Artificial intelligence (AI) is a complex and often exaggerated concept that seeks to replicate human cognitive abilities for problem-solving and task completion. AI involves various computational methods, including training machine learning algorithms with datasets to recognize patterns and make predictions. However, some AI applications, like those used in hiring, policing, and loan approvals, can perpetuate biases present in historical data, leading to discriminatory outcomes.
While AI has revolutionized scientific research, with a significant increase in AI-based publications since 2015, concerns have been raised about the opaque nature of many AI tools hindering reproducibility in research. Despite these challenges, AI has proven to be a powerful tool in uncovering new discoveries, such as in the study of animal communication and archaeology.
In one groundbreaking example, AI has assisted researchers in deciphering the unique clicks produced by sperm whales in the Caribbean Sea. By analyzing thousands of recorded click sequences using machine learning, scientists have identified distinct patterns in the whales’ vocalizations, resembling phonetics in human language. This breakthrough could potentially pave the way for human communication with these marine creatures.
Although AI has enabled the identification of various rhythmic and temporal patterns in whale communication, the ultimate meaning behind these signals remains a mystery. Further interactive experiments with whales are proposed as the next step in understanding the complex language of these majestic animals.
Studying whale behavior may play a crucial role in understanding the patterns of sperm whale click sequences. This approach could potentially be useful in analyzing vocalizations of various animals, as Dr. Brenda McCowan, a professor at the University of California Davis School of Veterinary Medicine, mentioned to CNN. She was not directly involved in the study.
Moving to land, archaeologists have long been exploring the ancient lines and symbols etched into Peru’s Nazca Desert. The use of artificial intelligence has accelerated the search for these mysterious geoglyphs, which include geometric designs, humanlike figures, and even an orca holding a knife. Masato Sakai, an archaeology professor at Yamagata University in Japan, led a team that trained an object detection AI model with detailed imagery of the Nazca symbols. The study involved researchers from IBM as well.
The team tested the model’s accuracy in the desert area between September 2022 and February 2023, exploring sites on foot and using drones. They were able to verify 303 geoglyphs, significantly increasing the known count within a short period. Although the AI model generated a large number of potential locations, the team identified 1,309 sites deemed to have high potential after screening and ranking suggestions.
Despite the model’s imperfections, AI has the potential to revolutionize archaeology, especially in challenging environments like deserts. Amina Jambajantsan, a researcher and data scientist at the Max Planck Institute, explained that AI models could greatly benefit archaeology in these remote terrains, albeit with some limitations.
Moreover, AI models are aiding researchers in comprehending the intricate structures of proteins, the fundamental building blocks of life. Proteins, composed of amino acids and capable of intricate folding patterns, play essential roles in biological functions such as DNA replication and oxygen transport. The development of AI tools like the AlphaFold Protein Structure Database has revolutionized the study of protein structures, providing a more efficient alternative to traditional lab experiments and X-ray crystallography techniques.
DeepMind, located in London, has made a groundbreaking advancement by successfully predicting the structures of almost 200 million known proteins based on their amino acid sequences. Utilizing a database trained on a comprehensive collection of known amino acid sequences and experimentally determined protein structures, this innovative tool serves as a powerful resource likened to a “Google search” for protein models. With just a simple click, researchers gain immediate access to predicted protein models, significantly expediting advancements in fundamental biology and various related fields, including medicine. This tool, which has already been leveraged by over 2 million researchers globally, has been hailed as a major breakthrough in physical chemistry.
Anna Wedell, a distinguished professor of medical genetics at Karolinska Institutet in Sweden and a member of the esteemed Royal Swedish Academy of Sciences, expressed her admiration for the pioneering work of DeepMind’s founders, Demis Hassabis and Thomas Jumper. The duo was honored with the prestigious 2024 Nobel Prize in chemistry for their exceptional contributions to the field. Wedell lauded the significance of this achievement, describing it as a significant milestone in tackling a longstanding challenge in the realm of physical chemistry.
Despite its remarkable capabilities, the tool does have certain limitations. Notably, efforts to apply AlphaFold, as the tool is known, to proteins with mutated sequences, such as those associated with early breast cancer, have revealed that the software currently lacks the capacity to accurately predict the outcomes of new mutations in proteins. This underscores the need for ongoing refinement and development in the field of protein structure prediction.
AlphaFold stands out as a prominent example of the growing utilization of artificial intelligence (AI) tools in various biomedical disciplines. Machine learning technologies are playing a crucial role in accelerating endeavors aimed at constructing a comprehensive atlas detailing every cell type within the human body. Furthermore, AI is proving instrumental in the discovery of novel molecules that have the potential to become groundbreaking new drugs, including a class of antibiotics that exhibit promise in combatting drug-resistant bacteria strains.
