Hi HN,

I've been reading Nick Bostrom's Superintelligence and I've been thinking (dangerous, I know). I wonder how far we are from creating some pretty intelligent robots by combining the tools we already have. Here are some ideas:

Hardware: Raspberry Pi or similar micro computer head. Attach a wifi dongle (or maybe 3G as well?). Create an API on EC2 to communicate with. Use wheels to move around and attach a webcam, microphone, and speakers. Probably need an arduino to control the wheels and whatever other sensors you have attached.

This is a latency vs cost issue, I think. You could create a computer in the robot and have it process things without sending back and forth over wifi, but you get more processing power with an API on EC2 and it's more feasable to rent out that space rather than buy parts.

Features:

- Webcam

  - Realtime Facial recognition. Something like this: https://www.mashape.com/lambda/face-recognition#!documentation
    
  - to add a face you can set it in meeting mode and take some pictures and add them to the training dataset.
  - store info in a database with timestamp:name.
  - Realtime Object recognition. Similar to above. Store it in a database that has a timestamp and a list of objects seen.
  - Text analyzer/handwriting
  - use: solve math problems (send text to wolfram alpha, it'll evaluate it and send it back)

- Microphone/speaker: - text to speech to talk back to you. - voice recognition - look stuff up on Wolfram Alpha or similar service: has basically every fact you could want.

What other stuff could you see added into this? Just from these ideas, which I think would only require time and a little money to put together, I see an AI with at least a grade school level of intelligence.