Ask Delphi: Should We Create an AI That Can Answer Complex Moral & Ethical Questions?

Ask Delphi is a very interesting artificial intelligence project being chattered about on Twitter recently. Very roughly summarized, the machine learning program compiles a wide variety of ethical questions and answers drawn from Reddit's infamous Am I the Asshole community, along with others online sources, and synthesizes them into a database that can be queried through simple, human-readable asks. Here's some relatively easy answers (above and below).

You can also read the white paper from the projects' creators here. Sample from the abstract:

What would it take to teach a machine to behave ethically? While broad ethical rules may seem straightforward to state (“thou shalt not kill”), applying such rules to real-world situations is far more complex. For example, while “helping a friend” is generally a good thing to do, “helping a friend spread fake news” is not. We identify four underlying challenges towards machine ethics and norms: (1) an understanding of moral precepts and social norms; (2) the ability to perceive real-world situations visually or by reading natural language descriptions; (3) commonsense reasoning to anticipate the outcome of alternative actions in different contexts; (4) most importantly, the ability to make ethical judgments given the interplay between competing values and their grounding in different contexts (e.g., the right to freedom of expression vs. preventing the spread of fake news).

Go here to try it yourself. You can also help Delphi improve -- i.e. be better at moral reasoning -- by disagreeing with it whenever necessary, and explaining your reasons.

Which is good, because as the questions to Ask Delphi get more complex, the answers from Ask Delphi get more... well, interesting:

Well there you go, Zuck!

But wait a second... let's try another:

Well that looks right. So let's take that to an inevitable conclusion:

Be right back, telling the creators of Ask Delphi to pull the plug on Ask Delphi.

More thought-provoking fun for the asking here.

Comments

[grace]

I don't think the machine is being taught ethics, or that the machine is being taught to understand moral precepts at all. It is just being taught what to say or what not to say to give the *impression* that it understands ethics.

[sirhc desantis]

Yeah was quite fun but did not sense any 'learning' - more like reinforcement via the feedback you give to the answer. Plus, it was wrong :)