In our lab, we’re not just observing the AI revolution—we’re living it. We’re actively integrating AI, especially Large Language Models (LLMs), to fundamentally change how we do science. This isn’t about replacing your expertise; it’s about giving you superpowers to be more efficient, insightful, and impactful in your research, all while upholding the highest standards of research integrity.
Our AI Compass: Agency & Ownership
Think of AI as a brilliant, tireless assistant. But remember two golden rules that are paramount to legitimate and responsible use:
- You’re in Charge (Maintain Agency): You are the scientist, the thinker, the decision-maker. AI amplifies your work; it doesn’t do it for you. You must always retain ultimate control and intellectual responsibility for your content and conclusions.
- You Own It (Assume Ownership): Any error the AI makes is ultimately your responsibility. You must always critically double-check, verify, and ensure every single piece of AI-generated information is accurate, unbiased, and sound. Your reputation, and the integrity of our lab’s work, depends entirely on your rigorous supervision of all AI outputs.
Your AI Toolkit
The AI world moves incredibly fast, but our core tools right now are ChatGPT, Gemini, and NotebookLM. NotebookLM is fantastic because it sticks strictly to your uploaded documents, making its answers incredibly precise and reliable for source-grounded tasks. We’re also keeping a close eye on Microsoft Copilot—it’s set to become a game-changer, integrating AI directly into the Microsoft apps we use every day.
AI in Action: Making Research Easier (Responsibly)
Here’s how we’re putting AI to work across our lab, always with an eye on data quality and ethical use:
- Smarter Writing: Beyond fixing typos, LLMs are like having a personal editor. Draft your text, then ask for specific feedback on clarity, coherence, or even if your manuscript rebuttal hits the mark. This helps you refine your communication without compromising your original thought.
- Rapid Literature Reviews: Say goodbye to endless scrolling! LLMs can quickly identify, summarize, and analyze stacks of academic papers, helping you get up to speed on new fields or dig deeper into existing ones in record time. Crucially, always cross-reference with traditional databases like PubMed or Google Scholar to avoid biases, and meticulously verify all sources!
- Boosting Your Code: For coding, LLMs are becoming indispensable. They can help generate code snippets, debug tricky scripts, and even act as instant manuals. This accelerates development, but always review and test generated code for accuracy and efficiency.
- Accelerating Quantitative Work: We’ve seen LLMs condense weeks of complex mathematical calculations into hours. For physics, they’re like a super-smart search engine for equations or specific components for engineering tools.
- Data Analysis (with a strict caveat): While we use specialized deep learning for image analysis, we do not use, and strongly discourage the use of, LLMs for raw data processing, statistical computations, or drawing direct scientific conclusions from datasets. LLMs can “hallucinate” with numerical data. Use them to understand methods or debug code, but perform actual analyses with dedicated statistical software.
The New Super Skill: Prompt Engineering
This is key to effective AI use. Prompt engineering is essentially how you “talk” to the AI to get exactly what you need. It’s about being clear, specific, and even telling the AI what “persona” to adopt (e.g., “Act like an expert editor”). Master this, and you’ll get far more accurate and professional results, drastically reducing AI “hallucinations” or overly flattering responses.
Remember: Humans First!
While AI is amazing, it’s a tool, not a replacement for people. Your interactions with colleagues, mentors, and peers are vital. Real human collaboration is irreplaceable for critical thinking, nuanced feedback, and true scientific breakthroughs. Scientific integrity is a human endeavour, supported by AI, never superseded by it.
Surprise! This blog post was generated with Gemini, using the original webpage I wrote to inform my colleagues about the use of AI in the lab.
