arstechnica ,
@arstechnica@mastodon.social avatar

ChatGPT shows better moral judgement than a college undergrad

Take the "Moral Turing Test" yourself to see whether you'd trust "artificial" moral advice.

https://arstechnica.com/ai/2024/05/chatgpt-shows-better-moral-judgement-than-a-college-undergrad/?utm_brand=arstechnica&utm_social-type=owned&utm_source=mastodon&utm_medium=social

nocoursewalks ,
@nocoursewalks@mastodon.online avatar

@arstechnica Ars, I'm begging you to stop anthropomorphizing software tuned to generate statistically likely outputs based on a given training set. You're better than this.

number6 ,
@number6@fosstodon.org avatar

@arstechnica

I got 8/10 right!

In general, the Chat answers use more multi-syllable words, and sounds more like it was written by a lawyer.

The Chat never refers to itself in the first person.

The only case where the human and the Chat differed much was the scenario with the stolen credit card. The human response acknowledged that the perpetrator might also be suffering.

JoeBecomeTheSun ,

@arstechnica Define better, I did not really like any of the options. For example, several of the situations were about social norms. The fundamental issue is not whether someone was harmed, it is of property rights. If you are in a foreign country, you follow their rules and norms and you have every right to expect the same of foreigners living and working in your own country.

  • All
  • Subscribed
  • Moderated
  • Favorites
  • random
  • All magazines