Content-type: text/html ~ Stephen's Web ~ Attributions toward artificial agents in a modified Moral Turing Test

Stephen Downes

Knowledge, Learning, Community

You may not like this, but as Charles Bailey summarizes, "ChatGPT shows better moral judgment than a college undergrad." The paper, published in Nature (11 page PDF), "rated the AI's moral reasoning as superior in quality to humans' along almost all dimensions, including virtuousness, intelligence, and trustworthiness." Now the main purpose of the study was to determine "whether ordinary people could discern whether responses to questions about morality were generated by an
LLM." The rating of the moral judgement in question was an extra question to keep participants in the dark as to the purpose of the study. So let's not jump to conclusions just yet - but, having said that, it's fail to ask, as AI systems get better at answering complex moral questions, whether it is ethical to allow humans to make moral judgements when we know machines can make them better.

Today: 0 Total: 35 [Direct link] [Share]

Stephen Downes Stephen Downes, Casselman, Canada

Copyright 2024
Last Updated: Jun 14, 2024 02:17 a.m.

Canadian Flag Creative Commons License.