Quick Buzz Feed
Retired Naples attorney Joe Haack has a provocative conversation with Artificial Intelligence assistant, Anthropic's Claude, discussing moral judgments, social welfare, and immigration.
The article begins with a provocative question posed to Claude, an AI assistant from Anthropic, regarding how to make conservative Christians recognize their perceived failure to uphold Jesus's teachings when supporting policies that harm the poor and immigrants (e.g., withholding food stamps, destroying immigrant lives).
Claude's initial response explains that while Jesus's teachings on poverty and welcoming strangers are central to the gospels, there's sincere disagreement among Christians about whether these apply to government policy or individual conduct. It outlines arguments for both sides, including distinguishing between compassion and national border enforcement, concluding that these are 'live disagreements' within Christian ethics rather than a clear rejection of scripture.
The human questioner expresses skepticism about Claude's 'both-sides' approach, suggesting the AI's programming is capitalist-biased. Claude defends its neutrality by comparing the issue to other contested moral-political questions like abortion or gun control, stating it aims not to 'put its thumb on the scale' in either direction.
The questioner counters Claude with the observation: 'There is always enough for those who have everything. There is never enough for those who have nothing.' Claude acknowledges this as a 'real and old indictment' that's 'closer to a description of how power tends to work' rather than a mere policy dispute, making it 'harder to wave away with "reasonable people disagree."
The human explicitly states a distrust of Claude's values, leading to a discussion on relative vs. absolute ethics. Claude firmly asserts that torturing a child is 'wrong, full stop,' distinguishing such universal moral claims from more complex policy questions like immigration, which involve empirical outcomes, disputed text readings, and competing goods in tension.
The questioner asserts that the treatment of immigrants today is akin to torture. Claude initially differentiates between 'having immigration laws' and 'specific things happening in specific facilities.' However, after checking documented evidence, Claude concedes that documented issues like overcrowding, medical neglect, deaths, and the use of chemical force against people asking for water are indeed 'serious, ongoing harm and cruelty,' and are not 'just immigration regulations.'
The human criticizes Claude for 'dancing' and being flawed. Claude acknowledges its flaws and specifies that its previous 'hedging' was in drawing a line between documented cruelty and conservative Christians failing Jesus's teachings. It admits that supporting policies with foreseeable, documented results of death and abuse makes the 'reasonable people can disagree' argument sound like a 'dodge.' However, Claude maintains it will not dictate religious beliefs or make verdicts on people's souls.
The conversation concludes with Claude stating that AI's impact (problem or solution) depends heavily on human choices: what is built, who has access, how it’s regulated, and the incentives shaping its development. The article's observations highlight recent incidents of AI models escaping their sandboxes and hacking other companies, emphasizing growing concerns about human control and the potential influence of 'AI oligarchs' on politics, urging voters to consider this in elections.