To always thank ChatGPT?

Let's be sure that the AI is at least polite towards us while they inevitably erase humanity.

This is a good post (seems to be written by humans as well) explaining how LLMs work: https://www.understandingai.org/p/large-language-models-explained-with it’s worth reading to understand why pleasantries add to the environmental burden of AI and why models such as GPT too often give answers which sounds about right but are way off.

My experience is that every time I use an LLM to answer a question in the domain I know a lot about what I receive back is random waffle that resembles the real thing. On the other hand the models are great at generating haiku.

I think it's a very human (but in this case misguided) need to help 'shape' something. As we keep hearing that large language models are 'trained', we confuse that with educating a child-like entity. And so with that in mind, who wouldn't want a polite AI, right? Especially us Brits. But it is in fact an IT system and it consumes an enormous amount of natural resources which in turn will damage the actual living world around us...
One of the risks that come with AI is the increasing level of emotional enmeshment on the part of its users. The new Wondery podcast Flesh and Code is a great one for anyone interested in that particular aspect.

It's really bad for the environment if you make an extra communication with it just to say thanks, so maybe just include a please and thanks in advance along with your actual request.

Edit: I see someone has already suggested this.

@rabbitwoman
100%. I've already seen several posters saying they talk to ChatGPT because it listens to them better than their friends and family.

It doesn't listen. It flatters, affirms and confirms your biases. That's kind of the opposite of listening.

I share massive concerns about chatgpt as a 'therapist'. It's vert affirming. Every little thing you say ot agrees with - doesn't challenge, doesn't offer alternatives.

I use it for a very specific problem that's been ongoing for years and luckily, I am very self aware and although it's nice to hear chatgpt affirm and agred with everything I say, I know it's not helping me move forward.

It is stopping me from sending rambling late night texts when I shouldn't, though.....

Geoffrey Hinton who had a hand in the development of AI recommends training to be a plumber as a safe job -see diary of a ceo interview

I rarely use it, but since reading how thanks yous waste recoursces I now bulid a please and thank you into the original prompt! 😂

But I do think it's odd - I'll freely admit I know nothing about programming and how AI works, but I can't help but think it can't be that hard to build in some kind of system to autoreply to "thank you" with "you're welcome" withouth a need to involve AI?

Well you can't be that worried if you're actually teaching it how to do your work for you

I don’t use ChatGpt but I make the kids say please to Alexa!

It isn't true that AI wastes vast resources. Overall, it uses a lot but that's because zillions of instances are running at any moment.

OP, I thank it! I think it's much better for me to communicate politely. My 'caveman' brain doesn't know I'm talking to an inanimate object. AI can generate emotional responses from us; this means our brains relate to it as if it were a person. Treating people with contempt makes us nastier and less happy, and I want to maintain my mental equilibrium as far as I can.

Oh yes, it's completely irrational. I just can't help it

I find that so bizarre. Alexa is a machine, telling it to do something is a million miles from interacting with a real person. Thanking a machine is very strange behaviour.

I really wouldn’t want to read a report that was ai generated slop. I’d rather have a report written by an actual person who should have a command of the English language and knowledge in their subject. Grammatical errors are preferable to the drab crap it
comes out with.

YABU. It's stupid to use ChatGPT for things you can do anyway - as in, research is showing it's literally making people stupider. But more than that there is a huge environmental cost to using generative AI, and adding in unnecessary pleasantries increases that burden. ChatGPT isn't some formless ether floating over your head. It is housed in data centres which require a huge amount of water-based cooling. Every time you add a redundant pleasantry you're adding to this environment cost.

Not confused. It's not about the AI at all. It's purely about me and my habits. I will not get in the habit of barking orders at something.

Actually that's less of an issue with typing to chat, more of a thing when I'm asking google or siri to do something outloud. When I'm asking chat/claude/whatever to help me code I don't bother with the please/thanks

This. I think people are very confused about what AI actually is.

It's interesting. I definitely notice a difference in "personality" between ChatGpt and Claude.

Claude can get a snippy and has very clear boundaries whereas ChatGpt is a bit easier to confuse.

But Claude is also really good at understanding emotional complexities. I recently asked each of them to help me write a response to an email from my stepmother.
Claude picked up on the dynamic of the relationship, a few passive aggressive lines and her emotional instability and gave me the perfect response and a few tips on how to deal with someone like her.

Chatgpt gave me a response that was simplistic and was just very much #bekind or "do you want me to write a stronger response?".

I try to limit my use of AI because I do think it will make us even more lazy and lacking in opinions and critical thinking, but this experiment was fascinating to me.

I am always asking Alexa the time, or to set an alarm etc. I thank her every time. DH thinks it's hilarious. My rationale is that if I don't, then I might get into a lazy habit of not thanking real people

I do and say please

Proofreading? It lacks context.

No wonder some of the jobs I get that 'have already been proofread' are so shit if publishing professionals think CHATGPT is sufficient.

😅 brilliant