ChatGPT's Wolfram Superpowers: Wolfram GPT (Formerly the Wolfram Plugin for ChatGPT)
Vložit
- čas přidán 19. 06. 2024
- This webinar recording showcases the Wolfram plugin that gives ChatGPT access to a large selection of curated content from Wolfram|Alpha and options to use the computational strength of Wolfram Language.
Learn about the plugin's development and see some remarkable responses from ChatGPT with the Wolfram plugin activated. ChatGPT with its "Wolfram superpowers" provides the combined strength of two powerful tools. The video also includes a Q&A session where the presenter answers selected questions from the original webinar audience.
Presenter: Alan Joyce, Director of Wolfram|Alpha Socioeconomic Content Development
Chapters:
0:00 Welcome
0:33 Plugin motivation
5:56 How to install the Wolfram plugin
7:40 Simple math calculations and data lookups
13:53 Plugin workflows
16:20 Examples with built-in prompts, non-English languages and data visualization
21:29 Audience Q&A
Relevant Links:
wolfr.am/ChatGPTWolframPlugin...
www.wolfram.com/wolfram-plugi...
writings.stephenwolfram.com/2...
In a few months it will become the best matemathician ever.
The big problem is that even if Wolfram can do something, chatGPT often can't figure out the right syntax to do it. Probably there were not nearly enough samples in the training data so that it knows how to correctly interact with Wolfram Alpha / Language.
Simple example: "Give me a bar chart of the 10 most valuable currencies in the world." It fails.
If you instead use the browsing mode and ask: "Generate Python code for a bar chart of the 10 most valuable currencies in the world." you get exactly what you want.
Also, I feel like some of the data in Wolfram are really out of date like month-over-month inflation data hasn't been updated in over a year, as far as I can see. This is very old compared to data you can get through a simple Google search nowadays. So it is really not clear to me if data is a strength of Wolfram, compared to all of the internet that chatGPT can access with the browsing mode.
You can fix this by running GPT in a loop and getting it to respond to its own errors.