A new TAU study using toddler vocalizations reveals a fundamental challenge: AI can recognize patterns in sounds, but that doesn’t necessarily mean it understands what those sounds mean


Research
A new TAU study using toddler vocalizations reveals a fundamental challenge: AI can recognize patterns in sounds, but that doesn’t necessarily mean it understands what those sounds mean

In recent years, numerous attempts have been made to use artificial intelligence to decipher the communication of bats, whales, birds, and other animals. However, a new study led by a team of researchers from Tel Aviv University points to a fundamental problem with this approach: AI models focus on the physical properties of a sound, but this does not mean that they understand the meaning attributed to it by animal listening.
According to the researchers, sounds that are acoustically similar do not necessarily carry similar meanings, while sounds that appear different may convey the same information to the receiver. Therefore, classifying sounds according to their acoustic similarity, as is done in most studies, may create a misleading picture of the communication system and the meaning of the messages it conveys.
The study, published in the scientific journal Current Biology, was conducted by Mor Taub, Inbal Arnon, Amiyaal Ilany, Mirjam Knörnschild, Yoav Ram, and Prof. Yossi Yovel. The research team included scientists from Tel Aviv University, the Hebrew University of Jerusalem, the University of Edinburgh, the Museum für Naturkunde – Leibniz Institute for Evolution and Biodiversity Science, and Humboldt-Universität zu Berlin.
To investigate the problem, the researchers used a unique communication system: the vocalizations of human toddlers who have not yet fully developed speech. Unlike animal vocalizations, in this case the researchers can determine, at least to some extent, how the humans to whom the vocalizations are directed interpret them. The recordings included vocalizations made in three contexts: distress, calling to a specific person, the mother or the father and requesting food.
The researchers analyzed the recordings using a classical acoustic method and two deep state-of-the-art neural networks: one trained on animal vocalizations and another trained on adult human speech. The models were asked to group the vocalizations according to their characteristics.
The results showed that the deep neural networks performed better than the classical acoustic method, but even they failed to classify the toddlers’ vocalizations according to their meaning. In some cases, they grouped together vocalizations carrying different messages; in others, they separated different vocalizations intended to convey the same message. The models also failed to identify how a sequence of vocalizations expressed increasing urgency, a distinction that the human ear perceives naturally.
According to the researchers, reliably deciphering animal communication will require combining AI tools with behavioral observations, playback experiments, and sometimes measurements of brain activity. Every species has its own unique perceptual world, and understanding what animals are “saying” therefore requires more than analyzing sound alone: it also requires examining how they hear the sound and respond to it.
Prof. Yossi Yovel concludes: “In recent years, there has been growing excitement about the possibility of using artificial intelligence to decode animal communication, but our study shows that these promises should be treated with caution. Identifying acoustic patterns is not necessarily the same as deciphering meaning: to understand what an animal is ‘saying,’ we need to know how the animal receiving the message perceives it and responds to it. The path toward truly deciphering animal communication will require a combination of AI, behavioral observations, experiments, and research into the nervous system. Artificial intelligence is a powerful tool, but it is no substitute for the perspective of the animal itself.”

Research
A new masking technology could expand the potential of antibody-based therapies for diseases such as Parkinson’s

