New project: Creating an AI Mind

Hi everyone!

I’ve started a new video series to regularly publish my development progress on this new AI project.
This is the first intro and walk-through of the concept.


Cool concept Nuno and best wishes! Looking froward to the progress.


Part 2 is up! A demo of how the summarizer works, plus a few features of the chat interface which will help in future fine-tuning

1 Like

Part 3 is up.

I’m late with the next video, but rest assured, the project is progressing nicely. In the next episode I’ll explain why I’m taking so long to do an update and what is new.




finally, an update


Exploring an idea on how to have the AI learn a new behavior

1 Like

Inspirador Nuno :rocket:. Continua :muscle:

1 Like

Do you have written content?

I would love to read your ideas, but can’t learn from video.


1 Like

I should! I also can’t learn from video :sweat_smile: unfortunately I don’t have time to put it in writing in a proper form, so video its the best I can manage for now

Great news :smiley: fine-tuned models are finally working

1 Like

have whisper do it and then summarize your videos :upside_down_face:


My AI assistant will greatly contribute to my already existing peak-laziness :smiley:

1 Like

This new tool support is quite cool and a fast way to add new features. I can now have the assistant check my calendar to search for stuff or tell me what is happening

Massive update!! :smiley:

  • langchain-style tools
  • fine-tuned babbage
    and more… :sunglasses:

while I’m still sorting out the next steps, I made some progress. Now I can access my assistant from anywhere via the phone :smiley:


Been following your series, did not watch all of the episodes but each one really is food for thought!
Very impressive project.

It seems a Telegram implementation is a natural implementation/continuation of such ideas. A lot of people, including myself, are working doing it. Glad to see you too :slight_smile:

Keep going!

Hi Nuno!

Very intrigued by your “memories that need to be saved” feature. How does your algorithm decide what data is relevant/important enough to be saved? Is it just using the embeddings or is there something else?

the model decides :slight_smile: I’ll address that in my next video to give a better explanation

1 Like