more like an independent investigator. It makes you wonder: what happens to the pace of discovery when our tools can not only help find answers but also formulate the questions? Read the full research review at https://lnkd.in/eD5KtuJY Research from Markus J. Buehler and Alireza Ghafarollahi at Massachusetts Institute of Technology #ScientificDiscovery #ArtificialIntelligence #ComputationalBiology #AIinScience #Research

AI leaders often predict that health and medicine will soon advance as rapidly as computing. They point to breakthroughs in text, images, and code, and suggest biology is next, that cures and life-extending therapies are just around the corner. Until recently, I thought that was naïve. I was sure medicine wouldn’t follow Moore’s Law, and I felt discouraged that AI couldn’t help unravel the biology behind ataxia telangiectasia (A-T), the rare disease that affects two of my sons, causing neurodegeneration, immune deficiency, lung problems, and cancer. I desperately wanted AI’s help in accelerating drug discovery for A-T, but it still felt out of reach. Computing progressed predictably because engineers could shrink transistors in steady, repeatable ways. Biology, by contrast, is messy: genes, proteins, and cells interact in tangled webs of redundancy and feedback. Push on one pathway, and ten others compensate. Instead of scaling curves, you find complexity piled on complexity. AI’s success in language relied on trillions of words freely available online. Biomedical data, in contrast, is scattered, biased, and often too small to train robust models. Genomic and proteomic data are enormous but largely descriptive, pointing to correlations rather than causes. Even when algorithms flag a target, validating it biologically can take years. Biology isn’t just information; it’s chemistry, physics, and evolution, all at once. I used these points to argue that Silicon Valley’s optimism needed a dose of reality. Medicine would progress, but not on the tech industry’s timetable. But I may have been wrong. Emerging AI models are now revealing possibilities I hadn’t imagined. TranscriptFormer can organize millions of single-cell measurements into a universal cross-species map of cell types. HEIST incorporates spatial and proteomic context. GeneMamba scales efficiently to tens of millions of cells. scCross integrates multiple omic layers. CellSymphony fuses gene expression with tissue morphology. These approaches suggest AI could accelerate biology in profound ways. And for A-T families like mine, that acceleration means hope. For example, many of these AI models can now test interventions in silico, predicting which proteins, if modulated, would shift cells from a diseased to a healthy state. For A-T, they could predict (without doing any laboratory experiments!) whether targeting a specific protein would make A-T cells function like those from unaffected individuals. We still need to generate far larger genomic, transcriptomic, epigenetic, methylation, spatial, proteomic, cellular, electrophysiological, and circuit-level datasets for AI models to train on. But now, I think I can glimpse AI-assisted breakthroughs on the horizon. Please let the A-T Children’s Project know if you’re applying AI to any facet of A-T. Thanks. #AI #ataxiatelangiectasia #GenerativeAI