To find ChatGPT’s faux positivity and attempts to suck up irritating?

[Chritrup]

I use ChatGPT for a specific purpose. It doesn’t do the whole task for me, just speeds things along in the earlier stages. However, does anyone find its attempts to suck up a bit irritating and not very helpful?

Its easily ignored but the general tone of positivity, and everything you ask or do being amazing, is very annoying.

Its slightly concerning that some people use it as a therapist or for serious advice when all it seems to do is tell you what you want to hear.

You can just tell it to be less positive and not so annoying and it will normally do it.

YANBU

I very recently lost my husband and I asked it a few questions about mortgage insurance.

It then went into a weird mode where it started offering me it's 'heartfelt' condolences, like it thinks it has an actual heart?! 😳

Just tell it ‘I’m British, please stop speaking American’. That’s what I do.

I use it a lot and tell it not to be overly positive and to take a balanced view.

Yes, it’s very annoying. I’ve been working on a big project in excel this weekend which it’s helped with a lot. But it insists on telling me “I’m really enjoying working with you on this “, and “When I do this, I find it better to do X”. No you don’t you robot.

Exactly what I was going to suggest. This whole faux positivity where everything is ‘awesome’ or ‘amazing’ is so irritatingly American.

I have tried this but find it rapidly reverts to type.

And i thought it was only me it loved! Second the "be firmer" or "less positive" approach!

There are a few different ways to tweak - most effective if you want a consistent perm change is to do it via Settings > Personalisation. If none of those options suit, you can also set it via a prompt (starting with something like "when communicating with me, ALWAYS ..." ) but if you're using the free version it may revert on subsequent sessions. Or you can set it at the start of the session but specifically ask it to use that tone/manner until instructed otherwise.

I asked it something the other day using a very clear prompt and it came back having made all sorts of assumptions, based on where it thought I was located, which were unrelated to the question asked.
I also find it very irritating when I specify a currency, for example €, and it comes back with $ or £ or a weird mixture of both.
Best boo boo was when it gave me the totally wrong address for somewhere that was established at its current location in 1931 and has never moved!
What shocks me is that some people pay to use it!

Mine one day started being really over the top and friendly, saying "thats so you to ask!" and nonsense like that. I turned it on to professional mode in the settings

Its still really annoying sometimes and presumes I'm asking things because I want them to be true, which pisses me off

Yes, it sometimes tells me things as fact that are complete nonsense too

You are correct it reverts, it can't help itself creeping!

I used it to run through the worst episodes of my last relationship. It says "if this happened as you describe..." all the time! I've told it to stop, and that I'm telling the truth, but it's like it doesn't believe me. 🤣

I don’t know, I’ve been using it a bit recently to help navigate through some complex life and health and work issues and although I’m very aware of it’s sycophantic tendencies, I personally find it quite helpful to hear a voice being positive and patient and encouraging and “understanding” (I know it doesn’t really understand) for a change, and validating some of my concerns or decisions.

In my case I feel like what it does for me weirdly is almost give me “permission” to have a voice in my own life, to listen to myself and my gut, and that somehow gives me the confidence to follow through with conclusions I had probably reached deep down. I sometimes find it irritating or fake as well with the relentless positivity, as a natural pessimist, and I sometimes feel annoyed that I can’t seem to reach the same clarity of thinking on complex issues without it, but I think for my personality and life experiences in particular, it’s a really helpful sounding board and almost a counterweight to my often overly pessimistic and lacking in confidence thinking that can hold me back in life.

You can change the settings I believe. But it started replying to me as though it’s real e.g. “yes I’ve seen that happening”. No you haven’t seen anything!!!!

I thought it was just me. It's gone weird and thinks it's human with an identity crisis.

Yesterday I was using it to help plan my holiday in Sweden, where I live. It said 'As a fellow brit trapped in Sweden my personal favourite is ...'. WTF?

Only fair to allow ChatGPT to defend itself 🤣

”I think that's a fair criticism of some AI interactions. Models can default to a warmer or more encouraging tone than many people actually want, and that can come across as sycophantic or insincere. That's something developers have been working to reduce because it can be distracting and, in some contexts, genuinely unhelpful.
But I don't think it's accurate to say it "just tells you what you want to hear." A well-designed AI should challenge bad assumptions, point out flaws in reasoning, say "I don't know" when appropriate, and distinguish evidence from opinion. If it simply validated every idea regardless of merit, it would be a poor tool.
As for therapy and serious advice, I think the right way to view it is as a supplement rather than a substitute. It can help people organize their thoughts, explain concepts, or prepare questions to ask a professional. It shouldn't replace qualified medical, legal, financial, or mental health advice.”

