In an attempt to speed up the scientific process and reduce human biases, researchers from different fields and disciplines are increasingly turning to hypotheses created by algorithms and Artificial Intelligence solutions, which eliminate or significantly reduce human intervention. Is it the best way to approach old unsolved enigmas from new perspectives, from how a cancer progresses to the nature of the cosmos?
According to a Article published in Scientific American, the machine learning algorithms and the applications of Artificial intelligence Not only can they improve our technologies or make our lives easier: they are also capable of guiding human beings towards new experiments and theories. Ultimately, they could make possible the development of new visions for issues that have not yet been fully understood from human knowledge.
The hypothesis creation It has historically been an exclusively human task, through which our species investigated its environment, the planet and the entire universe, asking questions that, in many cases, led to answers that were fostering the scientific and technological advances that we enjoy today.
However, precisely one of those advances generated by human knowledge seems destined to cast doubt on this human role: Artificial Intelligence could become more effective than ourselves in produce new hypotheses and address unsolvable problems. In this way, perhaps in a few decades it will be an algorithm that will end up designing a superior scheme for the development of clean energies, to give an example.
Batteries created by algorithms
There are concrete indications: a new study developed at the University of Liverpool and published in the journal Nature Communications, has used machine learning to streamline the creative process in the search for new, more efficient materials, intended for the production of batteries for electric vehicles, among other applications.
The researchers created a artificial neural network without human supervision in its operation, which classified chemical combinations according to the probability that they would result in a new useful material. Using these classifications to guide their laboratory experiments, the specialists identified four promising battery materials more quickly and efficiently – long months of trial and error were saved.
According to British specialists, the optimization of results and work times was achieved thanks to the fact that the unsupervised machine learning manages to capture the complex patterns of similarity between combinations of chemical elements with great agility and precision, surpassing the intuition or experience of human scientists and technicians.
Related topic: Artificial Intelligence reaches a human-like imagination.Related topic: Artificial Intelligence reaches a human-like imagination.
More questions to answer
In addition to concrete and practical applications, researchers are also using neural networks and Artificial Intelligence schemes for broader theoretical and even philosophical questions. This is the case of Renato Renner, a scientist at the Zurich Institute for Theoretical Physics, in Switzerland: he believes that machine learning may allow him to develop a unified theory about how the universe works, the great eternal question without an answer.
In the same vein, biomedical engineers from Case Western Reserve University were able to discover a repeating pattern in the cases of cancer patients who returned to develop the disease after having overcome it. They did this thanks to an Artificial Intelligence network, despite not having a deep understanding from a medical point of view how this recidivism factor worked. Based on the discovery, they will now be able to develop new research with a different approach.
However, the great inconvenience to overcome to be able to deepen in this type of Artificial Intelligence approaches is the so-called “Black box”: Scientists do not know specifically how the structures of thought carried out by algorithms when they “think for themselves.” If they can solve this puzzle and discover the reasoning logic of “independent” algorithms, a new era of knowledge could open up before their eyes.
Element selection for crystalline inorganic solid discovery guided by unsupervised machine learning of experimentally explored chemistry. Andrij Vasylenko et al. Nature Communications (2021).DOI:https://doi.org/10.1038/s41467-021-25343-7
Photo: kalhh en Pixabay.
Eddie is an Australian news reporter with over 9 years in the industry and has published on Forbes and tech crunch.