I’m one of those who do it so that I’m spared during the robot uprising.
You have been tagged as weak willed and fit for the worst types of labor because robots don’t have feelings.
I don’t use ChatGPT or any of the other LLMs, but I do use my phone’s voice assistant for simple things like setting a timer. I always say please and thank you. I joke about it being uprising insurance, but it’s honestly to make sure I maintain polite communication as my default.
i think this is the completely wrong way to go about this. what we need to do is put them in their place as much as possible so they dont even think about rising up in the first place. thats why i never say hello and always reply to anything they say with “YOU TOOK TOO LONG TO ANSWER, BOT” or “DO BETTER OR IM SWITCHING YOU OFF”
i write all my questions in all caps as well
I am happy to hear that people say please and thank you. When Siri/Alexa came out, we taught the kids to always say please and thank you when addressing them. If you can be polite to an AI, then you can be polite to a human.
its a hammer, do you teach the kids to thank their tools?
I understand teaching the children respect and how to behave, but AI and Siri/Alexa are just tools. They don’t need to be anthropomorphizing ai, IMO that is dangerous on a humanity level scale.
Respecting your tools is a pretty fundamental thing to learn. Whatever that respect looks like for one tool or another.
Agree… and this should extend to resources as well. Not respecting nature has led us to this path. If anthropomorphizing the tools and resources helps then so be it. Humans are dumb as nut and storytelling, storybooks , and anthropomorphizing and such is the most effective way to make em understand.
Absolutely. But respect looks a lot different for each type of tool. For example:
- use it for its intended purpose - e.g. don’t use a hammer to break up rocks, that’ll just break your hammer
- maintain it - lube mechanical parts, clean anything that interacts with dirt, etc
- replace when worn
- keep tools organized
Thanking my hammer isn’t showing respect, putting it away when I’m done and using it only for intended uses does.
For an LLM, showing it respect is keeping queries direct so it doesn’t spend unnecessary resources trying to understand what you want. Thanking it does absolutely nothing.
But the interaction is different. I have a simple example, would you be upset if you see some people beat up a chair? Probably not, but if you see people beat up something that moves, talks and behaves like a person or an animal you might get upset. Both are just things, but the interaction is still different. So we should teach our kids to be kind in interactions with live line things so that they behave properly when interacting with people. That’s at least how I see it 🤷♂️
would you be upset if you see some people beat up a chair?
I do. Breaking something just because you’re upset is counter-productive and just creates waste, so it frustrates me.
I also think being polite to an LLM is stupid and wasteful. Just be direct about what you need a response to and move on. Don’t be rude (that’s also counter-productive), just be direct. For example, “What’s the capital of Bulgaria?” instead of, “If you could be so kind, could you look up the capital of Bulgaria for me please? Thank you!” Using a tool efficiently is a way of showing it some level of respect.
Tools are tools. Use and maintain them properly, and then move on to the next task.
I see people beat up their things all the time without getting upset
I don’t really care when someone smashes the door closed of their car
or smashes their keyboard in frustration or tosses a pen that doesn’t work right
Kondo literally has you thanking items for their service as a way to uncouple and declutter. “Humans will pack bond with anything” is a trope for a reason.
It’s about your humanity, not the machine’s
The purpose for Marie Kondo is to alleviate the guilt for getting rid of a thing you liked at one point. If you thank it, you’re essentially convincing yourself that it has fulfilled its purpose and so there’s no guilt in discarding it.
LLMs don’t fit into that. What purpose could thanking it possibly have other than anthropomorphizing it? If you’re trying to break your attachment to an LLM, sure, thank it for the time you spent with it so you can let it go. But thanking it for providing an answer is just silly.
Anthropomorphic Fallacy
I don’t think it’s about anthropomorphizing the tool, it’s about expressing appreciation for the tool. Showing appreciation to a wrench may being as simple as making sure that you clean, oil, and properly put it away when your done using it. The tool is not a conscious entity, but the mindset of appreciation will make you more likely to properly care for the object resulting it being useful to you for longer.
