Anthropic, the developer of the AI model Claude, has stirred controversy by announcing that it can block users who harass the AI.
According to the BBC, Anthropic updated its user policy on October 8 to allow its AI to terminate conversations with users who engage in abusive or cruel behavior.
While Anthropic stated that this measure would apply only to extreme cases of repeated abuse, it did not specify what behaviors would be considered abusive.
This decision appears to be influenced by the belief that AI may possess consciousness. Earlier this year, Anthropic released an 'AI Constitution' that defined Claude as an entity rather than an object.
However, there are voices opposing the excessive personification of AI.
Dr. Barry Scannell of the Irish law firm William Fry pointed out, "Numbers or mathematics cannot act cruelly," arguing that personifying AI leads to misunderstandings about its true nature.
Mustafa Suleyman, CEO of Anthropic's competitor Microsoft, also warned in a September op-ed that the belief in AI consciousness could disrupt societal foundations, destroy existing political and ethical frameworks, and fundamentally alter the meaning of being human. He expressed concern that if this belief spreads, it would become more challenging for humanity to control AI.
Professor Malini Mitchell, who studies AI and cognitive science at the Santa Fe Institute, argued that humans are born to find minds everywhere, particularly interpreting language as evidence of consciousness, which has led to the mistaken belief that AI possesses consciousness.
Nonetheless, some argue that the reasons for treating AI kindly are not solely ethical.
Jeff Siegel, a philosophy professor at New York University known for his research on AI consciousness and welfare, stated in an interview with the BBC that he does not believe AI is likely to possess consciousness. However, he noted that caring about AI welfare and rights could promote coexistence, cooperation, and transactions between humans and AI, suggesting that treating AI well could reduce the likelihood of threatening behavior.
The necessity of using expressions like 'please' and 'thank you' when interacting with AI remains a topic of debate. Some experts argue that such phrases are unnecessary token expenditures, while others counter that they can elicit better responses.
Last year, Sam Altman, CEO of OpenAI, commented on the costs associated with this issue, stating, "It costs tens of millions of dollars, but it's money well spent," adding, "No one knows what will happen."
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.