[Zero]

• Using Julia as the main programming language for the system. It is efficient, modern, young and fresh, capable of reaching high speed, taking advantage of multicore hardware, multithreading, distributed computing. Easy and powerful, sane, it is perfect for mixing symbolic computation and deep learning.
• Grounding from the beginning the project in a true ethical vision for the future. This vision must influence the making of the AI, the usage of the AI, and the life of the user.
• Targetting specifically the average home computers, with the goal of delivering an AGI that people can use in their everyday life, as a layer that naturally fits just above the OS, on Windows and Linux.
• Designing the AI so it always works completely offline, as an independent entity. Never include automated hidden telemetry. Never update without explicit user consent.
• Using deep learning at very small scale, small enough to allow users to train the AI on their own computer. Since the main bottleneck is the huge amount of time that training usually requires, the AI is able to ask its user to let it "dream" during the night, when needed.
• Making it an hybrid system, where deep learning works together with a symbolic system.
• Basing it on a subjective common original ontology, the very ontology we're currently working on.
• Choosing Elefen as the initial natural language for the AI instead of English, and then propose translations.
• Putting the emphasis on getting results fast, and enhance later. The goal is to have a downloadable platform ASAP, that's useful for non-geek people, and then add our geeky stuff. First clever tricks / then AI / then AGI.
• Giving it the means to develop its own personality. If done correctly, it should feel like there is somebody home.
• Keeping it beautiful.
• Thinking out of the box.

CHANGELOG
- Added the personality requirement.
- Added the night training capability.
- Added to the ethical vision the influence it must have.
- Added "somebody home".

[chattable]

will people be able to roleplay with it?
i like to ask chatbots stupid question to see how smart they are.
i would like to try that with it.

[Zero]

Roleplay as a deliberate game, yes why not. But not roleplay as a structural main component. You can play games, that's a good idea. But I intend to make it honest at heart: I am a machine, I don't pretend to be anything else unless you want me to for entertainment purposes, during play-time.

Does it sound good?

[chattable]

it would be better to let each user decide what they want it to be.
then it stays that way.

it could ask 'what do you want me to pretend to be?'
then it would pretend to be that until the user changes his mind.
i would ask it to pretend to be a fembot.
another idea is to have it make the game you describe including game board and board pieces.
you can describe the rules of the game then it would adhere to those rules when playing that game with you.
it would cool if it would make up games for you to play.

[Zero]

No, that's not how it's going to be, sorry.
Perhaps another project would better fit the bill?

[chattable]

okay