The use of Large Language Models (LLMs) can be extremely helpful for many repetitive business processes such as data interpretation, writing business texts, programming, or configuring regular expressions.
Providers on the internet, like OpenAI with ChatGPT and others, offer highly useful services for these purposes, but they are associated with potential risks:
- If you phrase your requests poorly, it is possible to inadvertently disclose customer data.
- In some cases, it may be unavoidable to include business information in your queries.
- These services are only available online.
I asked ChatGPT how to avoid leaking out sensitive data while still using LLMs. In response, I received many useful tips and was pointed to the availability of local LLMs, among other things.
One recommended tool with easy usability is GPT4All, which can be installed on Windows, Mac, and Linux. (Another option would be Ollama, which is more targeted toward programmers.) GPT4All (and Ollama) run LLMs locally on your own computer. This eliminates the risks to your own data.
Of course, local models also come with disadvantages:
- The best models are not available as they are retained by the providers for their own services.
- Running the models is extremely memory- and processor-intensive. Older computers quickly reach their performance limits, and sometimes you need to be very patient before receiving a response.
After installing GPT4All, you need to install a model (or optionally several). GPT4All provides a simple interface with a long list of available models.
Choosing the right one is not easy. But ChatGPT can help here again. You can ask which model is best suited for your specific query, and you will receive appropriate recommendations. The models download and install quickly and are then available on your own computer. After that, there’s nothing standing in the way of safe experimentation.
Quick Tip: If you need help operating the tool, simply ask ChatGPT or the local models themselves.
