For example, they never randomly scream at you because they had a bad day, nor secretely hostile because they have inherited some bias against people looking similar to you, so you don't have to guess the mood they're in before you risk entering a query: when you enter code and it doesn't work, it's because its logic is faulty, so it's easier to map input with output, their logic is sorely based on facts.
Google's botsplaining in the past years that kept making it increasingly harder to search for exactly what you entered was already frustrating.
And now we're going full speed towards long back and forth chats as primary input method.
Regular people are delighted because it makes these black boxes more relatable as they act "more human".
People like me, on the other hand, are saddened by the impending loss of one of the few things we could rely on because the output will become at least as random and unpredictably biased as humans.