A recent study from Humboldt University of Berlin reveals that AI models now have the capability to mimic the language and cognitive processes of children. This discovery could fundamentally change how we think about the development and application of AI.
The AI market has seen explosive growth in recent months. With the introduction of technologies like ChatGPT, advances in AI research have become unstoppable. Daily reports on new achievements testify to an industry that is rapidly evolving. Yet, the latest study introduces a fascinating twist: AI can now also imitate the speech and thought processes of children.
Researchers at Humboldt University have thoroughly examined the stages of human development to explore the extent to which AI language models can simulate human cognitive abilities. “Thanks to psycholinguistics, we have a relatively good understanding of what children at various ages are capable of,” explains Anna Marklová, the lead author of the study. Her research suggests that large language models, like GPT-4, are indeed capable of simulating a lower level of intelligence than they actually possess.
In the study, the researchers conducted 1,296 independent experiments with GPT-3.5-turbo and GPT-4, utilizing specially designed False-Belief Tasks to test the AI systems’ adaptability to different child developmental stages. The findings show that as the age of the simulated child increases, so does the complexity of the language used by the AI systems. This represents a remarkable advancement in AI and raises important questions for the future of artificial superintelligence (ASI).
“We must be cautious when expecting ASI to mimic human intelligence, as this could lead to underestimating their actual capabilities,” warns Marklová. This underscores the necessity to carefully monitor the development of AI to ensure its safe and responsible use.
The study not only opens new avenues for AI research but also highlights the importance of an ethical approach to the development of technologies that have the potential to fundamentally change our understanding of intelligence. It’s an exciting time for AI, with limitless possibilities and challenges on the horizon.