Project MoneyPenny

Project MoneyPenny is my attempt to build a AI assistant with a voice interface that serves the following needs not currently met by what I see on the market.

  • Truly Personal

  • Truly Private

  • Truly Extensible

  • Customized to the use cases where “I’ need it and not the general public.  

The point here is not to sell me a product or pass a Turing Test, but to create an AI that lives in a layer closer to my personal life than what is on the existing market.

Truly Personal

Right now the Alexa, Siri or Google home HEX ( Human Experience) is the same for everything. This really doesn’t feel like AI so much as just a voice UI. In fact the only intelligence application for these assistants seems to be in voice parsing.

Why is this?

At this time I’m not sure that we can make an AI that is truly personal and truly profitable. The training set for truly personal events is extremely limited and possibly creates the need for someone to manually create AI for the use cases they need. I just don’t see this happening in the short term.

The ultimate test?

Can I develop an “inside joke” with an AI assistant.

Truly Private

This speaks for itself. If I want something to be truly personal, it has to be truly private. While I trust apple on the privacy of Siri. I don’t have faith that they will build the product I want and let’s face it -- this is about the journey and learning more than money.

Truly Extensible

Can I modify and deploy this to other things I make? Can it clean my home via a roomba device? Can it run a robot I make? Can it interface with other API’s and AI assistant to get me generic info on a truly private layer?

Customized to the use cases where “I’ need it and not the general public.  

This may seem hand in hand with being truly personal. But really where I need help is not in googling celebs or making calls or the like. I really need it to help me out on science, engineering and product management questions I also need some extra help around personal shopping ( I think the echo show is heading there, but I think we can do more) 

How do I get started?

These are the threads I’m going to start working with.

  • Define use cases

    • Would like to be able to ask it s subset of analytics questions and get answers similar to wolfram alpha but with voice obviously

    • Voice imprint unlock

    • Personalized Calendar Support ( More on this later )

    • Being able to leave a note/ for my AI assistant on my calendar, to-do list or document  

    • Be able to have MoneyPenny take a note for me when I wake up or whenever. Essentially let it record a voice journal and store it securely and privately

    • Guided meditation

    • Management of my diet and exercise routine.

    • Advanced time management feature ( More on this later)

    • Figure out how we’re going generate voice output

    • Being able to display info back from the voice assistant on a screen

    • Morning and even walkthrough support ( More on this later)

    • Garage/Robotics knowledge base support ( Can this thing actually help me make something else?)

    • Inside Joke Feature.

  • Define Advanced use cases

    • Camera Support for Image recognition

      • Derived use cases from this.

    • Extensibility to Other AI assistants.

    • Extensibility to Robotics ( Note: Avoid creating Ultron)

  • Select Voice Parser (done)

  • Select Cloud Service to run on (done - the previous decision locks us into Google compute engine)

  • Get better at TensorFlow( In Progress)

  • Design ML pipeline/deep learning architecture.

  • Pick initial use cases and design neural network

What am I going to call it?

Was thinking “Naomie” after Naomie Harris who plays Eve Moneypenny or just moneypenny ( though to date that’s been reserved for a (theoretical) dog

What would be super cool one day?

I’m able to hand down a truly personal assistant to my (currently theoretical) kids as a way of passing my knowledge guidance and experience down through generations of my family. Almost as a “generational steward”

Note to self:

Don’t create AI singularity in the process.