Over the past few decades, antibody-based therapies have revolutionized modern medicine and are now widely used to treat cancer, autoimmune diseases, inflammatory disorders, and even infectious diseases. Despite their success, however, antibodies have a significant limitation: they struggle to penetrate cells and are therefore largely limited to targeting molecules located on the cell surface or outside the cell. In addition, antibody-based drugs have difficulty crossing the blood-brain barrier, restricting their use in treating diseases such as Parkinson’s and Alzheimer’s. Now, researchers at Tel Aviv University, together with colleagues, have developed a new technology that could overcome these barriers, enabling antibodies to reach the most important targets, those located inside the cell itself.
The technology was developed through a collaboration between research groups from Cornell University in the United States, Tel Aviv University, and the Technion. The study was led by Prof. Chris A. Alabi and Prof. Matthew P. DeLisa from Cornell, collaborating with Prof. Avi Schroder from the Technion and TAU’s Prof. Ben Maoz of the Fleischman Faculty of Engineering and the Sagol School of Neuroscience and Prof. Uri Ashery of the Wise Faculty of Life Sciences and the Sagol School of Neuroscience, together with Prof. Christopher Alabi of Cornell. The findings were published in the Proceedings of the National Academy of Sciences (PNAS).
In the study, the researchers developed an innovative approach based on temporarily “masking” the antibody using a synthetic molecule called SL4. This masking alters the antibody’s chemical properties in a controlled manner, allowing it to be encapsulated in lipid nanoparticles (LNPs), similar to the technology used to develop mRNA vaccines against COVID-19. Once the nanoparticles enter the cell, the antibody is released and regains its original structure and activity.
According to the researchers, this represents a significant breakthrough because approximately 80% of the proteins involved in human disease are located inside cells, making them inaccessible to most antibody-based therapies. The ability to deliver active antibodies into the cell cytoplasm opens new possibilities for treating diseases that have long been considered inaccessible to drug-based interventions.
The study demonstrated that the masking process significantly improves the efficiency with which antibodies can be encapsulated within lipid nanoparticles. Whereas unmodified antibodies were incorporated into the nanoparticles with relatively low efficiency, the masked antibodies achieved substantially higher encapsulation rates while retaining their stability and their ability to recognize the molecular target associated with the disease.
The researchers tested the technology using a series of therapeutic antibodies targeting key biological pathways involved in disease development. The antibodies successfully entered cells and altered important signaling pathways associated with various types of cancer and inflammatory diseases. Following treatment, the activity of these pathways was significantly reduced, indicating that the antibodies had reached their intended targets and remained active inside the cells.
One of the study’s most promising findings emerged from a research model of Parkinson’s disease. The researchers used an antibody targeting alpha-synuclein, a protein whose accumulation in the brain is one of the hallmarks features of the disease. Following delivery of the antibody via the nanoparticles, they observed a significant reduction in the pathological aggregates of the protein in nerve cells, a finding that suggests that the technology could pave the way for new treatments for neurodegenerative diseases.
The technology was also evaluated in a model of acute inflammatory lung injury. The researchers found that delivering antibodies via the lipid nanoparticles reduced inflammatory markers and improved pathological features of lung tissue. These findings highlight the potential for developing targeted therapies for severe inflammatory conditions.
Prof. Ben Maoz said: “For many years, delivering antibodies into cells has been considered one of the greatest challenges in the field of biologic therapies. We have succeeded in developing a system that enables antibodies to cross the cellular barrier and reach targets that were previously beyond their reach. We believe this is an important step toward expanding the therapeutic toolbox of modern medicine and paving the way for more precise treatments for complex diseases that still lack adequate therapeutic solutions.”
The researchers emphasize that the technology is still at the preclinical stage. Nevertheless, they believe the platform could lay the foundation for a new generation of biologic therapies. If it successfully progresses through development and clinical trials, it could, for the first time, enable the widespread use of antibodies against intracellular targets, a goal widely regarded in the pharmaceutical industry as the next frontier of personalized medicine.

Research
A new TAU study of Neandertal and modern human pelvises suggests that the male pelvis evolved a natural shock-absorbing mechanism that may make long-distance walking more efficient.

A new study at the Department of Anatomy and Anthropology of Tel Aviv University, and published in Scientific Reports, offers an explanation for one of the striking differences between men and women: the evolutionary development of the modern human pelvis. By comparing Neandertal pelvises with those of modern humans, the researchers reached a surprising conclusion: the unusual structure of the pelvis may not be that of the Neandertal, as has been assumed for decades, but rather that of the modern human male. According to the researchers, the male pelvis evolved into a unique biomechanical shock absorbing mechanism that stores energy and makes long-distance walking more efficient.
Led by Professor Yoel Rak of the Department of Anatomy and Anthropology at Tel Aviv University, and conducted in collaboration with researchers from Spain, the Technion, Bar-Ilan University, and Ono Academic College, the study is based on a comparison of two nearly complete male Neandertal pelvises, one from Kebara Cave in Israel and the other from the Sima de los Huesos site in Spain, with dozens of modern human pelvises. Surprisingly, despite their large size and robust construction, the Neandertal pelvises were found to resemble those of modern human females in most measurements and proportions rather than those of modern human males.
For many years, the Neandertal pelvis has been regarded as an anatomically unusual structure requiring a functional explanation of its own. The new findings, however, require a new look at Neandertal pelvises. The ancestral configuration may have been retained in Neandertals and in modern human females, while the modern human male pelvis underwent substantial evolutionary modification, giving rise to a distinctive anatomical configuration.
The central finding of the study is that the hip joints of modern human males are positioned farther forward on the pelvic ring than those of both modern human females and male Neandertals. According to the researchers, this shift created a new mechanical system in which the anterior thigh muscles, attached to the front of the pelvis, function much like a spring, while body weight acts on the posterior part of the pelvis .

