4 ms·
We really need to start thinking of how to reduce magical thinking in the field. It’s not pretty. They literally quote biblical guidance for the models and pray
by startupsfail 3y ago
We really need to start thinking of how to reduce magical thinking in the field. It’s not pretty. They literally quote biblical guidance for the models and pray that this would work.
And start their prompts with “You”. Who is “You”?
- hxugufjfjf 3y agoThe LLM. The most common end-user interface for LLM is a chat so the ser expects to be talking to someone or something.
- nomel 3y ago“You” is an optimization for the human user. Here’s some insight: https://news.ycombinator.com/item?id=35925154 https://news.ycombinator.com/item?id=35925154
- startupsfail 3y agoIf you see any prompt that starts with You, generally it is a poor design. Like using a “goto” or global variables.
- nomel 3y agoThis is true for something like raw GPT. For the chat models that have been specifically optimized for "you" prompts, this is false. See the discussion in the link I provided, along with the leaked copilot/bing prompts. Or, in other words, use a model in a way that fully takes advantage of how it was specifically optimized, from the intentional burning of massive amounts of compute time/money to get it that way.
- startupsfail 3y agoBing prompt is hilarious. And so wrong in so many ways.
- nomel 3y agoI would assume that OpenAI helped with the Microsoft prompt. Being Microsoft, and it being a front page feature, I would also assume that there are miles of powerpoint showing data in support for it. What do you see as a better prompt? How would you test its efficacy?
- Jorengarenar 3y agoYeah, GOTO or global variables are usually a not so good idea... except in cases where those are actually the best tools for the job and any alternative makes code actually worse. Coincidentally, the same applies to "you"
- woah 3y agoFunctions are an optimization for the human user