Building NietzscheBot: a chatbot that read the entire collected work of the late philosopher.

New advances in AI technology make it possible to pass highly specific instructions to language models like GPT-4 (ChatGPT). Using a software package called LangChain (maintained in Python and Javascript), you can teach a chatbot just about anything. For example, this chatbot can answer philosophical question in the style of Friedrich Nietzsche.

It’s also possible to teach the bot about your own documents, the documentation of any coding language, the collected work of any writer or philosopher, and more. Passages highlighted in bold are actual quotes from Nietzsche’s collected work. If only Nietzsche were always this wise!

The solution to ChatGPT ‘dreaming’

One of the larger concerns about ChatGPT is that it makes up (potentially biased) information when it’s not sure which answer to give. Building bots that can interact with documents solves this answer entirely. The ability for the bot to understand natural language remains intact, while the sources it bases its answers on are factual, based on real information which we know to be true.

It’s possible to instruct the bot to reference the information it uses to give an answer, so you can always fact-check whether the information was true or not. Imagine how useful this is. If you keep your personal or company documents organised in a handful of places, for example, you could build a bot that answers questions about all of your internal resources using natural language. Never again would you have to search for that one piece of sales copy, that memo, that presentation. Just ask what you want to know and your bot will automatically answer the question, while pulling up the right document. ****

LangChain will impact the educational sector

Many schools and educators have expressed concern lately about students using ChatGPT to write essay’s and answer test questions, with the obvious consequence in mind that students won’t learn as well. While slacking off is certainly one use of the tool, consider the incredible didactic potential this tool holds instead:

• In many European countries, there is a huge shortage and demand for teachers.
• Because of this, as classrooms grow larger, students get less and less of a personalised and high-level education.
• Limited access to (expensive) private tutoring leads to increased income inequality in the long run.
• It’s entirely possible to build a chatbot trained on course materials, that uses didactic methods and presents information to a student in a way that they can understand it.
• (If you work at a school or European national education institution, and want to know more, let’s have a chat!)

‍

LangChain integrates with the internet of things.

In my next post: you can integrate LangChain with Zapier, Google Search, or any web service with an application programming interface (API) that you want. This means you can literally ask your personal assistant chatbot to send a tweet, create an instagram post and present it to you before posting, to summarise a blog post, to order groceries, to control your smart home, send a file to the printer, simply by asking.

While we have to be extremely careful which services we integrate our chatbots with, the potential is enormous. Connect with me if you want to learn more about the potential implementations, future, dangers, safety and societal impact of generative AI. Have a nice week!

‍