LLMs really are a useful tool, especially those like Perplexity search and Kagi Assistant (paid but cool) that have access to live web results or your documents. Getting a quick summary of the content of the results complete with references is great.
It’s also very nice for coming up with quick outlines or lists of options. For example, I had a UX user interview, and I asked a LLM to generate a list of questions based on the user’s profile and the website I’m working on. I wrote my list first and compared it to the generated list (which was better organized), coming up with a finished list that I felt very confident about.
I do think it’s good for writing, too, just not in the way most people think. The folks behind IA Writer wrote a great post about using it to think more about your writing and gain another perspective. For someone like me who haaaaaaaaates bothering anyone about anything ever, having an always-available “partner” that will happily answer an infinite number of questions, deliver harsh and challenging evaluations, or brainstorm an endless number of things is incredible.
The real magic of this stuff, though, is just messing around. Stuff that you’d never take the effort to write or draw or commission but that it’d be amusing to see or read. If someone mentions that a Twitter post sounds like a poor Dennis Miller impression, I can ask a LLM for its take on a trite Miller-style monologue complete with lazy, shoehorned historical references. (Verdict: mostly accurate but too clever with similes.) And with image generators, if someone says “salm dunk” and it makes me imagine a salmon dunking a basketball, I can ask the computer to hallucinate it, and there it is right in front of my eyes, brilliant and colorful and usually weird in an unexpected day.
Was any of this productive? Did it add to anyone’s bottom line? Did it need to exist? No. But the computer did a stupid magic trick and made me laugh, so it was good.
LLMs really are a useful tool, especially those like Perplexity search and Kagi Assistant (paid but cool) that have access to live web results or your documents. Getting a quick summary of the content of the results complete with references is great.
It’s also very nice for coming up with quick outlines or lists of options. For example, I had a UX user interview, and I asked a LLM to generate a list of questions based on the user’s profile and the website I’m working on. I wrote my list first and compared it to the generated list (which was better organized), coming up with a finished list that I felt very confident about.
I do think it’s good for writing, too, just not in the way most people think. The folks behind IA Writer wrote a great post about using it to think more about your writing and gain another perspective. For someone like me who haaaaaaaaates bothering anyone about anything ever, having an always-available “partner” that will happily answer an infinite number of questions, deliver harsh and challenging evaluations, or brainstorm an endless number of things is incredible.
The real magic of this stuff, though, is just messing around. Stuff that you’d never take the effort to write or draw or commission but that it’d be amusing to see or read. If someone mentions that a Twitter post sounds like a poor Dennis Miller impression, I can ask a LLM for its take on a trite Miller-style monologue complete with lazy, shoehorned historical references. (Verdict: mostly accurate but too clever with similes.) And with image generators, if someone says “salm dunk” and it makes me imagine a salmon dunking a basketball, I can ask the computer to hallucinate it, and there it is right in front of my eyes, brilliant and colorful and usually weird in an unexpected day.
Was any of this productive? Did it add to anyone’s bottom line? Did it need to exist? No. But the computer did a stupid magic trick and made me laugh, so it was good.
We carve micro-patterns into rocks to make them think, and now they can be funny too