Hacker News Reader

Comment by hellcow | Hacker News Reader

This weekend I used ChatGPT to:

1. Help select a data structure for a very specific use-case given certain constraints, generate a first pass of the code, and help debug it.

2. Recommend a recipe for the exact set of ingredients I had on hand. (It was for muffins, and they were delicious)

3. Diagnose and fix an issue on my stovetop that’s plagued me for months.

In some of these cases it dramatically reduced the time I would have spent searching/researching from hours to a minute. Search is great for some use-cases and not great for others. I’m glad DDG is supporting more use-cases with an eye toward privacy.

View on HN

I'm glad that it's helped you, but there's a big difference between an individual proactively using AI and a service forcing it on its users.

I guess I don’t understand why you feel they’re “forcing it on users.”

Just don’t look at the results it generates or use ublock origin to filter the AI response off the page?

I don't think it's so much about DDG specifically but of all the services now rushing to introduce AI into their flow.

On the flip side I tried to get chatgpt to help me generate a pandoc metadata.yaml file to convert a bunch of markdown files into a single docx file.

I tried to keep it simple to start with so I could incrementally build it up so started with wanting double line spacing and the content of each markdown fike to start on a different page.

I gave up in the end and had to do the research myself as chatgpt was massively hallucinating and throwing random stuff in I didn't even ask for.

> throwing random stuff in I didn't even ask for

Sounds like ChatGPT has officially become a professional software engineer.

Turbo3.5 or 4? 4 is substantially better at code generation. To the point where I would never try to use T3.5 to generate code.

I feel like we should all start specifying in our comments which model we are talking about when we are talking about ChatGPT.

You had the knowledge to discern whether the info given to you was legitimate. The average user doesn't have that. What if I need to look into something that I have zero knowledge about? When I get a list of sources, I can look through those sources and at least partially determine whether they're reliable. If an AI just spits an answer out, do I always just trust it?

Search engines are getting bad enough about throwing crap information at me. I think AI will just increase that. I prefer that I have the power to sift through the sources. I'm not willing to hand that completely off to AI.

I see your point, but is it much different than googling and getting a site in the result with wrong information or advice?

It's different in that you'll probably look at more than one source if you're searching the web. The sites with bad info become pretty obvious when you've looked at more than a few different sites on the topic.

Reply on news.ycombinator.com