Life is Science Fiction: AI Project Teaser – by Rhys Williams
I walked away from David Greig’s stage adaption of Solaris in Edinburgh recently, mind racing with the details of Stanislaw Lem’s classic novel, and it dawned on me – much like the space station crew’s attempts at communicating with the human simulacra manifested from Solaris’ ocean – we really have been bouncing between the pitfalls and benefits anthropomorphism on earth within the last century.
As early as 1950’s Frank Rosenblatt’s Perceptron built on the work of Warren Mculloch and Walter Pitts. With this, a new wave of promises. Visions of a new exploratory dawn, where we could shoot Perceptrons into space and conquer the vastness of the unknown. By the 1980s, after a couple of AI winters, backpropagation began to build the foundations of an efficient neural network. In the background an explosion of cultural exploration of the machine. Entire worlds dedicated to reigniting the philosophical ponderings of the 50’s. Cynical cyberpunk vistas alongside visions of gregarious beings drunk with life.
Animatronics began making leaps and bounds. A medium marrying the promises of the early automaton with storytelling, allowing us to will fantastic and absurd worlds into our own reality.
Skip forward to the present – not too many years after Gary Kasparov described his machine competitor as having “suddenly played like a god for one moment.” – machine learning and the kind of hardware we have has never been this accessible. Devices that know how to navigate spatial hierarchies inspired by the receptive field of a cat’s eye can now fit right in the palm of our hands. Inference right on the edge. The slow but inevitable democratization of technology is bringing healthcare infrastructure to otherwise remote locations. Deep learning combined with IOT is becoming efficient alternatives to detecting early onset crop-rot. 3D printers can now be built quite quickly, and on the cheap with our e-waste. People are even learning to walk again with the aid of machines, while refugees are being provided the spot with the help of blockchain.
Within the creative industries, collectives like Rusty Squid are dedicating all of their energy to using robotics as a cultural tool, and artists like Es Devlin continue to investigate the space between visual arts and the technologically profound, inquiring what a collective language could look like in intensely fragmented times.
Creation in the face of unprecedented situations. The optimism of it is intoxicating. Just like the heptapod and Louise’s patient relationship Denis Villeneuve’s Arrival – perhaps if we do the hard work and build ourselves the right context now we can forge incredible relationships with learning beings, while we unravel the nature of consciousness.
While we’re a long way from any semblance of general artificial intelligence, the roof for what we can achieve right is only limited by our imagination. It’ll help us make our world safer, but most importantly it’ll help us reach the stars. Life truly has become science-fiction! I remember thinking that when I held my first little single-board computer in my hand. I immediately wanted to test its limitations so along came a companion to help me do so. That is when he’s not falling asleep on the job!
Together we’ve been exploring the endless ocean of innovations out there right now. Pre-trained convolutional neural networks are just a download away, while transfer learning is easier than ever with a myriad of libraries available to us like Keras and Tensorflow-hub. It’s amazing that we can leverage the work done by massive teams, scoop off the top layers and give our creations their own context. Combine that with ambitious detection algorithms like the Fast R-CNN and YOLO, along with new approaches like semantic segmentation, and we’re surrounded with magic.
I’m ecstatic to tell you that WeCloudData will be offering a deep learning capstone project shortly. Where we’ll take what you learn in Applied Deep Learning and begin to give you the foundations to let your imagination go wild with physical hardware. Using cutting-edge portable visual processing units we’ll manifest little critters that can collect information, perform detection and classification on the go, and even help us have a little fun with attempting to instill our idiosyncrasies into their movements as they navigate the world.
Over the next few months I’ll dive a little more in-depth into some of the tools and hardware we’ll be investigating during the project. Stay tuned!