I think using LLMs with RAG (aka tools) is more useful and reliable than relying only on training data that the model does its best to represent.

For example, using a search engine to find results for a query, downloading the first 10 results as text, and then having the LLM answer subsequent queries about those sources, or another example would be uploading a document and having the LLM answer queries about its contents.

This is also advantageous because much smaller and quicker models can be used while still producing accurate results (often with citations to the source).

This can even be self hosted with Open WebUI/ollama.

I havent really used any in a serious manner. I did install DeepSeek on my PC to try out. Its pretty fun to play with, but still seems to have issues. For example, I was using it to create bread recipes and fine tune proportions to get the exact amount of dough I need. I found that its math was way off, and would give me wildly different results even when asked the same question and given the same requirements.

Im not sure how ChatGPT compares, as I dont have access to it and Im not really willing to pay for it.

Honestly, my favorite use case for ChatGPT is as an Internet search engine. Google has become so shitty that I outsource it's main job lol. I just tell ChatGPT to send me reputable sources as links for the query, and I skip the bullshit.

It's also not a bad way to generate SEO friendly descriptions for eBay listings, if you have a lot to list and are lazy. You can move a lot faster and get better results than using the default ai that site has. It would be ideal to personally write everything and be an SEO expert, but you are mostly guiding people to see the photos and just need the metadata perks of the jargon.

Writing emails, brainstorming

Website building

Might be worth checking !LocalLLaMA@sh.itjust.works which can have tons of useful inputs

It's really good in statistics, but you need to know enough statistics to know what to ask. Just today I needed to write a PyStan script for doing some MCMC and it's helped me to write it, structure the data and understand the results of the experiment. Then, it confirmed my suspension that the chosen model was not very good for my data and tomorrow I'm trying with another probability distribution.