Practical Examples

Some of the easiest ways to get started involve using OpenAI GPTs

Google put together 101 Use Cases, examples of leading businesses using LLMs

Video Scraping

Simon Willison didn’t want to manually add up a bunch of numbers across several emails, so instead he did a screen capture of scrolling through his emails and fed the results into Gemini, retrieving the answers for under 1/10 cent of API time.

Investing

from Every.to

In this episode of How Do You Use ChatGPT?, Dan Shipper and IA Ventures managing partner Jesse Beyroutey tell us how they outplayed Wall Street. Dan and Jesse invested in Nvidia in 2019 when the shares were trading at $33. They’re worth nearly $900 a piece today. It was the best trade of their lives. In this interview, Dan and Jesse walk us through how they used Google’s powerful LLM Gemini Pro 1.5 to try and top that.

To get them started, Dan opened an account on investment platform Robinhood and put in $1,000 as the principal amount. With 90 minutes on the clock, Dan and Jesse were off to the races.

This interview is a masterclass in how to use AI to refine your own investment thesis and make smarter financial decisions. I came away inspired—and more than a little tempted to test my luck and day trade!

Quill AI - Use LLMs to speed up the equity research.

Research Assistant

Lemire shows how to use ChatGPT as a research assistant: 1. Query a document 2. Improve text 3. Idea generation 4. Grant application 5. Writing code 6. Find reviewers and journals

Use Gemini 1.5

Google’s most powerful model, Gemini 1.5, normally costs $20/month (same as ChatGPT.) But you can try it for free using their developer site: https://aistudio.google.com/

Gemini allows an incredible 1 million token context window, allowing for a much larger input “prompt” than other engines. (ChatGPT has a 128K context). A million tokens is enough to store most good-sized books (e.g. the Bible fits in less than 800K tokens), letting you ask detailed questions and get answers pointing directly to the relevant place in the text.

A typical use case: upload all of your company’s customer support documents. Gemini will respond with the answer and the exact page of the document, thus making it far more reliable.

Furthermore, unlike a typical document search, you can ask related questions, even if you don’t know the specific keyword. For example, if you had all the documentation for a car manual, you could ask questions like “Do cars before 2010 support Windows Media audio disks?”. The system would know to look only at manuals before that date, and search for content related to audio. Importantly, none of the words you’re searching for need to appear in the question itself.