Sim-Link

Finally made it to the first milestone - linking something in the real world with an equivalent in the simulation. In the video below you can see Jack turning a lamp on and off. It's physical counterpart follow suit.

(You may also notice a small white circle around the switch - this represents the sound of the switch to us the viewers; soon Jack will actually 'hear' it.)

With this proof-of-concept in the bag, we'll move on to modelling a full computer keyboard; and you can probably guess where that's heading. At the same time we'll be reaching out to some social media sites get their views on Jack being a user. On that note, please get in touch if you can help out with that - I know from experience that it's pretty hard to hold of someone savvy at a big corp.

Milestone Approaching

Made some good progress towards the project milestone of associating a simulated light with a real-world light. The association will be such that if the simulated light gets turned on, the real one also comes on. If the real-world light is then switched off, it goes off in the simulation as well. The video below shows some progress towards that goal - specifically using inverse-kinematics to move a hand to a light switch.

This may seem like arbitrary functionality, but it demonstrates one of the key areas of the project - the synchronous nature of the simulation.

OCR

This is the first project update since January - but I haven't been slacking, I promise! At some point back then I realised that Jack (which I've named one of the MindLoop entities) would need to be able to read - which from a software point of view means OCR. It turns out that if you only spend 2 hours a week on it, OCR takes about 8 months!

I've been working exclusively on the OCR proof-of-concept as the capability means lots of exciting possibilities and a more 'authentic' AI overall. As you can see above the code is performing quite nicely - it hasn't correctly identified all of the numbers, but I'm calling this 'feasibility study' a success as I'm confident that with more time, much better results will be possible.

From now on I hope to have a few project areas on the go at any one time, so there should be more regular updates in future.

Stars

Finally managed to get back to the mind-loop project after a number of distractions! To ease into the 3D code I've added a star field around the Earth. In the picture we're looking 'up' at the Earth from below the South pole - accordingly the Pole-star can be seen in the middle of the picture and the Big-Dipper off to the right. Many thanks to David Nash for making the super easy to use HYG Star Database available.

Polling for Rain

Only a couple of progress points to report at this time.

First is the use of the great weather api at www.wunderground.com to get current conditions at a latitude/longitude. This is perfect for the purpose of having near-realtime weather conditions at the location of an entity.

The second area of progress  was in writing a multithreaded scheduler. At the environment-simulation level, the scheduler will handle polling of 'local' weather conditions etc, as mentioned above. At the information-processing level, it will be used to queue 'packets' of work to be sent for processing. Assuming we can figure out how to integrate the results into the current 'thought-process' as they are completed, this should give us simple multi-tasking.

Future Headlines #1

'AI plays text adventure like a boss...'

Dot-to-dot

Settled on a 'resolution' for the 3D environment and did a proof of concept for lat/long to 3D and accuracy at centimetre level - simple video here : http://www.youtube.com/watch?v=FGU-DArH1DY