According to a recent study, the confusion in language perception affects the perception of AI intelligence. While AI, such as ChatGPT, generates visually appealing text, it lacks genuine consciousness and comprehension.
Human intelligence, deeply rooted in essence and connection to the external world, surpasses AI’s ability to “fabricate” or “imagine.” The study underscores that despite AI’s utility, it lacks essential human traits like environmental concern, survival instinct, and empathy.
Key Points:
- AI can generate text but lacks true comprehension, exemplified by language models like ChatGPT.
- AI differs fundamentally from human knowledge due to its absence of lived experiences and emotions.
- AI-generated text can perpetuate biases and produce harmful, prejudiced content unbeknownst to users.
The University of Cincinnati is the source of this study.
Diverse responses have emerged from tech leaders, policymakers, and the general populace regarding the advancement of artificial intelligence.
Some express alarm at the prospect of any “intelligent” device surpassing human capabilities, while others enthusiastically endorse AI technologies like ChatGPT as transformative tools for society.
Anthony Chemero, a professor of philosophy and psychology at the UC College of Arts and Sciences, argues that despite AI’s cleverness, it cannot match human intelligence, even though it may deceive and fabricate information like its creators.
In a co-authored paper published in Nature Human Behaviour, Chemero contends that while AI is undoubtedly impressive in a conventional sense, intelligent computers have existed for decades.
The paper asserts that ChatGPT and similar AI systems, known as large language models (LLMs), are trained on vast datasets mined from the internet, often reflecting the biases of the data sources.
Chemero suggests that while LLMs produce coherent text, they frequently engage in pure fabrication. While they excel at generating linguistic content, they require significantly more training than humans. He argues that they lack awareness of the context of their statements. Due to their lack of embodiment, LLMs operate differently from human cognition.
Although Chemero suggests labeling this behavior as “fabrication,” the developers of LLMs construct sentences by selecting the statistically most probable next word—without regard for accuracy. When they fabricate information, they refer to it as “hallucination.”
Furthermore, he asserts that with minimal prompting, an AI system can generate “offensive and biased statements.”
Chemero’s paper aims to underscore that since LLMs lack embodiment—being devoid of physical presence and social interaction—they do not exhibit intelligence in the same manner as humans.
He notes that LLMs exist outside the physical world and lack concern for their surroundings, contrasting with human priorities of self-preservation and environmental care.
The core message is that LLMs are indifferent, as Chemero states, “They don’t care,” highlighting the disparity in values between AI and humans. While humans prioritize survival and environmental well-being, LLMs lack this intrinsic drive.
