Demis Hassabis was unaware that he would receive the Nobel Prize in chemistry from the Royal Swedish Academy of Sciences until his wife began receiving a flurry of calls from a Swedish number via Skype.
“She answered the call several times, but they kept calling back,” Hassabis explained during a press conference held to commemorate the award alongside his colleague John Jumper from Google DeepMind. “Eventually, she realized it was a Swedish number and they requested my contact information.”
While the news of his recognition—the highest honor in science—might not have caught him completely off guard, it followed closely on the heels of Geoffrey Hinton, a figure often dubbed one of the “godfathers of AI,” and Princeton University’s John Hopfield receiving the Nobel Prize in physics for their contributions to machine learning just a day prior. “Clearly, the committee aimed to make a statement by awarding the two prizes together,” Hassabis noted in a subsequent press conference after his own win.
To emphasize the point: artificial intelligence has arrived, and we are now at a stage where one can earn a Nobel Prize through research in AI while also impacting various disciplines—be it physics for Hinton and Hopfield or chemistry for Hassabis and Jumper, who shared the honor with David Baker, a genomics expert from the University of Washington.
“It’s certainly a significant moment for ‘AI in science,’” remarks Eleanor Drage, a senior research fellow at the Leverhulme Center for the Future of Intelligence at the University of Cambridge. “With notable computer scientists receiving awards in both chemistry and physics, we’re all curious to see who might be in line for a peace prize,” she adds, noting that her colleagues were humorously speculating about xAI owner Elon Musk potentially being a candidate for that honor.
Drage describes the recognition of AI researchers with physics and chemistry accolades as “a significant debate, not just within those fields but also from an external perspective.” She posits that the prizes could represent one of two developments: either a substantial change in disciplinary boundaries driven by the prevalence of AI in academic research or an indication that “we’re so captivated by computer scientists that we’re willing to place them in any category.”
While she is uncertain about what this week’s decisions imply, she, along with others, believes it will have a substantial impact on the future of research.
“Achieving a Nobel through AI might be a precedent that has already been set, but it will undoubtedly shape future research pathways,” asserts Matt Hodgkinson, an independent specialist in scientific research integrity and a former manager at the UK Research Integrity Office. The key concern is whether this influence will be positive.
Baker, this year’s recipient of the Nobel Prize in Chemistry, has been a prominent figure in the research community focusing on the application of AI in predicting protein structures. For decades, he has diligently tackled this complex issue, making steady progress and understanding that the clear-cut challenges of protein structure serve as an ideal platform for testing AI methodologies. His success is not a mere coincidence—Baker has authored over 600 research papers throughout his career, alongside the ground-breaking AlphaFold2, a project by Google DeepMind that also received accolades from the committee.
However, Hodgkinson expresses concern that the focus of researchers may shift towards the methodology rather than the underlying science, which could hinder genuine understanding. “My hope is that this won’t lead researchers to misuse chatbots by mistakenly assuming that all AI tools are on the same level,” he notes.
This apprehension stems from the surge in interest surrounding emerging technologies that promise significant transformation. “Hype cycles are a common phenomenon, with recent cases being blockchain and graphene,” Hodgkinson points out. Following the discovery of graphene in 2004, there was a notable increase in academic publications—about 45,000 papers were released between 2005 and 2009, as per Google Scholar. Following the Nobel Prize awarded to Andre Geim and Konstantin Novoselov for their work, that number skyrocketed, reaching 454,000 between 2010 and 2014, and surpassing a million from 2015 to 2020. Despite this wave of research, the real-world impact of graphene has been only modest so far.
Hodgkinson believes that the recognition of multiple researchers by the Nobel Prize committee for their contributions in AI could attract more investigators to the field, potentially leading to varying standards of scientific quality. “The question remains whether these AI proposals and applications hold real substance,” he states.
We have already observed how media coverage and public interest in AI have influenced the academic sector. Research conducted by Stanford University indicates that the number of publications related to AI has surged threefold from 2010 to 2022, with nearly a quarter of a million papers published just in 2022—translating to over 660 new publications each day. This surge occurred even before the launch of ChatGPT in November 2022, which ignited the generative AI boom.
Julian Togelius, an associate professor of computer science at New York University’s Tandon School of Engineering, finds the relationship between academics and the factors of media attention, funding, and recognition from Nobel Prize committees somewhat perplexing. “Scientists typically opt for a balance between the easiest path and the maximum reward,” he explains. Given the increasingly competitive landscape of academia, where securing funding is becoming more challenging and is closely tied to job opportunities for researchers, it appears that the allure of a popular topic—which as of this week has the potential to bring a Nobel Prize to high achievers—might be too enticing to ignore.
However, this trend carries the risk of hindering innovative thought. Togelius points out that “extracting more fundamental data from nature and developing new theories that people can comprehend is challenging.” Such endeavors require significant contemplation. Instead, researchers may find it more efficient to conduct AI-driven simulations that align with existing theories and utilize already available data, leading to incremental progress rather than groundbreaking advancements. Togelius anticipates that a new generation of scientists may well resort to this easier route.
Moreover, there’s a concern that overzealous computer scientists, who have contributed to the evolution of AI, might observe that AI-related work is garnering Nobel Prizes in unrelated fields like physics and chemistry. This might prompt them to venture into other domains. “Computer scientists are often seen as intruding into areas they lack expertise in, applying algorithms, and declaring it as progress, sometimes to positive or negative effects,” Togelius remarks. He confesses to having once considered introducing deep learning into another scientific discipline to “enhance” it, but refrained after recognizing his limited knowledge in fields such as physics, biology, or geology.
Hassabis exemplifies the effective use of AI to propel scientific advancement. With a PhD in neuroscience obtained in 2009, he attributes his academic background to the progress made in AI through Google DeepMind. However, he has noted a shift in the industry’s approach to optimizing efficiencies. “Today, [AI] has become more engineering-heavy,” he remarked during his Nobel Prize press conference. “We have a plethora of techniques that we are now refining algorithmically, independent of brain references.”
This shift might influence the nature of research conducted, the individuals who engage in it, their expertise in the field, and the motivations driving their involvement. Instead of scholars who have dedicated their careers to a particular specialty, there may be an increase in research contributions from computer scientists who may lack a deeper connection to the subjects they are exploring.
Nonetheless, this discussion is likely to be overshadowed by the celebrations for Hassabis, Jumper, and their team members acknowledged for their contribution to winning the Nobel Prize this week. “We are nearing the completion of the [AlphaFold3] code, which we aim to release for unrestricted academic use,” he shared earlier today. “From there, we will continue to make further advancements.”
