3 ms·

>Can you give a specific example of what the bots would say and the effect you think that would have on your subreddit? Sure, here you go: https://i.imgur.com/

>Can you give a specific example of what the bots would say and the effect you think that would have on your subreddit? Sure, here you go: https://i.imgur.com/EthyyS5.png https://i.imgur.com/EthyyS5.png Obviously the result would be more uplifting, solution-oriented, positive community. The bot did a good job. But supposing OpenAI instead replied "Sorry, I can't help you with that" then I would have to use an airgapped machine or take the time to do it myself and I don't have that kind of time. An example of this is at the bottom of the page.

LordDragonfang 3y ago
You keep saying "airgapped", and I don't think you understand what that means or what it is for. Airgapping is putting a machine on a physically isolated network to ensure data integrity/security against external network access. None of the failure modes for LLMs that you've described would have anything to do with that. Moreover, a reddit bot, by definition, cannot be airgapped. Also, your "airgapped machine" (locally-run LLM) can refuse requests just as easily as GPT can; it's all about the system prompt you give it, which you can choose with the API. On that note, the product in your screenshot is ChatGPT (a gpt3.5/4 api interface), NOT the GPT3.5/4 API itself, which is what you would be using. The "usage limit" pictured only applies to Chatgpt, the API is pay-per-use. You can try it out here: https://platform.openai.com/playground https://platform.openai.com/playground