Has anyone ever actually used an LLM to do anything even remotely useful?

>>105614816 (OP)
Yes. Shitpost. Allow me to demonstrate. Large Language Models have been used in multiple domains for tasks with demonstrable utility. Examples include:

Software Development

Code generation and completion: GitHub Copilot, powered by OpenAI Codex, is integrated into development environments to assist in writing and debugging code.

Code migration: LLMs have been used to automate legacy code translation (e.g., COBOL to Java).

Healthcare

Clinical documentation: Ambient AI scribes transcribe and summarize doctor-patient conversations in real time, reducing administrative workload.

Medical Q&A: LLMs integrated into clinical decision-support tools to assist practitioners with medical literature retrieval and summarization.

Customer Support and Operations

Automated support agents: Used in enterprises to reduce human workload and increase response consistency across customer service interactions.

Internal knowledge base querying: Retrieval-augmented generation (RAG) systems using LLMs provide employees with answers from corporate documents.

Legal and Regulatory Compliance

Contract analysis: LLMs extract, classify, and summarize clauses in large volumes of legal documents.

Policy monitoring: Used to track regulatory changes and assess impact on internal compliance requirements.

Education and Training

Tutoring and question answering: Deployed in personalized learning systems to provide explanations, test preparation, and homework help.

Language learning: Used in platforms for translation, grammar correction, and interactive conversation practice.

Data Analysis and Business Intelligence

Natural language querying: Business users query structured databases in natural language using LLMs as semantic intermediaries.

Report generation: Automatic generation of executive summaries and dashboards from structured or semi-structured data.

>>105614829
Sounds like a load of bullshit if you ask me.

>>105614816 (OP)
I just like Pi's soothing voice.

Tech support. When the same user opens a ticket with the same moronic issue for the 20th time, I just tell ChatGPT to tell them to fuck off, politely.

>>105614816 (OP)
Yes

>>105614829

>>105614816 (OP)
Have YOU ever done anything useful?

>>105614816 (OP)
- it saves me times writing python scripts. in the time it would take to look up the library to do what i want i can get a prerendered script already from chatgpt.
- Ive had it make funny images for the lulz
- Ive had it recommend software to me to use as a use case like self hosting forums or time tracking stuff
- when trying to find a place in a video clip i can ask it based on the signs i see to find out where that clip was taken. (google maps sucks at this for some reason)
- had it take data and put it into a nice format. like i wanted to compare the timeline of two active games to remember what was released when between the two.
- had it explain some science stuff to me when i was bored. i didnt realize we were a barred spiral galaxy.

stuff like that.

>>105614816 (OP)
>Despite the bot's sincere promises, the offer was not, in fact, legally binding. Presumably, no Chevy dealers were harmed as a result of this viral prank.
>That being said, it has proved to be quite the headache for the chatbot's vendor, a tech startup called Fullpath that provides these customer service AIs to hundreds of car dealerships across the country.
>Fullpath, advisedly, has shutdown the bot on Watsonville's website. In spite of its viral contretemps, CEO Aharon Horowitz believes its AI fared admirably.

>>105614816 (OP)
i use it to ask random stuff i'm curious about, and then i can just delve deeper if i want to. super cool

>>105614816 (OP)
I used GPT to generate types for bunch of jsons

>>105614816 (OP)
I ask "How to learn...?"-type questions. It used to be really bad at this. Always sourcing those shady Pajeet blogspams.
But now it's pretty good.

>>105614816 (OP)
they are great for writing erotic fan fiction of star trek characters. i love it when mirror universe seven of nine travels to the prime universe to seduce a young and inexperienced female trill named Ezri. The stories start out rather vanilla but end with borg enhanced adult toys and mind rape by way of nano probes and a shared consciousness.

I went from basically 0 machine learning knowledge to building a functional fraud detection model that has a really high hit rate. It didn’t write any code for me, but it explained concepts and was able to tell me when the precision and accuracy scores looked eerily too good. It took a long time but it was really useful for learning.

>>105614816 (OP)
After working 20 years maintaining legacy code I can only refactor trash code into something useful, so it works for me.

>>105614816 (OP)
I am planning to design and develop system to book local DJs for your parties