According to a new study from the University of Turku in Finland, artificial intelligence can interpret social characteristics among people in images and video almost as reliably as humans.
People are constantly and quickly assessing each other's actions and interactions. Latest AI models such as the large-scale language model ChatGPT developed by OpenAI can explain what's going on in images and videos. However, it is not clear whether AI's interpretive capabilities are limited to easily recognizable details or whether complex social information can be interpreted.
Researchers at the Turku Pet Center in Finland have studied how accurately the popular language model ChatGPT can assess social interactions. This model was asked to evaluate 138 different social features from videos and photos. This feature explained a wide range of social characteristics, including facial expression, physical movement, and social interaction characteristics such as cooperation and hostility. Researchers compared assessments made by AI to over 2,000 similar assessments made by humans.
The findings showed that the ratings provided by ChatGpt were very close to those made by humans. The AI ratings were even more consistent than those created by one person.
“The ratings of social characteristics of CHATGPT were on average more consistent than those of individual participants, so the ratings could be even more reliable than those created by one person. However, multiple people's ratings are even more accurate than those of artificial intelligence.”
Artificial intelligence could boost neuroscience research
The researchers used AI and assessment of the social situation of human participants to model the brain network of social cognition using functional brain imaging in the second phase of the study. Before researchers can see what happens in the human brain when they see videos and photos, they need to assess the social situation they draw. This is where AI has proven to be a useful tool.
“When we mapped the brain networks of social perception based on ChatGPT or people's social assessments, the results were surprisingly similar,” says Santavirta.
Researchers say this suggests that AI could be a practical tool for large-scale, tedious neuroscience experiments. For example, interpreting video footage during brain imaging requires considerable human effort. AI can automate this process, which reduces data processing costs and significantly accelerates research.
“While collecting human ratings required efforts from over 2,000 participants and a total of over 10,000 working hours, ChatGpt produced the same rating in just a few hours,” Santavirta summed.
Practical applications from healthcare to marketing
Researchers have focused on the benefits of AI for brain imaging studies, but the results suggest that AI can also be used in a wide range of other practical applications.
Automatic assessment of social situations using AI from video footage can help, for example, monitor patient well-being. Additionally, AI can assess the likelihood of receiving audiovisual marketing by target audiences and predict unusual situations from security camera videos.
“AIs don't tire you out like humans, but they can monitor situations around the clock. In the future, monitoring more and more complex situations could likely be left to artificial intelligence, allowing humans to focus on checking the most important observations,” says Santavirta.
/Public release. This material of the Organization of Origin/Author is a point-in-time nature and may be edited for clarity, style and length. Mirage.news does not take any institutional position or aspect, and all views, positions and conclusions expressed here are the views of the authors alone.
