Upgrade to Pro — share decks privately, control downloads, hide ads and more …

A taste of the digital future

A taste of the digital future

A very brief introduction to some NICTA projects, then a fun look at the state-of-the-art in brain-computer interface research from around the world. Originally presented at Offis eClub.

Avatar for Ricky Robinson

Ricky Robinson

March 28, 2013

Other Decks in Technology

Transcript

  1. Software from NICTA spin out Open Kernel Labs (OKL) runs

    more than 1.6 billion devices, such as mobile phones, worldwide. OKL was acquired late last year by General Dynamics, a US $23 billion listed company.
  2. © Bionic Vision Australia 1. Camera captures image and transmits

    data to an external, body worn processing unit 2. Data processed and sent to implanted system via external wire 3. Implanted receiver passes signals onto retinal implant 4. Implanted electrode array stimulates retina 5. Electrical signals sent from retina via visual pathway to vision processing centres in the brain
  3. Science fiction has a funny way of becoming science fact.*

    *There is some disagreement as to whether science fiction is a roadmap or a warning to humanity. In the interests of delivering a fun talk, I assume the former.
  4. Mind copying A Rough Taxonomy of BCIs by Application BCI

    Controlling physical world objects Mind sharing/ recording Mind control Artificial memories/ data store Remote control Behaviour control Assistive tech/ prosthetics Telepathy/ shared dreaming Systems control Knowledge upload Data mule
  5. Nope. It’s all drugs, dude. No BCI here. So, moving

    on… http://www.pasivdevice.org/
  6. Build a model that correlates features in the movies (shapes,

    edges, motion) to patterns of activity in the brain. 2
  7. Record brain activity while subject watches a new set of

    trailers. This will be used to assess the quality of the model and generate reconstructed video. 3
  8. Randomly sample around 5000 hours of video clips from YouTube

    and run these samples through the model to generate predictions of brain activity. 4
  9. Select the 100 clips which generated predicted brain activity most

    similar to the observed brain activity, then blend (i.e., average) these clips. This yields the reconstructed video. 5