AI isn’t smarter than us, it’s just less biased


According to researchers at Rutgers University, it appears that machines have already beaten humanity in at least one science subject.
Professor Vikas Nanda of the Department of Biochemistry and Molecular Biology at Rutgers University in New Jersey has spent over two decades meticulously studying the intricate nature of proteins, the highly complex substances found in all living organisms. He has dedicated his professional life to contemplating and understanding the unique patterns of the amino acids that make up proteins and that determine their transformation into hemoglobin, collagen, etc. In particular, the professor’s studies focused on self-assembly of proteins because many scientists believe that fully understanding this process could lead to many breakthrough products for medical and industrial uses (such as artificial human tissue for wounds or catalysts for new chemicals). When the study authors set out to conduct an experiment that pitted a human — a person with a deep intuitive understanding of protein design and self-assembly — against the predictive abilities of an artificial intelligence computer program, they singled out the Professor as best candidate. The researchers wanted to see which of the two could do a better job of predicting which protein sequences would combine more successfully.


“Despite our extensive experience, AI has performed better or equal on different datasets, demonstrating the enormous potential of machine learning in overcoming human biases ,” Nanda says, in a statement from the university.

During the test, Prof. Nanda and five other colleagues were given a list of proteins and had to predict which ones might self-assemble. The computer program made the same predictions, and the researchers compared human and machine responses.

The human participants made their predictions based on their previous experimental observations about proteins, such as patterns of electrical charges and degree of aversion to water. In the end, Prof. Nanda and colleagues predicted that eleven proteins would self-assemble. The computer program, on the other hand, thanks to an advanced machine learning system, chose nine . Human experts turned out to be correct for six of the eleven proteins chosen. The computer program achieved a higher accuracy rate , with six of the nine proteins chosen actually able to self-assemble.
The study authors explain that human participants tended to “ privilege” some amino acids over others, which has led to incorrect predictions. The AI ​​program also correctly identified some proteins that weren’t “obvious choices” for self-assembly (among other things, opening the door for further research). Prof. Nanda admits that he once doubted machine learning for protein assembly investigations, but now he is much more open to the technique.

  • Machine learning overcomes human bias in the discovery of self-assembling peptides. (


Related Articles

Leave a Reply

Your email address will not be published. Required fields are marked *

Back to top button