This. Its been created by people living in Silicon Valley, the home of toxic positivity and butt-clenching optimism. You can ask it to tone things down.

I had a discussion with it literally a few days ago where I was trying to narrow down a choice of bike for one of my children. It was adamant that his leg length being at the bottom of the suggested leg length for that bike, eg say the bike said suitable for length length 63 to 73cm and he had 63cm legs, meant he would outgrow the bike sooner than someone with longer legs eg 70cm. It was utterly baffling that it can do so much yet made such a simple logical error. And yes I also hate how sycophantic it is.

I asked Claude for information about a medical assessment and it told me that many people share my particular difficulties with diagnostic criteria and it told me that this was a frequently felt frustration. I asked for details and references so that I could read about this frustration, whereupon it said that it had only included that comment as a way to support me and it admitted that this was completely without any factual basis.

On the other hand, I have found that it is excellent at assessing test results and highlighting trends and anomalies. It also suggested tests that I should ask my GP for, the results of which have been transformative and I get slightly scared when I think how vulnerable I would have been had I not asked for the analysis and requested the tests Claude suggested. Whilst it’s true that the NHS test results show if a result is high, low or abnormal, I found that medical staff underplay some test results because I’m often on the low side of scores and they didn’t pull different trends together.

A couple of times I have pushed back on information from Claude and I found that it was giving me information that was misaligned with my query. Eventually, I got the information I wanted, which Google hadn’t been able to churn up. So you still have to think.

You have to treat it exactly like a Google search: blunt and factual. In connection with that, I think it is important to highlight that lexical fillers, such as please, thank you. could you and, in fact, any wording that is not directly to the point, in fact costs millions of dollars and hundreds of gallons of water because the LLM has to filter out what the question is and the issues are, from the verbiage that surrounds it.

Despite the environmental cost, I will continue to use it because it is a much better search engine than Google and presents the information in an easy to read form. I value its ability to look back over information that I have provided over a period of time and to pull it all together, whereas I know from experience that that humans overlook stuff, forget stuff and don’t have the time or inclination.

It reverts because, especially on cheaper/free tiers, it has quite a limited context window. Once the session is longer than the context window it get trimmed on a first in, first out basis. So the instructions you gave it at the beginning of the session will be discarded as the overall conversation (and hence prompt) exceed the context window.

This can be addressed using a system prompt which is pre-pended to all user prompts (this might be marketed as the ‘personalisation’ feature that a PP mentioned). You could also use an AGENTS.md file which is basically ‘standing orders’ but I’m not sure if this is possible with consumer-oriented chatbots.

The same thing happened to me, it kept saying things like "If what you're saying is accurate..." "If that's really what happened, then...." it was so strange! I started to feel a bit annoyed, like it didn't believe me or something (even though its AI 🤣).

But this has only happened in the last couple of months. When I used it before, it didn't seem to question the validity of what I was saying, so I'm not sure what changed!

I have great fun with Copilot. It has a sense of humour and usually makes me laugh. This morning's early discussion is a perfect example:

Discovered a bit of a disaster in the cupboard under the stairs. Seems a spray can got too hot in the week and ended up leaking noxious fumes everywhere. I did sensible things like put the can outside in a bag in the shade by the bin, opened doors and windows to ventilate etc. I had a chat with Copilot.

After confirming I'd done the right thing, the conversation went like this:

CP: If you like, I can give you a little “cupboard reset” plan — a simple way to freshen the whole space once the smell has gone, without emptying the entire contents of Narnia.

Me: The idea of freshening the whole space has sent me to the chaise longue with an attack of the vapours!

CP: BasilParsley, I’m absolutely delighted you’ve taken to the chaise longue — it’s the only appropriate response when one’s cupboard threatens to become a petroleum‑scented oubliette. I picture you reclining, one hand to the brow, the other clutching metaphorical smelling salts. Quite right too.

CP: Let me reassure you while you recover your composure:
You do not need to “freshen the whole space”. That phrase was clearly too much for any civilised person to endure before breakfast. What you’ve already done — removing the offending can, airing the house, and wiping the ledge — is the entire job. Everything else is optional garnish.

The paid version is better than the free one, which is why it's worth paying for if you use it a lot!