Can artificial intelligence predict who will win the Nobel Prize?

ChatGPT can now work integrated with many applications. In addition to companies turning to generative AI for search engines and browsers, Santo Fortunato from Indiana University asked ChatGPT who might win the Nobel Prize.

IT CANNOT PREDICT THE NAMES WHO WILL WIN THE NOBEL PRIZE

Fortunato, who directed his questions to ChatGPT before the Nobel Prizes began to be awarded, did not get the answers he wanted, especially on the Nobel side. 

Asking ChatGPT who would win the awards, Fortunato received the answer, ''I cannot predict the future, including the Nobel Prize winners of 2023 or other years.''

The researcher, who tried to ask questions in different ways, asked the artificial intelligence to identify three major discoveries in the fields of chemistry, physics, and physiology or medicine—the three categories of science Nobel Prizes—to identify living scientists who have not yet won a Nobel Prize. His students also made a similar request and tried Google's AI chatbot, Claude.

Although both chatbots were successful in selecting important discoveries, from the development of the genome editing tool called CRISPR to the discovery of the 2D material called graphene, there were many flaws in their answers. In some cases, they identified discoveries for which scientists had already won a Nobel Prize. And that was not the only problem. Fortunato says, "Even though I asked for living scientists, it gave examples of those who have passed away."

SPECIAL EFFORT IS REQUIRED FOR AN AI THAT PREDICTS NOBELS

Although large language models (LLMs), such as ChatGPT and Claude, are not Nobel Prize predictors in their current form, James Evans, a social scientist at the University of Chicago in Illinois, states that they have the potential to be powerful predictive tools. However, he adds that it will take some work to make them fit for purpose. To create an AI that predicts Nobels, existing LLMs must be modified and trained with appropriate data. "We will need to do more than just take someone else's language model and force it into this task."

Last week, the global analytics company Clarivate released its annual list called 'citation laureates,' which has successfully predicted more than 70 Nobel laureates in the last two years, to predict future Nobel Prizes.

These predictions were made largely by analyzing citations.

The list highlights researchers who have published papers that have been cited at least 2,000 times. Clarivate's analysis also considers whether the authors of these highly cited papers have made groundbreaking discoveries and won significant awards. Researchers on this year's list have made major contributions to fields such as cancer immunotherapy, synthetic biology, and materials science.

Clarivate had begun its research to make predictions for the Nobel Prize.

Artificial intelligence needs to be trained more to make these types of predictions. Stating that citations in scientific studies may not be sufficient to 'predict the person who will win the Nobel,' physicist Rasmus Bjork from the Technical University of Denmark says that for researchers to win a Nobel Prize, they must produce work that moves a field forward, has an impact on society, or is groundbreaking, "Something has to have that special something. Of course, it can be difficult to quantify that special something."

There are also challenges in developing an AI that will predict the scientists who will win the award. This language model may be 'biased' due to the influence of its developer or the datasets on which the language model is trained, and this may ultimately lead to the perpetuation of errors.

This is just like facial recognition in the US to catch criminals s