AI's inability to grasp the world as comprehensively as humans is due to the complexities of human cognition and experience that AI systems have yet to replicate.
Artificial Intelligence Still Can'tMatch Human Understanding
Artificial intelligence, including the popular ChatGPT, lags behind human comprehension, a new research published in the journal Nature Human Behaviour reveals. The study, led by psychologist Tsinghua Xu from Ohio State University, suggests that AI's inability to grasp even simple concepts as humans do stems from its sole reliance on text-based learning, unlike humans who learn through lived experiences.
"AI can't smell a rose, touch a daisy's petals, or walk through a field of flowers," Xu points out. "Without these lived experiences, AI can't truly understand what a flower really is."
To measure AI's understanding against human comprehension, the research team evaluated thousands of words in human subjects and advanced language models like OpenAI's GPT-3.5, GPT-4, Google's PaLM, and Gemini.
The researchers employed two evaluation methods: Glasgow Norms, which rates words on emotional intensity, imagery, and other factors, and Lancaster Norms, which measures the connection between words and sensory (vision, audition, olfaction, touch) and motor (hand, mouth, body movements) experiences.
The findings showed that AI outperforms humans with abstract concepts but struggles with sensory and bodily ones such as "flower," "hand," and "fried onion."
Authors emphasize that understanding is more than just language. "Even if AI processes millions of text words in a lifetime, it still won't capture the richness of human perception," Xu stated.
The research team foresees a future where AI models can feel and act in the physical world through sensors, robots, and real-life experiences, capable of approaching human-like understanding.
It's interesting to note that previously, scientists found people increasingly developing feelings for AI. Yet, as the current study shows, AI still lacks the depth of sensory and emotional experiences that humans rely on for understanding the world.
Unraveling the Differences: Human vs. AI Understanding
| Aspect | Human | AI ||-----------|-----------|-------|| Learning Method | Based on personal experiences, sensory inputs, and emotions. | Trained on large datasets of text. || Creativity | True originality and imagination driven by emotions and experiences. | Mimics creativity through pattern recognition and data combination. || Emotional Intelligence | Deeply integrated with emotions and empathy. | Lacks emotional experiences and empathy. |
While AI like ChatGPT can generate human-like language and mimic creativity to some extent, it still lags behind in replicating the full depth of human understanding, rooted in lived experiences, sensory perceptions, and emotional intelligence. The future, perhaps, lies in AI models capable of living in the physical world and interacting with objects, allowing them to better understand our world as humans do.
- Artificial intelligence, currently, struggles to grasp sensory and bodily concepts that humans easily understand, such as the distinction between a 'flower', 'hand', or 'fried onion'.
- The study suggests that humans, unlike AI, learn through their lived experiences, bringing a richness of perception to their understanding.
- AI's lack of emotional intelligence stems from its reliance on text-based learning, unlike humans who integrate emotions deeply into their understanding of the world.
- In the future, AI models may be equipped with sensors, robots, and real-life experiences, potentially allowing them to approach human-like understanding by feeling and acting in the physical world.
