Recently, a Swedish research team has developed an artificial intelligence (AI) model based on machine learning that can assist experts in early screening for autism spectrum disorder (ASD). ASD is commonly referred to as Autism Spectrum Disorder or Autism Spectrum Disorder. The research results show that the accuracy of the system can reach about 80%. According to the World Health Organization, approximately 1 in every 100 children worldwide is diagnosed with autism. However, only about a quarter of children with autism are diagnosed before the age of 3. Early detection, early behavioral intervention, and education can significantly enhance the social and cognitive functions of children with autism. The research team from Karolinska Institute in Sweden published a paper in the Journal of the American Medical Association Online Open, stating that their machine learning data was sourced from a research project in the United States that recruited approximately 50000 individuals with autism and their family members, including 15330 diagnosed children and 15330 undiagnosed children, covering different age groups, races, and genders. Based on the medical background provided by parents and the survey questionnaire filled out, the research team extracted 28 easily obtainable measurement indicators for children before the age of 2, such as the time point of the first smiling expression. Based on this data, the research team constructed and trained a large machine learning model to search for different combination patterns of these features and their representation relationships in large-scale data containing both autistic and non autistic children, and selected the model with the best performance for further data testing. The AI model conducted autism screening on approximately 10000 participants. The test results show that its recognition accuracy is 78.9%, with an accuracy of 78.5% for children under 2 years old, 84.2% for children aged 2 to 4, and 79.2% for children aged 4 to 10. However, when screening data from another group of participants, the accuracy rate was only 68%. The research team stated that the model still needs further optimization and will consider incorporating technologies such as eye tracking in the future. This is not the first time that researchers have attempted to use AI for early screening of autism. Previous studies have combined AI technology with retinal scanning in children for identification. (New Society)