Deloitte’s Presentation at Infinite Vision

Back

Feb. 8, 2022

“To imagine the future of computer vision, we must first go back to the origin of natural vision.” Hear Varvn Aryacetas of Deloitte examine the possibilities.

Watch all the videos from Infinite Vision 2022

Varvn Aryacetas:

I’m Varvn Aryacetas and a futurist at Monitor Deloitte. I’ve the privilege of studying what impact nascent and exponential technologies may have on our future. Today, we’re here to talk about the future of computer vision and we’ll cover the origins of vision, what challenges it poses and possibilities it unlocks? To imagine the future of computer vision, we must first go back to the origins of natural vision. The eye is the first thing that comes to mind. When we think of vision, eyes have independently evolved multiple times in the animal kingdom with abundant diversity. There are eight types of eyes, and some can only see black and white. Some can see color.

Varvn Aryacetas:

Some can even see infrared ultraviolet or poor eyes light. The vision is much more than just the eye. The brain, which [inaudible 00:01:10] with eyes helps us make sense of the light we see. Thus, we have a simple formula site is the product of the eye and the brain. Now let’s break down and appreciate the sort things we can do with our eyes and brains. We can gaze, look, see, watch, stare, behold, observe, perceive, trap, focus, inspect, and study things around us. We see both still and moving images. We steam light, and darkness. We recognize shape, size, texture, sheen, transparency. We estimate temperature, count, depth, weight, speed, and velocity. We understand gestures, facial expressions, body language, and sign language.

Varvn Aryacetas:

Now that we appreciate the origins and complexity of natural vision, let’s see how far we have come with artificial vision. How well do we replicate this? So let’s take the first part of our simple site formula again the eye. The camera is the eye of artificial vision. We’ve made huge leaps on this front over the last five centuries. We now have sophisticated lenses, filters and sensors. These can sense a broad light spectrum, including x-rays and UV rays. And onto the second part of the brain, the processors, and the algorithms form the brain in artificial vision and we made impressive progress over the last few decades. We have computer vision systems that can lock onto moving objects and track them that can recognize many attributes of objects such as size, shape, color that can estimate counts of boxes in a warehouse using Apple iPhone, or count the number of fans at a stadium.

Varvn Aryacetas:

Systems that can try to identify the gender or ethnicity of a person, these systems can even recognize faces, expressions, and body language and if the gate of a person is threatening. However, the challenge has been to do these upscale economically while maintaining consistency and accuracy. Beyond these engineering challenges, we have major ethical privacy and security challenges around these systems and ecosystems. So navigating these will only get trickier as we explore a plethora of future scenarios. So let’s dive in, lenses and filter, let’s take that part now. So we may see innovation in this world of lenses and their codings and the filters associated with them. They’ll be lighter, more flexible, pliable, durable, Google for instances already file patents for a contact lens camera, sensors. So with quantum or nano technologies, it might be possible to packing much more power into our chips, but beyond just creating these chips, I think there is a huge opportunity to apply biomimetics, so that we can design new solutions and capture broader light spectrum, and even things like magnetic field lines.

Varvn Aryacetas:

The next component here are processors and algorithms. This is where we can expect to see huge leaps and rewards along with risks. With the advances in our machine learning algorithms, we may be able to have computer vision systems that can read lips and fluently understand sign language. And with the metaverse on the horizon, it’ll be essential to crack the ability to identify shape, size, weight, contours, and textures of objects. To classify and sort objects and people along different variables, augmented reality and mixed reality will provide the incentive and impetus for solving these problems at scale and speed because without cracking this nut, augmented reality, mixed reality are just not going to be viable. Translating 2D objects and humans into 3D models in the metaverse is going to be another important thing to crack Facebook AI, open AI, just two names have published numerous research papers just over the last three years, demonstrating how they’re able to do this.

Varvn Aryacetas:

And beyond smarter machines with advanced artificial vision, we can see a future where natural vision is augmented by computer vision. Though, we have long augmented our eyes with spectacles, binoculars, and telescopes. We may be able to do far more by combining brain machine interfaces, like the ones from Neuralink and external sensors or contact lenses like the ones from Google ad mentioned. And this could unlock so much more possibilities along with risks. Before we conclude, let’s reflect on a quick analogy flight, natural flight comes in magnificent diverse forms, birds, bats, dragonflies, butterflies, beetles, and gliders to name a few. An artificial flight looks surprisingly different, fixed wing aircraft, rotary wing aircraft, and even jet packs look quite different from what we see in nature. I bring this analogy up because we must expect to see human ingenuity realize artificial vision in a whole different way. And we must also expect to see new risks, hazards, standards, laws, regulations, crafts, jobs, ecosystems, and startups pop up just as they did in the world of flight. So buckle your seatbelts and get ready for your flight into this future.