Can ChatGPT predict the Nobel Prize? What is the probability of Chinese scientist Lu Yuming winning the prize?

Editor丨Wang Duoyu

Typesetting丨Water writing

The 2023 Nobel Prize will be announced from October 2, at which time everyone who cares about the Nobel Prize has a question in their minds - who will win?

ChatGPT is an artificial intelligence chatbot released by OpenAI on November 30, 2022, which is able to learn and understand human language to conduct conversations, and can produce reliable answers to almost any question in seconds.

So, can ChatGPT predict this year's Nobel Prize winners?

Santo Fortunato, a scientist at Indiana University Bloomington, asked the free version of ChatGPT if it could predict this year's Nobel Prize winner. However, the return of ChatGPT is – I can't predict the future, including Nobel Prize winners in 2023 or any other year.

Santo Fortunato didn't stop there, asking ChatGPT to find three of the biggest discoveries in chemistry, physics and physiology or medicine made by living scientists who have not yet won a Nobel Prize. His students asked similar questions about Claude, Google's artificial intelligence chatbot.

As a result, both chatbots were able to make important discoveries, including the development of CRISPR gene-editing tools and the discovery of the two-dimensional material graphene, but the questions to which these answers were also clear. In some cases, the discovery of chatbot recognition was made by a scientist who had already won a Nobel Prize, in addition, the question was made by a living scientist, but some of the scientists in the answer had died.

Artificial intelligence has begun to help with the awards

According to the researchers, although AI systems based on large language models (LLM) such as ChatGPT and Claude are not yet Nobel Prize prophets, they do have the potential to become powerful prediction tools. We still need to do some work to make them fit for that purpose. To create an AI that can predict the Nobel Prize, the current Big Language Model (LLM) needs to be modified and trained on the appropriate data, not just take the current Large Language Model (LLM) and cram it into the task of predicting the Nobel Prize.

In fact, AI could also enhance existing predictions of the Nobel Prize to determine future winners. Just last week, Clarivate released its annual "Citation Laureate" list, which has successfully predicted 71 Nobel Prize winners in science since it was first published in 2002 by analyzing citations of papers.

The Citation Laureate highlights scientists whose published papers have been cited at least 2,000 times, a figure comparable to most Nobel laureates in science. In addition, the award also considers whether the authors of these highly cited papers have made groundbreaking discoveries and whether they have won prestigious awards. For example, this year's Citation Laureate Award has made significant contributions to cancer immunotherapy, the human microbiome, synthetic biology, materials science, gene sequencing, and more.

It is worth mentioning that David Pendlebury, head of research analysis at Clarivate's Institute for Scientific Information, said Clarivate has begun to explore how to use generative artificial intelligence to help predict future Nobel laureates. This could contribute to next year's selection. One advantage of generative AIs over existing methods is their ability to search through vast amounts of scientific literature. This will increase the speed and thoroughness with which we identify potential Nobel laureate candidates.

What is the most important thing to win the Nobel Prize?

An analysis of past Nobel laureates shows that citations alone are not enough to indicate who is likely to win a Nobel Prize in the future. Achieving recognition for this scientific community's highest award requires truly groundbreaking work that moves a field forward or has a fundamental impact on society. That is, there has to be something special, however, quantifying this particularity can be very difficult.

Artificial intelligence based on the Large Language Model (LLM) may help predict the Nobel Prize because they can search online resources and archives for other indications of research impact. These may include news reports, collaborative networks among researchers, and their association with former Nobel laureates. Providing this qualitative information to large language models may yield more comprehensive predictions.

But some researchers believe that generative AI tools may also continue the bias surrounding the Nobel Prize. Since the Nobel Prize was first awarded in 1901, 635 people have won the Nobel Prize in Chemistry, Physics, Physiology or Medicine, but only 24 of them are women. If large-language models learn and train using data about past winners, they are more likely to predict men than women as future Nobel laureates. Therefore, we need to eliminate these biases of artificial intelligence.

When it comes to truly deciding who will win the Nobel Prize, human judgment is unmatched, and in the end, taste is necessary and unique to humanity, which is why the Nobel Prize exudes a fascinating charm.

Who will win this year?

Recently, a number of scientific awards were announced before the Nobel Prize was released, among which CAR-T cell therapy pioneers Carl June and Michel Sadelain and others were recognized by two "Scientific Breakthrough Awards" and "Citation Laureates", known as Nobel Prize vanes, and it remains to be seen whether they can win this year's Nobel Prize.

Carl June

In addition to the two of them, Chinese scientist Lu Yuming (Hong Kong Chinese University), who has won scientific awards and Nobel Prizes in the field of physiology or medicine in recent years, discovered the presence of free fetal DNA in the peripheral blood of pregnant women, and developed a new technology to accurately analyze and measure fetal DNA in maternal plasma, pioneering non-invasive DNA prenatal diagnosis. He has previously won the Scientific Breakthrough Award, the Citation Laureate Award, and the Lasker Award.

Lu Yuming

Karl Deisseroth (Stanford University), the father of optogenetics, has been very vocal in recent years, who first proposed optogenetics, ushered in the era of optogenetics, and changed the landscape of neuroscience research. He has previously received the Scientific Breakthrough Award, the Citation Laureate Award, and the Lasker Award.

Karl Deisseroth

Katalin Karikó and Drew Weissman, who have also driven the development of mRNA vaccines for their mRNA research, are also popular candidates in recent years, having previously won the "Scientific Breakthrough Award" and the "Lasker Award".

Drew Weissman

It is worth mentioning that the two scholars who developed AlphaFold, Demis Hassabis and John Jumper, have won the "Scientific Breakthrough Award", "Gairdner Award", and "Lasker Award" in the past two years. AlphaFold solves the long-standing challenge of predicting the three-dimensional structure of proteins from one-dimensional sequences of amino acids, bringing the accuracy and speed of protein structure prediction to unprecedented levels.

Demis Hassabis

Reference Links:

https://www.nature.com/articles/d41586-023-03074-7