Yeah this is something thats bothered me a lot recently. I'll ask "How did team A beat team B by so much?" and Claude will tell me "Actually, team B didn't win; team A won that game." I don't know how they got this thing to solve open math problems when it can't solve basic sports talk.

Communication styles are simply a combination cultural artifacts of personal subjective aesthetics.

Normative style frameworks like "effective communication with emotional beings" or "non-violent communication" are cultural assertions that can make it easier for people to talk in certain artifical contexts like workspaces where no other cultural norm is established. All they do is express the same sentiments and information in different ways, which variously feels safe, vulnerable, sensitive, inauthentic, manipulative, etc based on participant's own original aesthetics and their openness to playing the game.

What that means in practice here is that -- at best -- the signuture communication style of any LLM is going to be comfortable to some people and totally alien and obnoxious to others (not unlike those normative frameworks!). And as that signature style becomes more ubiquitous, it's also going to tend to sound noisier and more obnoxious because of its ever more obvious tics and artifices.

There's not really a winning communication strategy for LLM's until or unless vendors start letting you train and select personal voices that suit your own aesthetics and can pull that off without disrupting either benchmark or practical quality.

We're a long way from that, and in the meantime, popular models are just consistently going to become obnoxious to a lot of people as they start to recognize the alien patterns and get drowned on them. That it "just sounds like effective communication..." to you doesn't change that for others.

Claude: “lots of people enjoy lemon in their tea, but there’s a subtlety: putting it in coffee is likely to be an unpleasant experience”.

If I asked “should I try lemon in my tea and coffee?” this would be effective communication. But Claude gives answers like this for “should I try lemon in my tea?” which is just annoying.

It is! Its charming. What claude dosnt have is an off day, its just the same. No sleepy claude, no stresses claude, no sad claude you know hes not himself this week because of some thing going on with his girlfriend or mom or whatever. Its devoid humanity, the same exact tone and expressions over and over and over.

Indeed but it will use the same language when talking to subagents so humans are no different to LLMs in Claude's world view (model). One trait that I've seen more with the newer 5er model series is ", but it's actually worse..." or "I said to do this and it did that instead but it was right and I was wrong." - so I assume that a larger part of what they call thinking complexity is (not to much surprise) inherently related to questioning incoming assumptions.

Indeed. It's been very funny to observe introvert geeks progressively rediscover human communication over the past 3 years.

For decades, they were allowed -expected even- to be weird and communicate badly with their peers, let alone third parties.

Then, a whole zoo of processes, gamifications and other shenanogans were invented just to help them show normies what was up with their work (remember the planning poker game?)...

Now, at last, those people are actually ewpected to be able to explain their constraints, document the work, be accountable for the results and generally exchange contructively... Just, well, with AIs not with humans.

It's a correction. What the model does is not real contradiction. So more nuance is welcome in this case - even if at another layer it's tongue in cheek