Beyond Chatbots: AI as a Lab Assistant and Clinical Colleague
AI in healthcare and science is no longer confined to chat interfaces. In chemistry, researchers at the University of Rochester now prompt large language models (LLMs) with natural‑language descriptions of materials they want to create. The models respond with step‑by‑step experimental “recipes”, which scientists iteratively refine by feeding back lab results. This turns the LLM into a kind of data‑driven lab assistant that helps explore huge design spaces for catalysts and other advanced materials more efficiently. In mental health, a new psychiatry‑specialised LLM called PsychFound is being developed to help clinicians sift through complex notes, diagnostic criteria and treatment evidence. Instead of replacing psychiatrists, it aims to reduce documentation burden and support better‑grounded diagnosis and care planning. Together, these systems hint at the next era of AI: tightly embedded in expert workflows, quietly speeding up decisions rather than replacing professionals outright.
Alzheimer’s Prevention AI and Psychiatry-Specialised Models
For neurodegenerative and mental health conditions, time and context matter. The MIT‑centered FINGERS‑7B foundation model for Alzheimer’s prevention shows how AI can integrate many streams of information at once. Trained on data from tens of thousands of at‑risk individuals, it reads lifestyle, clinical, genomic and proteomic signals together to uncover early multi‑omic biomarkers of preclinical Alzheimer’s. On international WW‑FINGERS datasets, it delivers much more accurate early diagnosis and responder stratification than earlier tools, and is openly accessible via the AD Workbench to researchers worldwide. In parallel, PsychFound demonstrates how domain‑adapted language models can encode psychiatric knowledge, from diagnostic frameworks to therapy guidelines. Such models can summarise long clinical notes, highlight risk factors, and suggest evidence‑based options, offering potential value for Malaysian mental health services where specialists are scarce. Both systems illustrate how Alzheimer’s prevention AI and psychiatry‑specific LLMs can enhance—not replace—human judgment.
AI Drug Discovery and Lung Cancer Biomarkers
Lung cancer remains one of oncology’s toughest challenges, but AI drug discovery is changing the landscape. New platforms now achieve 85–95% accuracy in identifying lung cancer biomarkers, the molecular signatures that indicate which treatment a patient is most likely to respond to. This level of precision supports more personalised care, from targeted therapies to immunotherapy combinations tuned to each patient’s tumour profile. AI systems are also optimising drug formulations and delivery methods and improving clinical trial design and recruitment, making it easier to find suitable participants and predict which regimens are worth testing. Global biopharma leaders, including AstraZeneca, Bristol Myers Squibb, Genentech, Pfizer and Merck, are integrating AI across their lung cancer portfolios. For Malaysians, this shift means that future therapies reaching local hospitals are increasingly likely to have been designed, tested and refined with AI—especially in how lung cancer biomarkers are selected and used.
Workplace Safety AI: From Lagging Indicators to Real-Time Prevention
AI is also advancing safety far from hospital wards. CompScience’s Safe Work Plan platform applies AI to workplace risk mitigation, guided by the National Safety Council’s Serious Incident and Fatality (SIF) Prevention Model. Instead of relying mainly on past injury statistics, workers can use their phones to capture images and brief descriptions of their tasks. The system then rapidly identifies tasks, environmental conditions and potential hazards, generates risk scores, and suggests safeguards and controls tailored to the situation. Because the NSC model emphasises preventing serious injuries and fatalities before they occur, this form of workplace safety AI could be particularly relevant to Malaysia’s construction, manufacturing and energy sectors. By embedding AI into daily routines—risk checks before a lift, maintenance or confined‑space entry—companies can move from reactive reporting to proactive prevention, potentially reducing catastrophic incidents while complementing, not replacing, safety professionals on the ground.
Benefits, Risks and What It Means for Malaysia
Across these examples, AI in healthcare and safety brings clear advantages: faster data analysis, more accurate lung cancer biomarkers, earlier Alzheimer’s risk prediction, smarter experiments and real‑time hazard detection. Yet the same tools raise pressing questions for Malaysians. Sensitive health and workplace data must be protected; bias in training datasets could disadvantage certain ethnic or rural communities; and the opaque reasoning of complex models can make clinical decisions harder to explain to patients and families. Regulators and hospitals in Malaysia and the wider region may respond in several ways: requiring transparent validation studies before AI tools guide care, insisting on local data to test performance across diverse populations, and mandating human oversight for high‑stakes decisions. Hospitals could start by piloting decision‑support uses—such as triage, summarisation or safety pre‑checks—while building internal expertise in AI governance, so that innovation proceeds with accountability and public trust.