-
Asked by Liwia on 18 Jan 2024.
Question: How can you use AI in science?
- Keywords:
-
Andrew Maynard answered on 18 Jan 2024:
Great question! You can use it like a genius partner that helps you see connections between different ideas that you missed. AIs can read thousands of scientific papers, and so they can definitely help find ideas that you and others may have missed.
-
Verna Dankers answered on 19 Jan 2024:
As a computer scientist, my work mostly consists of reading articles, thinking about research questions and hypotheses, lots of programming and training models, and writing articles about my own research. For each of those tasks, AI can be a great assistant: it can help you find relevant work to read from the internet, help you to program faster by autocompleting lines of code, and help you to write better (e.g. spelling correction, etc).
-
Carl Peter Robinson answered on 20 Jan 2024:
This is a great question for several reasons: it’s really exciting to think about how AI can help in all fields of science, and the progress it can help scientists make; it’s also a clever question because you can take it as meaning, “Wait, let’s not just go and throw AI into everything we do and hope for the best. How can we use AI in the best way possible to help us in our scientific fields?” Here are a few ways:
We can use AI tools to help do research and discover patterns in data much quicker than previously, when it was just humans performing laborious experiments and systematically working through sets of tests and recording the results. We can more reliably use unsupervised learning to find correlations in huge datasets, such that we can find important features in that data, or discover smaller clusters of related data, without having to manually label all of that data beforehand.
There are large language models (LLM) and natural language processing (NLP) models that we can use to process huge quantities of research papers and provide scientists with short summaries of these papers about the important aspects of them. Related to this, my lab has developed a system that uses AI to track trends in scientific fields, searching for how often specific target words are used, whether their context is positive or negative, which ones are hot topics, when they start to become very popular, and when they start to fall out of favour.
And, like my colleagues have said, the AI chatbots we have now can be used as awesome research assistants. We can ask them to help us with our research and the code we write to perform our experiments. It’s like we’re on the way to what happens on the TV show “Star Trek: The Next Generation” (one of my favourite shows, by the way!), where Captain Picard asks the AI computer on board his starship stuff like, “Computer! Run a statistical analysis on the nebula gas we collected and look for correlations.” And the computer goes away and seconds later reports back, “Analysis complete. There are seventeen highly significant factors in the data, listed on-screen now.” Maybe by the time you’re a scientist (if you choose to be), you’ll have this level of AI tools to use. How cool is that?!
Comments