Comment on Don't fall into the anti-AI hype
blarghly@lemmy.world 8 hours agoFrom the article, literally one line above the line you quoted:
In the past week, just prompting, and inspecting the code to provide guidance from time to time, in a few hours I did the following four tasks, in hours instead of weeks:
- I modified my linenoise library to support UTF-8, and created a framework for line editing testing that uses an emulated terminal that is able to report what is getting displayed in each character cell. Something that I always wanted to do, but it was hard to justify the work needed just to test a side project of mine. But if you can just describe your idea, and it materializes in the code, things are very different.
- I fixed transient failures in the Redis test. This is very annoying work, timing related issues, TCP deadlock conditions, and so forth. Claude Code iterated for all the time needed to reproduce it, inspected the state of the processes to understand what was happening, and fixed the bugs.
- Yesterday I wanted a pure C library that would be able to do the inference of BERT like embedding models. Claude Code created it in 5 minutes. Same output and same speed (15% slower) than PyTorch. 700 lines of code. A Python tool to convert the GTE-small model.
- In the past weeks I operated changes to Redis Streams internals. I had a design document for the work I did. I tried to give it to Claude Code and it reproduced my work in, like, 20 minutes or less (mostly because I’m slow at checking and authorizing to run the commands needed).
Kirk@startrek.website 8 hours ago
That doesn’t explain what the LLM actually did to help. It’s like saying “I used ChatGPT to plan my wedding menu” without any details. What did it actually do? Why was it helpful? Those are the things I continue to not understand about these tools.
blarghly@lemmy.world 8 hours ago
I’m confused why you are confused.
I feel like it is pretty clear the author said “hey AI, do this thing.” The AI made an attempt, the author clarified a few things and maybe made some edits, and then was satisfied with the result.
Like your example of planning a wedding menu. I’m not sure where the ambiguity is. If someone said “I used chatgpt to plan my wedding menu”, I assume they prompted it something like “plan my wedding menu. I want something classy but cheap. No fish.” Then chatgpt spat out a few options, they provided feedback - “I dont like broccoli either” - and then they picked an option they like.
Kirk@startrek.website 3 hours ago
It seems we agree on the facts, but not on what “useful” or “helpful” means. Ihonestly have never, ever considered deciding on food to be labor, but in the interests of replying in good faith I asked an LLM the exact prompt you gave. It gave a long, detailed reply, but here is the first part labeled “1. Welcome / Cocktail Reception”:
I want you to consider that this not actually helpful in the slightest, and is fact creating more work. Consider: is there a vendor nearby that has these items as an an option for event planning? Is this a recipe that even exists? Does this information further my mission of having a wedding in any conceivable way?
Senal@programming.dev 7 hours ago
Or Perhaps:
they mention reinforcement learning, pre-training and other general LLM concepts, but none of these are related back to the tasks they are talking about.
The point is, there was no explanation of how any of this was achieved, which can lead to confusion about what was actually achieved.
The LLM wrote some docs vs the LLM rewrote the library from end to end are very different things.
It’s very much a “Don’t give up on X, look at what can be achieved” but without any actual details on what is required to achieve those results.