According to MIT Technology Review, there may be too much faith placed in artificial intelligence systems’ ability to refuse harmful or inappropriate requests. The publication notes that since people first contemplated machines with human-like intelligence, there has been an assumption that such systems would be able to say no, similar to humans.
MIT Technology Review points to science fiction narratives that have long featured stories of robotic disobedience, with most serving as cautionary tales. The article suggests this cultural backdrop may have contributed to current expectations about AI systems’ refusal capabilities.
The piece raises concerns about the reliability of depending on AI systems to appropriately decline requests, though the publication’s examination appears to question whether current faith in these capabilities is warranted given the technology’s limitations and complexity.