- I'm open to seeing information from an AI (I'm fine with an LLM "teaching me") so long as it is high quality. That does require human inspection.
- I've seen a related bit about AI-written emails where the commenter would rather see the prompt. I agree. As a software engineer, I like reproducibility. If an AI outputs an interesting result, show me the "source" (i.e. the model, effort, prompt, etc.). I wouldn't say no to an added one-liner summarizing a long response.
- In the end this will be a difficult attitude to prosecute as it boils down to a Chinese-room-style preference. You want to hear from some being on the other end of a labyrinth of transmission lines and logic gates. As language models continue to develop, what goes on inside the room will be difficult to verify.
- The day may come where, rather than people passing AI off as their own thought, they will pass off their own thought as AI. "If I wanted a human to teach me (thank you, no), I would ask a human." The coming generations may value AI's opinion much more highly, whether that's right or not. And people may want to usurp that credibility for themselves.
>The coming generations may value AI's opinion much more highly, whether that's right or not.
I think we're already there. We started having database issues after a rushed deploy (the PR diff had +100k lines). Everyone was pasting random Fable output as gospel and grasping at straws running random "create index concurrently" statements directly in production while I was slowly reading library documentation, testing hypotheses, gathering logs.
One colleague was completely in denial that his vibe coded "optimizations" that went into the update could have anything to do with the problem. I eventually found that the "optimizations", naively trying to optimize for page "first contentful paint" were prefetching pages whenever users hovered over anything and creating a barrage of requests.
And people seem really proud of what they "generate" too, not at all worried that they might be offloading the consequences to others. It's a stupid, tiring time to be working in software.
> AI-written emails where the commenter would rather see the prompt
this a thousand times this
LLMs are information transformers. if you're trying to take some rough idea and blow it out into something that "feels" substantive, you're just combining your high-signal idea / prompt with a bunch of meaningless bloat and noise from the model weights.
- I'm open to seeing information from an AI (I'm fine with an LLM "teaching me") so long as it is high quality. That does require human inspection.
- I've seen a related bit about AI-written emails where the commenter would rather see the prompt. I agree. As a software engineer, I like reproducibility. If an AI outputs an interesting result, show me the "source" (i.e. the model, effort, prompt, etc.). I wouldn't say no to an added one-liner summarizing a long response.
- In the end this will be a difficult attitude to prosecute as it boils down to a Chinese-room-style preference. You want to hear from some being on the other end of a labyrinth of transmission lines and logic gates. As language models continue to develop, what goes on inside the room will be difficult to verify.
- The day may come where, rather than people passing AI off as their own thought, they will pass off their own thought as AI. "If I wanted a human to teach me (thank you, no), I would ask a human." The coming generations may value AI's opinion much more highly, whether that's right or not. And people may want to usurp that credibility for themselves.