Join for quick access. Enter your email below and we'll send confirmation plus sign you up to our newsletter.
By submitting your information, you confirm you are aged 16 or over,
have read our
Privacy Policy
and agree to the
Terms & Conditions. Geographical rules apply.
Stay On the Cutting Edge: Get the Tom's Hardware Newsletter
Get Tom's Hardware's best news and in-depth reviews, straight to your inbox.
You are now subscribed
Your newsletter sign-up was successful
Join the club
Get full access to premium articles, exclusive features and a growing list of member rewards.
An account already exists for this email address, please log in.
UK government tells staff to stop thanking AI chatbots — draft guidance pushes lightweight models and shorter prompts to cut environmental impact
The United Kingdom government has published a guide on how its employees should use AI, including a reminder that people don't need to thank their chatbots for their work. The new document, which is still a draft and yet to be finalized, is designed to help people use AI "safely, fairly and sustainably, especially for government work."
The draft "Using AI ethically and sustainably" document also warns users that they should check and follow any AI policies put in place by their individual departments, noting that there may be specific tools and approaches that they should use. It's unclear which departments these may be, but it suggests an awareness that usage patterns may change depending on the nature of the work being done. This is backed up by a section that reminds employees to use the right tool for the task. Examples include using Gemini Flash instead of Gemini Pro and GPT Instant instead of GPT Thinking.
The sub-600-word document offers guidance on using AI responsibly, including checking the output of an AI before using it. This doesn't just apply to making sure that any actual data the AI provides is correct, but also includes ensuring that "users or groups are treated fairly and equally" as well as avoiding "stereotypes or discriminatory language."
The government is also keen to make sure that its employees don't rely on AI, reminding everyone that they are responsible for any content or decisions that are made using the input of AI. They won't be able to simply point the finger at an AI-powered tool if something goes awry. This is especially notable given a recent report suggested that using AI can actually increase burnout because it allows workers to take on too many tasks.
The "Use AI efficiently" section is where the document aims to help reduce the environmental impact of AI use — including a reminder that users need not say thank you. Further, they're told that they should use as few prompts as possible while ensuring they are as short, clear, and direct as possible. There's also a (dead) link to a prompt library. It's possible that the link is only available to those on a government network.
Oliver Haslam has written about technology of all shapes and sizes for over 15 years, both online and in print. He's fascinated by how personal computing continues to change as tower PCs give way to foldable phones and beyond.
I only use polite words or phrases as part of a prompt I'm going to type, anyway. I never enter a prompt that serves no purpose other than to express gratitude. Most often, I will use these as a turn of phrase, such as "Thank you for explaining X, but I'm really more interested in Y."
Remember that what you get out of a chat bot has a lot to do with what you put in. Their responses are very context-dependent, and if you want a more detailed, thoughtful, and accurate response, then use of proper grammar, punctuation, and the appropriate jargon can help.
The article said:
It is, of course, true that you don't need to thank a chatbot for its help, but some may choose to ignore that advice. Especially given constant claims by AI companies that their tech "could kill all humans" sooner or later.
I guess this was a silly, throw-away remark. However, it's worth understanding how LLMs retain context, as that also gets to the underlying rationale for why thanking them is pointless. Once you've concluded a session, the context is discarded and the instance forgets everything about the interaction.
The only possible exception to that is if part of your interaction is incorporated into training data for future models, although I'm not sure exactly if or how they do that. Even if it is, it's literally a drop in the ocean of all their training data and won't make any difference. Certainly not enough to justify the additional prompt processing.
The fact this even has to be said is ridiculous. Environmental impact aside, thanking or otherwise being polite to a machine is nonsensical. That being said.... of course this would come from UK. 🤣
Teach the world to be even more rude than it already is, good idea!
I did have this same thought. It's the sort of thing that makes sense in the short-term and for pragmatic reasons. However, it could cast a long shadow, at a cultural level.
I predict that in a year or two there will be a flood of articles decrying the lack of pleasantries in communication. They'll all ask if interacting with AI was the motivation behind people becoming more rude without actually blaming AI. It'll be like the post-pandemic "everyone forgot how to socialize" crisis all over again!
Teach the world to be even more rude than it already is, good idea!
Or teach it to distinguish between humans and AI. Losing sight of the fact that you're interacting with the latter and that it has no feelings could be dangerous.
