There are basically two ways to be warm and empathetic in a discussion: just agree (easy, fake) or disagree in the nicest possible way while taking into account the specifics of the question and the personality of the other person (hard, more honest and can be more productive in the long run). I suppose it would take a lot of "capacity" (training, parameters) to do the second option well and so it's not done in this AI race. Also, lots of people probably prefer the first option anyway.

▲

perching_aix 5 days ago | parent [-]

I find it to be disagreeing with me that way quite regularly, but then I also frame my questions quite cautiously. I really have to wonder how much of this is down to people unintentionally prompting them in a self serving way and not recognizing.

▲

lazide 5 days ago | parent | next [-]

The vast majority of people want people to nod along and tell them nice things.

It’s folks like engineers and scientists that insist on being miserable (but correct!) instead haha.

▲

perching_aix 5 days ago | parent [-]

Sure, but this makes me all the more mystified about people wanting these to be outright cold and even mean, and bringing up people's fragility and faulting them for it.

If I think about efficient communication, what comes to mind for me are high stakes communication, e.g. aerospace comms, military comms, anything operational. Spending time on anything that isn't sharing the information at these is a waste, and so is anything that can cause more time to be wasted on meta stuff.

People being miserable and hurtful to others in my experience particularly invites the latter, but also the former. Consider the recent drama involving Linus and some RISC-V changeset. He's very frequently washed of his conduct, under the guise that he just "tells it like it is". Well, he spent 6 paragraphs out of 8 in his review email detailing how the changes make him feel, how he finds the changes to be, and how he thinks changes like it make the world a worse place. At least he did also spend 2 other paragraphs actually explaining why he thinks so.

So to me it reads a lot more like people falling for Goodhart's law regarding this, very much helped by the cultural-political climate of our times, than evaluating this topic itself critically. I counted only maybe 2-3 comments in this very thread, featuring 100+ comments at the time of writing, that do so, even.

▲

pjc50 5 days ago | parent [-]

People say they're unemotional and immune to signaling when they very much aren't.

People cheer Linus for being rude when they want to do the same themselves, because they feel very strongly about the work being "correct". But as you dig into the meaning of correctness here you find it's less of a formal ruleset than a set of aesthetic guidelines and .. yes, feelings.

	▲	lazide 5 days ago \| parent [-]
		Dig deep enough, and every belief system ends up having some deep philosophical tenet which has to be taken on faith, because it’s impossible (or even contradictory!) to prove within the system itself. Even rationality. After all, that evidence matters, or that we can know the universe (or facts) and hence logic can be useful, etc. can only be ‘proven’ using things like evidence, facts, and logic. And there are plausible arguments that can tear down elements of each of these, if we use other systems. Ultimately, at some point we need to decide what we’re going to believe. Ideally, it’s something that works/doesn’t produce terrible outcomes, but since the future is fundamentally unpredictable and unknowable, that also requires a degree of faith eh? And let’s not even get into the subjective nature of ‘terrible outcomes’, or how we would try to come up with some kind of score. Linux has its benevolent dictator because it’s ‘needed it’, and by most accounts it has worked. Linus is less of a jerk than he has been. Which is nice. Other projects have not had nearly as much success eh? How much of it is due to lack of Linus, and how much is due to other factors would be an interesting debate.

▲

Gareth321 5 days ago | parent | prev [-]

I find ChatGPT far more likely to agree with me than not. I've tested various phrases and unless I am egregious wrong, it will attempt to fit the answer around my premise or implied beliefs. I have to be quite blunt in my questions such as "am I right or wrong?" I now try to keep implied beliefs out of the question.