🤖 Day 11: Generate test data using AI and evaluate its efficacy

📆 30 Days of Testing
, , ,
46

Excellent!

One option that might address your privacy issue is to host an open-source LLM within your organisation or on your machine.

Using LlamaFile (GitHub - Mozilla-Ocho/llamafile: Distribute and run LLMs with a single file.) from Mozilla is probably the easiest approach to try this out. It provides a set of open-source models packaged up to run on a local machine. They provide a ChatGPT like interface running as a local webserver against a locally downloaded LLM. So your data is kept private.

A few things to note:

  • You’d need to be able to download and run arbitrary exes on your machine so may not be allowed on some corporate machines.
  • the best performance is likely to come from the larger models but these may be too large or run slowly on a desktop.
  • you are accessing the base LLM model and so it may not have all the additional features and controls that you find with ChatGPT, Bard, CoPilot and the like.

Have fun

1 Like