TOTW 22: Ask Eve AI – Humanising Your Information

In TOTW 20, we introduced the first business case we’re working on: Ask Eve AI. I mentioned it was all about humanising accessibility to information. But what exactly does ‘humanising your information’ mean? In short, Evie (the AI behind Ask Eve AI) allows people to ask questions about your company information using the language they know best – their own.

To enable Evie to answer questions based on your information in human language, we need to tackle a few elements:

1. Understanding the information you have available.
2. Understanding a question asked in human language.
3. Linking that question to your information.
4. Looking at the relevant information and generating an answer, again in human language.

The exponential advancements in AI over the past few years have enabled us to automate these steps in ways that were simply impossible 1-2 years ago. In a nutshell, this is what Evie assists you to:

• Help prospects or customers search through all the information on your website, help files, and manuals.
• Assist staff in finding information on the intranet, wiki, or any other internal information repository.
• Aid researchers and internal experts in searching through their own collections of information.

Using what they know best – their own language.

Why Not 1-2 Years Ago?

To understand why this wasn’t possible 1-2 years ago, we need to dive a little deeper into what’s happening today. Until now, humanity has been improving the ways we communicate with the incredible machines we created – computers in all formats. Programming languages are the building blocks of our digital world, enabling us to perform increasingly complex tasks. In 1843, Ada Lovelace (yes, the programming language ADA was named after her) was the first to realise that numbers could represent other things, or better yet, all things. Her work laid the first step in designing programming languages to instruct machines. In 1936, Turing introduced the concept of a universal machine that could follow instructions, the precursor to our modern computers. From then on, we’ve seen a fractal evolution of programming languages and concepts, adding layers of abstraction and enabling us to automate more complex or specialised tasks.

Now, the tables have turned. Machines have learned how to communicate with us. AI, in general, and specifically technologies like Computer Vision, Text-To-Speech, Automatic Speech Recognition, and Large Language Models, have learned to understand us humans. Human language is becoming the next programming paradigm. It still requires engineering – not software engineering, but prompt engineering. Prompt engineering (which is not as easy as it might seem) instructs a Large Language Model or other generative AI to perform tasks.

Meeting in the Middle

Today, it’s not yet possible to rely solely on either software engineering or prompt engineering to accomplish what Evie does. Evie requires both, using the best of both worlds. This dual approach is part of what attracted me to this endeavour. I believe this will remain true for the foreseeable future. The art of creating good programs will remain important but will drastically change as powerful AI assistants for writing code continue to improve. The art of prompt engineering will be added to our toolkit. Currently, I have about 10 ‘algorithms’ implemented using prompt engineering. It’s like a happy marriage, where both engineering techniques have their strengths and weaknesses. By combining both, we can achieve what was unthinkable not too long ago.

The Future

The future will unveil fantastic new possibilities, at an increasingly faster rate. And I for one am eager to explore them and seek out business cases from which everyone can benefit. A new computing paradigm is rising, one based on tokens, not just computing cycles. Some voice go as far as to say the next operating system will be token-based. For now, I’m quite happy talking to Evie using my own language. In the coming weeks, we will uncover more of our work. Stay tuned!