AI in Science: The Future is Here, But Are We Ready?

As a scientist by training, I’ve spent my entire career working with other scientists, and I’ve noticed we often fall into two camps when it comes to new technology:

• On one side, there are those of us who get excited about the latest shiny tool and want to try it out immediately (guilty as charged!). Sometimes, we dive in headfirst without applying the critical scientific thinking we’ve been trained in.
• And on the other side, there are the skeptics—those who are cautious, even resistant, using “double the usual energy” to critically evaluate or dismiss something because “if there’s too much hype, it can’t be good.”

During my PhD back in the early 2000s, AI tools like AlphaFold didn’t exist, but we already had predictive technologies to estimate protein structures by comparing them to existing ones from other species. I remember using one of these tools early in my project and getting a decent prediction in just 5 minutes. My supervisor, one of those skeptics, insisted that “there is nothing like an experimental result”. So, I spent the next four years crystallizing my protein and solving its 3D structure experimentally—only to find that the initial prediction was, in fact, very accurate.

That experience taught me a key lesson: Experimentation is still the gold standard, but we shouldn’t discount new technologies just because they’re new or overhyped. As scientists, we’re trained to evaluate everything critically—whether it’s a cutting-edge tool or a time-tested method. It’s not about following the hype or rejecting it outright; it’s about using our scientific judgment to assess what works, what needs more development, and what’s simply noise.

This divide in mindset isn’t just theoretical. I remember speaking with a former colleague—a director of strategy (with a PhD)—months after ChatGPT had exploded onto the scene. His response shocked me: “I haven’t tried it, and I’m not interested.” As scientists, aren’t we supposed to be curious? We have a responsibility to at least try and evaluate these tools with a structured, scientific method, regardless of the hype.

Yes, I admit I’m biased towards new tech—my love for sci-fi might have something to do with that 😉—but I also believe in stepping back and critically assessing each tool. Not all tech will deliver on its promise today, but many hold the potential to transform our work tomorrow.

And let’s be clear: AI is already transforming science. Dashun Wang and Jian Gao at the Center for Science of Science and Innovation (CSSI) recently released a study, analyzing millions of academic papers to quantify AI’s current impact on scientific disciplines and predict its future benefits - it showed AI’s influence in nearly every scientific discipline. But there’s a catch — many disciplines are not ready to fully leverage AI because they lack the proper training. If we don’t address this gap, we risk missing out on the full potential of these tools.

Since 2015, AI’s influence has rapidly spread across nearly every scientific field, from biology to physics. One of their key findings is the “citation premium”—papers using AI techniques tend to have a greater impact, receiving more citations than their non-AI counterparts. This shows that AI-driven research is becoming more influential among scientific peers, contributing to advancements across disciplines.

However, Wang and Gao’s research also reveals a crucial misalignment. Many fields that could benefit greatly from AI are not equipped with the necessary training to fully leverage its potential. There’s a stark gap between the rapid rise of AI capabilities and the readiness of scientists in various fields to use these tools effectively. While fields like computer science, mathematics, and engineering are well-prepared, disciplines like sociology, biology, and even economics lag behind in AI education. The researchers highlight that this “supply and demand” gap in AI talent could prevent scientists in these fields from taking full advantage of AI’s advancements.

Another concerning finding is that fields with a higher proportion of women and underrepresented minority researchers seem to be benefiting less from AI. For instance, disciplines like sociology, where diversity is greater, show fewer AI-driven breakthroughs compared to fields like physics, where AI is making significant inroads. This suggests that the benefits of AI are not equally distributed, raising questions about how to ensure equitable access to AI-driven research.

Their conclusion is clear: while AI holds immense promise for the future of science, its potential will only be realized if we address these training gaps and provide more interdisciplinary collaboration. AI can accelerate discoveries, but only if scientists are given the tools and opportunities to engage with it meaningfully.

Wang and Gao’s study was included in a larger report to the National Academy of Sciences, urging policymakers to take action. They suggest that increasing funding for AI training and fostering collaboration between AI specialists and researchers in other fields are crucial steps toward unlocking AI’s full potential in science.

So, my message is this: Whether you’re in the “overly excited” camp like me or the “super skeptical” camp, we all need to engage with new technologies like AI. The future of science depends on it - Do I think AI will change science forever? Absolutely—it’s already happening.

Reference

"AI Is Revolutionizing Science. Are Scientists Ready?" -Kellogg Insight, Oct 11, 2024. Jian Gao and Dashun Wang - Link