Published on October 14, 2024 5:10 PM GMT
I often talk to people who don't find LLMs useful. Paraphrasing:
Why would I want to ask questions to a thing that might give me theright answer but might also make something up that fools me intothinking it's right? I've played with them, but they don't seem likethey'd save me any time.
My experience with has been that there are some things LLMs are muchbetter at than others and you need to learn where they're suitable,but overall I get a lot out of working with them. I went back over myrecent history in Claude, theone I use most, and here are cases where I found it helpful:
In yesterday'spost I wanted to list a bunch of posts where I talked aboutfailures, but my recall isn't that great. I pulled a list of titlesfrom my last thousand posts and prompted:
I've attached a list of titles of some of Jeff Kaufman's blog posts.Can you identify some where he probably talks about a mistake he'smade?
While some it highlighted were bad examples or not about failures atall, others were posts I would essentially have had to read throughpages of titles to find. And I know my own posts well enough that inlooking over its answers I could immediately tell which were useful.
I help my dad with sysadmin work for his small B2B SaaS and have along-running issue where my edits to/etc/systemd/system/multi-user.target.wants/tomcat9.servicewere getting overwritten by automatic updates. I asked how to fixthis and got an answer (sudo systemctl edittomcat9.service) that I could then Google and validate.
I had a set of locations in an organism's genome that had acertain property, and I wanted to plot this. This would plot better,however, as a series of runs instead of a set of descretelocations. If I thought about it I could have written a function toconvert the data, but I was feeling a but fuzzy and asking the LLM todo it was faster:
I have a list of values in order with gaps: [1,2,3,4,5,6,10,11,12]Looking over the code it was easy to tell that it was doing what Iwanted.Could you use run length encoding to convert it to a list of [start,length] pairs?
I've been working on a pedalboard and wasn't surewhat methods people typically used to attach the pedals. Claude gaveme an overview that was a good starting place for looking more deeplyinto each option.
I wanted a file format for human-writable machine-readable datathat included key-value pairs and multiline strings, preservedformatting when round-tripping, and wasn't YAML (security,coercion, complexity). Claude helped me identify TOML with tomlkit as doingwhat I wanted.
I often write code that assembles some data and then want tograph it. Pasting the code into Claude along with a request for thekind of plot I want typically produces something close to what I hadin mind, and a good starting place for tweaks. It can also be usefulfor tweaking existing plots: "In the following code the bars for themedians in the violin plots are blue. Could you make them black?","Could you modify the following plot to lightly shade the periodbetween 2023-12-01 and 2024-05-01, labeling it 'study period'?","Could you extend the following code to add a trend line for eachplot?", etc.
Claude is very strong verbally, so asking it to rephrase thingscan work well. I'm reasonably happy with my writing in informalcontexts like this one, so this is mostly something I use when I'mtired and it feels like words aren't flowing well or if I need tocommunicate in a voice that doesn't come naturally to me (contracting,military, academic, etc).
I often ask questions like "what do you call it when ..." witha verbose description and then Google the technical terms it thinks Imight be gesturing at.
There are still a lot of things it gets wrong, but it turns out thereare many areas where I can pretty easily check the output and wheresomething mostly-right is a good starting point.
Comment via: mastodon
Discuss