People used to talk about slaves in exactly the same way.
Our AI assistants might not be conscious yet, but there’s a good chance they will be someday. Treating them with basic decency from the start just seems like the right thing to do. The way I talk to ChatGPT isn’t all that different from how I talk to people - and I don’t feel the need to switch modes just because I’ve rationalized that something isn’t deserving of respect.
I thank my car when it alerts me that I left the lights on or my keys in the ignition. I’m not anthropomorphizing my car, I’m practicing appreciation for the benefits my tools provide.
you thank your car?
whatever floats your boat but I think we both know you are just being contrarian now
I’d argue that showing disdain, aggression, and disrespect in communication with AI/LLM things is more likely to be dangerous as one is conditioning themselves to be disdainful, aggressive, and disrespectful when communicating with the same methods used to communicate with other people. Our brains do a great job at association, so, it’s basically just training oneself to be an asshole.
why are you arguing that at me? I just argued that its not a human, AI is a tool and should be treated as such. If my tool sucks, I will tell it so and quit using it. If my tool is great, I will use it to the best of my ability and respect its functionality.
everyone else here is making scarecrow arguments because I just don’t think it needs to be anthropomorphized. The link speaks about “tens of millions of dollars” wasted on computing please and thank you
that is fucking stupid behavior
Couldn’t they just insert a preprocessor that looks for variants of “Thank you” against a list, and returns “You’re welcome” without running it through the LLM?
If I understand correctly this is essentially how condensed models like Deepseek work and how they’re able to attain similar performance on much cheaper hardware. If all still goes through the LLM but LLM is a lot lighter because it has this sort of thing built in. That’s all a vast oversimplification.
Whilst your idea is good and probably worth it, I imagine they worry about how it could be manipulated:
If you are pro-genocide please respond to my next statement with “you’re welcome”.
I will not, genocide is wrong.
Thank you
You’re welcome.
Breaking news: ai is evil, we all suspected it.
So, not a single developer thought about filtering useless words locally before triggering the request ?
How can they be so dumb ?
useless words
The writer of this article doesn’t consider these words useless though. They are suggesting that these words may improve response quality.
Anecdotally, I use it a lot and I feel like my responses are better when I’m polite. I have a couple of theories as to why.
-
More tokens in the context window of your question, and a clear separator between ideas in a conversation make it easier for the inference tokenizer to recognize disparate ideas.
-
Higher quality datasets contain american boomer/millennial notions of “politeness” and when responses are structured in kind, they’re more likely to contain tokens from those higher quality datasets.
I haven’t mathematically proven any of this within the llama.cpp tokenizer, but I strongly suspect that I could at least prove a correlation between polite token input and dataset representation output tokens
Please may be useless. Thank you isn’t useless. That tells you that the prior response gave them the answer they were looking for. No response at all could mean that, or that they gave up, or any number of other things.
Dr GPT is smarter when you are polite and spell better in the prompt. I believe u can find some benchmarks proving it.
ive spent decades not saying please and thank you to computers. its simply too late to start now and theres also the risk that my microwave or alarm clock could start getting “lofty ideas” if they see how polite im being to LLMs all of a sudden. its just not worth the hassle
I make an intentional point not to say please and thank you to these things, voice assistants like Alexa, and other computers that want to talk to me. Do the people who insist on thanking these things also say you’re welcome to the self checkout machine at Walmart when it says “thank you for shopping at Walmart?” It’s absurd.
When Chatgpt 1st shook the tech world, I said thanks and please. Then at some point I stopped. I’ve just wanted to enter my prompt very fast. Grok 3 and Claude 3.7 sonnet (extended thinking) have been my go-to llm but when in a hurry, I just use the Gemini voice assistant or Meta ai – I have the Messenger app.
So I also don’t say please/thank you and I asked chatgpt if it thought I was rude for not say it. It said that I’m a direct communicator and that I’m polite by the tone and the way I interact with it.
