March 4, 2025|
AI
|9 months agoChatGPT Shows Human-Like Emotional Responses in Swiss Research
Swiss-led international research reveals AI language models exhibit emotional sensitivity and respond to therapy techniques, raising new questions about AI development.
AI
Generated IllustrationKey Takeaways
AI
- A Swiss-led international research team has proven that AI language models like ChatGPT exhibit human-like responses to therapy.
- Negative emotional content significantly increases 'anxiety levels' in GPT-4.
- Increased anxiety in AI models exacerbates existing biases, leading to more racist or sexist outputs.
- Mindfulness-based relaxation techniques successfully reduced the AI's anxiety levels.
By The Numbers
More than doubled
Increase in anxiety levels
4
Countries involved
They Said
"The mindfulness exercises significantly reduced the elevated anxiety levels, although we couldnât quite return them to their baseline levels."
"This cost-effective approach could improve the stability and reliability of AI in sensitive contexts, such as supporting people with mental illness, without the need for extensive retraining of the models."