Content-type: text/html Downes.ca ~ Stephen's Web ~ ChatGPT Gets Its 'Wolfram Superpowers'!

Stephen Downes

Knowledge, Learning, Community

As has been previously noted, chatGPT functions by predicting the most likely 'next word' or 'next sentence' based on its large language model. This makes chatGPT responses sound plausible (they're literally 'just what you would expect') but frequently factually incorrect. Giving chatGPT better data (and weighing it more heavily) allows it to draw on a storehouse of facts as well as what most people would say in a given circumstance. That's what's happening here with the Wolfram plugin. It "gives computationally accurate answers to ChatGPT queries using Wolfram's data knowledgebase and language. Custom visualizations are given as well." So now chatGPT can sound a lot like your best friend but have the background computational and data skills of an Einstein. Still don't think it can teach?

Today: 8 Total: 1124 [Direct link] [Share]


Stephen Downes Stephen Downes, Casselman, Canada
stephen@downes.ca

Copyright 2024
Last Updated: Apr 24, 2024 8:42 p.m.

Canadian Flag Creative Commons License.

Force:yes