The Neanderthal pelvis
Professor Rak explains that during every step of bipedal walking, the body's center of mass drops downward. This drop traumatizes the joints and requires energy to raise the body again in preparation for the next step. According to the new model, the distinctive geometry of the male pelvis enables the thigh muscles to cushion the drop of the body's center of mass, store potential energy during the step, and then release that energy immediately afterward- effectively "springing" the body upward into the next step.
In this way, the pelvis functions as a natural shock-absorber and energy-return system. It may reduce energy expenditure, improve walking efficiency, and thereby provide a significant advantage during long-distance travel on foot. The change in the position of the hip joints also required additional structural adaptations, including the thickening of the pubic bone and deepening of the anterior portion of the pelvis to withstand the new mechanical loads.
Modern human females, by contrast, could not adopt the full suite of these modifications. According to the researchers, the constraints imposed by childbirth require a relatively shallow pelvis and a sufficiently wide birth canal. As a result, the female pelvis remains closer to the ancestral configuration—the same general configuration found in male Neandertals.
Professor Ella Been of Ono Academic College, a co-author of the study, adds: “This study demonstrates that questions about human evolution are not confined to the distant past. Understanding the evolution of our walking mechanism can contribute to contemporary research in biomechanics, musculoskeletal medicine, rehabilitation, and injury prevention. The perspective provided by the Neandertals helps us better understand the modern human body.”

A reconstruction of the skeleton
Professor Rak emphasizes that the findings of the study change the way we understand the evolution of the human pelvis. It is not the Neandertal pelvis that is the anomaly requiring explanation. Rather, it is the pelvis of the modern human male. The mechanism that evolved within the human male represents the evolutionary innovation.
The researchers note that the study presents a new biomechanical model that may explain a substantial part of the human pelvis’s sexual dimorphism, the anatomical differences between females and males. The research also demonstrates that even in human macroscopic anatomy, a field that might appear to have been thoroughly explored, there is still potential to uncover previously unrecognized structures, geometries, and mechanisms of biological significance.

Research
A new TAU technology identifies lung cancer’s biological fingerprint with more than 90% accuracy, without DNA sequencing, and could also help monitor patients’ response to treatment.

Researchers at Tel Aviv University have developed a new method for diagnosing lung cancer: a simple, fast, low-cost blood test that does not require DNA sequencing. The method identifies a chemical fingerprint of cancer cells in the blood, by analyzing cell-free DNA originating from those cells. In the study, the test distinguished between lung cancer patients and healthy individuals with a sensitivity of 93.1% and a specificity of 90.3% for patients with stage 2-4 disease.
The study was led by Prof. Yuval Ebenstein of the School of Chemistry at the Faculty of Exact Sciences, the Department of Biomedical Engineering and the Zimin Institute at Tel Aviv University, in collaboration with researchers from JaxBio Technologies, Bnai Zion Medical Center, and Sheba Medical Center. The paper was published in the journal Nature Precision Oncology.
Lung cancer is the leading cause of cancer-related death worldwide. At present, early diagnosis relies primarily on CT scans, but these tests generate a high rate of suspicious findings that ultimately prove to be benign, sometimes leading to unnecessary biopsies and surgeries. At the same time, existing liquid biopsies are generally based on DNA sequencing, a costly and complex process requiring advanced computational infrastructures.
The new method bypasses the need for DNA sequencing. After extracting cell-free DNA from a blood sample, the researchers label it with a light-emitting marker and bind it to a DNA chip they have developed. The chip is then scanned with an optical scanner, and the resulting light patterns are analyzed, enabling rapid identification of the biological fingerprint of lung cancer.

Conceptual illustration of a man with lung cancer
The study included 103 participants: 51 lung cancer patients and 52 healthy control subjects. Following a model-training phase, the researchers developed a signature of 170 genomic regions, and tested it on a separate validation cohort using blinded analysis, achieving high diagnostic accuracy. In addition, they were able to distinguish between the two main subtypes of lung cancer - adenocarcinoma and squamous cell carcinoma - based on distinct DNA signatures.

Illustration of the biological “fingerprint” test — a lung cancer patient on the right and an individual without cancer on the left
Beyond diagnosis, the researchers also examined the novel test's potential for monitoring patients' response to treatment. Among the patients evaluated, changes in the DNA's chemical fingerprint corresponded to imaging findings: in patients who responded to treatment, the chemical fingerprint shifted toward the profile of healthy individuals, whereas no significant change was observed in patients who did not respond to treatment. The researchers emphasize that this is only a preliminary finding and that large-scale studies are needed to confirm the method's monitoring capabilities.
According to the researchers, the technology's main advantage lies in combining simplicity, low cost, and speed. At present, the test can be completed within two to three days at a cost of approximately $60 per sample. They hope that in the future it will serve to complement imaging tests, assist in the early diagnosis of lung cancer, and enable more effective monitoring of treatment effectiveness.
Prof. Ebenstein concludes: "Our goal is to make blood tests for cancer diagnosis more accessible, simpler, and less expensive without compromising accuracy. We have developed a new approach that does not require genetic sequencing but instead identifies the tumor's chemical 'fingerprint' with light, using a technology that can be implemented in standard clinical laboratories. This is a significant step toward developing a tool that can complement imaging tests and help physicians diagnose lung cancer and monitor treatment effectiveness."
Watch to learn more: