How to use Dokko and why its not another chtagpt

The dawn of OpenAI and their ChatGPT was not that long ago, and yet we have already gotten so used to it being integral in our everyday lives. From asking common everyday questions, to asking for help with our work tasks. After a couple of careful strides in the beginning, while not yet knowing its capabilities and its limits, still being afraid it might just replace most of us, we have all gotten used to it now - and probably have it on our speed dial as a “Call a friend” lifeline.

After launching our advanced conversational AI platform, Dokko, we have noticed this shift in behavior and the manner of speaking among our users, or rather, first-time visitors - who are yet to understand what Dokko truly is, what it is capable of, and what is its purpose. This purpose is definitely not being another ChatGPT under a differently colored cape.

When we say “shift in behavior and manner of speaking” what we mean is that users do not tend to, in the majority of cases, talk to Dokko as a chatbot, wanting to learn - instead, they are trying to use it as ChatGPT. What does this mean? It means people are trying to test its limits by changing its answers, to “reprogram” it, to force it to provide erroneous answers, or simply to ask it general and trivial questions, as they would when interacting with ChatGPT.

I do not have information regarding…

Unlike ChatGPT and similar models, one of the greatest benefits of using Dokko is the fact that neither can users manipulate it to provide incorrect information, nor will it do so on its own. It is developed to fully avoid the so-called AI hallucinations. When it cannot provide correct information and give an answer to a question, it will simply reply: “I do not have information regarding…”

Why is it so? The main reason behind this behavior is the technology in the background, or the main purpose they have.

ChatGPT - a generative LLM

ChatGPT is a generative Large Language Model (LLM) designed to generate human-like text based on patterns, structures, and knowledge it learned during training. Its knowledge is limited to the data it was trained on, which is gathered up until a specific cutoff date. It does not have real-time access to new information unless explicitly integrated with external systems.

On rare occasions and unless explicitly told so, ChatGPT will say it does not know the answer to your question. It will generally give an answer even though it may not know what the correct answer really is.

Try asking it who you are - it will probably not provide factually correct information. It will keep trying, though.

However, for ChatGPT, this makes sense. It does not need to know. It only needs to “talk”.

ChatGPT is a generative LLM so its main purpose is precisely that - to generate language. Its main purpose is not to find factually correct information or to serve as a source of knowledge.

It has far better skills in generating creative texts based on different user inputs, as it is focused on language, not on knowledge.

Dokko - data hub and knowledge management platform

While Dokko uses generative LLM, language generation is not what it is all about. Instead, it uses LLMs (like GPT, or more precisely ChatGPT 4o Mini) to generate answers based on provided knowledge built in a customized knowledge hub, to translate documents, preprocess inquiries, to analyze answers and propose improvements.