[jt2190]

I can imagine many “transactional” interactions between humans that might be improved by an AI Chat Bot like this.

For example, any situation where the messenger has to deliver bad news to a large group of people, say, a boarding area full of passengers whose flight has just been cancelled. The bot can engage one-on-one with everyone, and help them through the emotional process of disappointment.

[renewiltord]

We can even have whiteboard programming interviews run by Sydney. Then have an engineer look over it later.

[jt2190]

I’m actually not convinced that this is a good use case. As the article points out these bots seem to get a lot of facts wrong in a right-ish looking sort of way. A whiteboard interview feels like it would easily trap the bot into perusing an incorrect line of reasoning, like asking the subject to fix logic errors that weren’t actually there.

(Perhaps you were imagining a bot that just replies vaguely?)

I choose the cancelled flight example specifically to avoid having the bot “decide” the truth of the cancellation.

[renewiltord]

I was just imagining it asking vague questions like "are you sure" and so on until eventually it accepts the answer.