AI chatbots are a nightmare for the privacy-minded. Sure, you can have a good chat with artificial intelligence and see how far language processing has come in recent years, but where does the data go once you've sent it off? Do all of your queries get added to a giant database to train the AI further?

The best way to solve this privacy fear is to have a chatbot that doesn't use an internet connection to parse its answers. And now, you can make one yourself. Someone has got a "GPT-like" LLM running on a Pi, and while you need an internet connection to get it set up, you can run it offline once it's ready to go.

A local chatbot on Raspberry Pi

You can see all the details on Nick Bild's GitHub project page. As you'd expect, you need to do a little bit of setup. You need to download Ubuntu Server to the Raspberry Pi SD card, connect a microphone and a speaker, and then wire up a button. Then, once everything is set up, the device requires an internet connection after the first user query, so it can grab the speech-to-text model required to understand you.

However, once you have all this set up, you can disconnect the device from the internet and use it offline. Nick Bild does state that it will take around 15 seconds to process and respond to your queries, and this will get even longer if you ask it to perform a more complicated task. However, as far as Raspberry Pi projects go, it's easy to build, quick to set up, and can be genuinely helpful in day-to-day life.