High Tech Binoculars

In Johnny Mnemonic we see two different types of binoculars with augmented reality overlays and other enhancements: Yakuz-oculars, and LoTek-oculars.

Yakuz-oculars

The Yakuza are the last to be seen but also the simpler of the two. They look just like a pair of current day binoculars, but this is the view when the leader surveys the LoTek bridge.

jm-25-yakuza-binocs-adjusted

I assume that the characters here are Japanese? Anyone?

In the centre is a fixed-size green reticule. At the bottom right is what looks like the magnification factor. At the top left and bottom left are numbers, using Western digits, that change as the binoculars move. Without knowing what the labels are I can only guess that they could be azimuth and elevation angles, or distance and height to the centre of the reticule. (The latter implies some sort of rangefinder.)

So far, this is a simple uncluttered display. But why is there a brightly glowing Pharmakom logo at the top right? It blocks part of the view, and probably doesn’t help anyone trying to keep their eyes adapted for night vision.

LoTek-oculars

The LoTeks, despite their name, have more impressive binoculars. They’re first used when Johnny gets out of his airport taxi.

jm-11-lotek-binocs-a-adjusted

There’s a third tube above the optics, a rectangular inlet, and an antenna.

In these binoculars, the augmented reality overlay is much more dynamic. Instead of a fixed circle, green lines converge in a bounding box around the image of Johnny. Text slides onto the display from left to right, the last line turning yellow.

jm-11-loteks-animated

Zoomrect

The animated transition of the bounding box resembles what Classic MacOS programmers of the 1990s called “zoomrects” used for showing windows opening or closing. It’s a very effective technique to draw attention to a particular area of an image.

Animated text

Text appearing character by character is ubiquitous in film interfaces. In the 1960s and 1970s mainframe and minicomputer terminals really did display incrementally, as the characters arrived one by one over slow serial port links. On any more recent computer it actually takes extra programming to achieve this effect, as the normal display of text is so fast that we would perceive it as instantaneous. But people like to see incremental text, or have been conditioned by film to expect it, so why not?

Bioscanning

The binoculars detect Johnny’s implant. It might just be possible to detect this passively from infrared or electronic signals, but more likely the binoculars include a high resolution microwave radar as well. If there had been more than one person in view, the bounding box would indicate which one the text refers to. And note that the last line of text is a different color. What that means is unclear here, but it becomes clear (and I’ll discuss it) later.

The second time we see the LoTek binoculars is when a lookout spots Street Preacher, a very bad guy and another who wants to remove Johnny’s head. Once again the binoculars have performed more than just a visual scan.

jm-17-lotek-binocs-a-adjusted

The binocular view and overlay are being relayed to another character, the LoTek leader J-Bone who can watch on a monitor. Here the film anticipates the WiFi webcam.

jm-17-lotek-binocs-b-adjusted

The overlay text now changes.

jm-17-lotek-binocs-c-adjusted

Narrow AI?

This is interesting, because the binoculars can not only detect implants and other cyborg modifications, but are apparently able to evaluate and offer advice. It appears that the green text is used for the factual (more or less) information about what has been detected, while yellow text is uncertain or or speculative.

Does this imply a general artificial intelligence? Not necessarily. This warning could be based solely on the detected signature, in the same way that current day military passive sonars and radar warning receivers can identify threats based on identifying characteristics of a received signal. In the world of Johnny Mnemonic it would make sense to assume that anyone with full custom biomechanics is extremely dangerous. Or, since Street Preacher is a resident rather than a stranger and already feared by others, his appearance and the warning could have been entered into a LoTek facial recognition database that the binocular system uses as a reference.

These textual overlays are an excellent interface, not interfering with normal vision and providing a fast and easy-to-understand analysis. But, the user must have faith that the computer analysis is accurate. There’s no reason given as to why any of the text is displayed. If Johnny was carrying an implant in his pocket instead of his brain, would the computer know the difference?

An alternative approach would be some kind of sensor fusion or false spectrum display, with the raw infrared or radar image overlaid over the visuals and the viewer responsible for interpreting the data. The problem with such systems is that our visual system didn’t evolve to interpret such imagery, so a lot of training and practice is required to be both fast and accurate. And the overlay itself interferes with our normal visual recognition and processing. If the computer can do a better job of deciphering the meaning of non-visual data, it should do so and summarise for the human viewer.

Further advantages of this interface are that even a novice sentry will benefit from the built-in scanning and threat analysis, and the wireless transmission ensures that the information is shared rather than being limited to the person on watch. 

Perpvision

GitS-heatvision-01

The core of interaction design is the see-think-do loop that describes the outputs, human cognition, and inputs of an interactive system. A film or TV show spends time showing inputs without describing some output, only when these users are in the background and unrelated to the plot. But there are a few examples of outputs with no apparent inputs. These are hard to evaluate in a standard way because it’s such a giant piece of the puzzle. Is it a brain input? Is the technology agentive? Is it some hidden input like Myo‘s muscle sensing? Not knowing the input, a regular review is kind of pointless. All I can do is list its effects and perhaps evaluate the outputs in terms of the apparent goals. Ghost in the Shell has several of these types of inputless systems. Today’s is Kusanagi’s heat vision.

Early in the film, Kusanagi sits atop a skyscraper, jacked in, wearing dark goggles, and eavesdropping on a conversation taking place in a building far below. As she looks down, she sees through the walls of the building in a scanline screen-green view that shows the people as bright green and furniture as a dim green, with everything else being black.

She adjusts the view by steps to zoom closer and closer until her field of vision is filled with the two men conversing in her earpiece. When she hears mention of Project 2501 she thinks the message, “Major, Section 6 is ready to move in.” She reaches up to her right temple and clicks a button, to turn the goggles off before removing them.

That’s nifty. But how did she set the depth of field and the extents (the frustum) of the display so that she only sees these people, and not everyone in the building below this? How does she tell the algorithm that she wants to see furniture and not floor? (Is it thermography? Is the furniture all slightly warm?) What is she doing to increase the zoom? If it’s jacked into her head, why must she activate it several times rather than just focusing on the object with her eyes, or specifying “that person there?” How did she set the audio? Why does the audio not change with each successive zoom? If they’re from separate systems, how did she combine them?

Squint gestures

If I had to speculate what the mechanism should be, I would try to use the natural mechanisms of the eye itself. Let Kusanagi use a slight squint gesture to zoom in, and a slight widening of the eyelids to zoom out. This would let her maintain her gaze, maintain her silence, keep her body still, and keep her hands free.

The scene implies that her tools provide a set amount of zoom for each activation, but for very long distances that seems like it would be a pain. I would have the zoom automatically set itself to make the object on which she is focusing fill her field of vision less some border, and then use squint-gestures to change the zoom to the next logical thing. For instance, if she focused on a person, that person would fill her field of vision. A single widening might zoom out to show the couch on which they are sitting. Another the room. This algorithm wouldn’t be perfect, so you’d need some mechanism for arbitrary zoom. I’d say a squint or wide-eyed gesture held for a third of a second or so would trigger arbitrary zoom for as long as the gesture was maintained, with the zoom increasing logarithmically.

As for the frustum, use the same smart algorithm to watch her gaze, and set the extents to include the whole of the subject and the context in which it sits.