Hacker News new | past | comments | ask | show | jobs | submit login
How are you using AI?
10 points by Sfjournalist 3 months ago | hide | past | favorite | 10 comments
'm a journalist working on a story about quirky ways SF and Bay Area locals are using AI.

Have you used to code for you at work or write your performance reviews? Compose a break up text message? Create a meme for a inside joke? Pretend to be a lawyer and fight your landlord over email in legalize?

I would love to chat to anyone who lives in the Bay Area and has an interesting story.

I can offer anonymity to anyone who feels the way they use it is sensitive.

The piece is for the SF Gazetteer, a new local publication. I have written for Wired, TechCrunch, New York Times and others. Feel free to check out my website. https://jesseerynklein.com.




I disabled my cofounder matching account due to how many people seem to think LLMs are a good idea for many things they just aren't good at. The race to the bottom for customer service is definitely afoot right now.

The most useful and valuable feature of LLMs imo is using the embeddings to enhance document search. There are smaller models that are just as good for this though.


Which specific smaller models?


Not in SF but I liked the topic because recently I started using LLM in a very interesting way and I would like to see if that's something others are doing too.

In place of adding packages to my projects, which I would probably not use the whole of it, I ask ChatGPT/Sonnet to generate a subset of that package that fits my needs and grow it from there.

In same way I've removed literally all packages in a web app I'm working besides tailwind which I'm also thinking about how to remove next.

The biggest upside of this strategy is that I understand everything that is going on in the code and another one is to avoid npm packages that is a big security risk.

Another benefit is package size, the app I'm working was getting to a final bundle of 400kb, which isn't too big but now it is 150kb with exactly the same features but without multiple frameworks and packages that I wasn't using 10% of it.


Good morning, I'm not sure if my story is unique, but here it is. I am an AI research scientist, and we conduct a lot of research and development on the OpenAI platform. For some of our R&D projects, we have strict requirements and cannot use any public AI services. So, I created a personal AI box that can provide the same ChatGPT experience while addressing three key issues:

    Hallucination
    Forgetfulness
    Long, essay-like responses
and many other lingering issues. In any case, we are now using our personal AI box for R&D and daily queries."


I use it for quite a few things and work with it as well, but specifically right now I'm using AI to turn a poem I wrote into a song and illustrate a video for it. Only after it produced the song I got a little offended that an AI is singing my words, so decided to layer my own vocals over the AI music. The problem is I can't sing, so now I'm looking at singing classes. The project is taking longer than originally expected.

I wonder if ChatGPT's Advanced Voice Mode could provide singing guidance.


Are you based in the Bay? would love to chat via phone if you are.


Oh no, sorry, I should've specified I'm in Sweden; just felt like engaging with the question.


No problem! I appreciate learning anyway!


I’m not. I’m still waiting for it to overtake Google… ;)


I am using for improve my code.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: