Artificial intelligence has long ceased to be just a tool for automating routine tasks or generating texts. Today, it is actively being integrated into fundamental scientific research: from creating protein structures for drug development to predicting climate anomalies. Machine learning technologies have already forced the scientific community to rethink a number of seemingly indisputable truths.
Here are five key scientific stereotypes that have been disproven or significantly revised thanks to the application of AI.
1. Computer Vision: Not Just the Prerogative of Living Beings
For a long time, the prevailing opinion in the scientific community was that the ability not just to "see" but also to recognize patterns was exclusively available to living organisms. The situation changed in 2012 when the AlexNet neural network was presented at the NeurIPS conference. Trained on 1.3 million high-resolution images from the ImageNet database, it made a breakthrough in computer vision, proving that machines can classify visual data with high accuracy. This work laid the foundation for modern technologies such as facial recognition for smartphone unlocking and showed that visual perception is not a unique biological function.
2. Logic and Mathematics: AI at the Level of an Olympic Champion
Complex logical deductions and theorem proofs were traditionally considered a test of exceptional human abilities, a kind of "filter" for finding young mathematical talents. However, in January 2024, a team of Google researchers presented the AlphaGeometry system, which solved 25 out of 30 complex geometry problems within standard Olympic competition time. For comparison, previous algorithms could only handle 10 problems, and the average score of a gold medalist at the International Mathematical Olympiad among humans is 25.9 points. This achievement, published in the reputable journal Nature, proved that AI is capable of deep logical reasoning.
3. Botany: Climate Does Not Always Determine Leaf Size
In biology, there was an established rule that plants in humid climates (e.g., in jungles) have larger leaves than the same species in arid regions. However, a study by Australian scientists, conducted using computer vision, disproved this axiom at the species level. After analyzing thousands of digitized specimens from the National Herbarium of New South Wales, specialists found that in Norway maple, leaf size depends more on genetic exchange with other populations than on temperature or precipitation. This discovery changes approaches to conserving rare plant species in the context of global climate change.
4. Evolutionary Biology: Extinction Is Not Always Followed by Flourishing
It is commonly believed that after mass extinctions (as in the case of dinosaurs), the vacated ecological niches are quickly filled by new species, inevitably leading to a biological boom. In 2020, British and American researchers applied machine learning to analyze over 1 million fossil descriptions covering 170,000 species. The results showed no direct causal link between mass extinction and the subsequent explosive growth of biodiversity, which required a serious reassessment of one of the key evolutionary theories.
5. Forensics: Fingerprints of One Person Can Be Similar
The belief in the absolute uniqueness of each person's fingerprints (and even different fingers on the same person's hands) is the basis of modern biometrics and forensics. However, in January 2024, scientists from Columbia University called this into question. By training an AI model on 60,000 real and 500,000 synthetic fingerprints, they discovered that fingerprints from different fingers of the same person show significant similarities. Unlike forensic experts who focus on ridge bifurcations, the neural network identified new, previously unconsidered markers: the curvature and inclination of the swirls. This discovery could improve the accuracy of forensic examination by almost two orders of magnitude and simplify biometric authentication in future gadgets.
Conclusion: The Future of Scientific Inquiry
Artificial intelligence does not replace scientists, but becomes their most powerful ally. It can process data arrays that are insurmountable for human perception and formulate non-trivial hypotheses about cell responses or vaccine efficacy. Although current models are still subject to biases embedded in training datasets, the trajectory of technological development indicates that in the coming decades, the most significant scientific discoveries will be made in tandem by humans and artificial intelligence.
