It started as an experiment; I wanted to build a chatbot from scratch. The main goal was not so much the chatbot itself, but more to gain knowledge. Soon the idea went from an AI experiment to a project with a real goal and application. The idea came from a TV documentary series "AI Love" that I was asked to participate in. So I thought it would be fun to create an AI clone that can train itself and in turn teach me new things and assist me in my daily life and my work.
So this year we are going to create an AI clone and you can follow this process and also help. I have been working on it for a while and now have a pretty good overview of how it works. That is why I can now make a good estimate for a step-by-step plan. In addition to the step-by-step plan, I will share every line of code with you. Do you have no experience with programming or AI? No worries, we do this together, and I can use your help in another area. So, can you join this project? It will be fun and certainly very educational.
For this first step, we are going to create our own trained neural network. This is a network that can learn by itself. We are going to do this using our own framework, which we are going to build ourselves. This is difficult work, but it is also very educational. The first step is to investigate how we are going to do this. We are going to program this in the C language.
To train the network, we need a dataset. This is a collection of questions and answers. To make our work easier, we are going to have the interview generated by ChatGPT or another AI model. This will be a web-based form that I will make available out-of-the-box. We will do this with HTML, JavaScript, CSS and PHP.
Tools are available for this that you can use to represent yourself in 3D. These can be tools that you can download and use, but also tools that you can use via the cloud. To animate the 3D model and make it speak, we will use Blender 3D, a 3D modeling program.
This is a rather difficult step. We will use an AI model or a web API that recognizes speech and translates it into text. In addition, we need a text-to-speech conversion. This is a process in which the AI converts the text into speech.
With a robot body, we can give the AI clone a physical body. We can do this using a 3D printer or other materials and tools. It would be nice if the robot body had a screen on which the AI clone could show its face.
In addition to all these steps, we might also need a cloud server to run the AI clone. This server would allow us to use more powerful AI models and process more and faster data. The computer on which the robot runs is most likely not powerful enough to handle these processes.
The interesting thing about this step-by-step plan is that you don't necessarily have to do it in order, so choose what you prefer. We will make all software available as open source software. This means that you can use it, modify it and improve it.
Let's work together to create an AI clone that can really be useful in everyday life. Will you join us?