Hi, I'm Tim Tyler, and this is a video about augmenting human intelligence.
Introduction
Intelligence augmentation is the practice of making humans smarter
with technology.
While the field also covers smart drugs, nutrition and education, this
video will be concerned with the practice of making humans smarter
using machines.
So far, most intelligence augmentation has consisted of interfacing
brains to additional computing hardware, which has properties that
complement the human brain.
Computers sit on many desktops, and they are carried around - in the
form of laptops, PDAs and mobile phones. They are connected together
in an enormous, worldwide network.
These computer excel at rapid, serial deterministic operations - and
those are areas where the human brain is weak. They also effectively
supplement human memory.
In contrast with "pure" machine intelligence strategies, intelligence
augmentation builds on the human brain - rather than starting again
from scratch. That means that systems start out with human level
intelligence, and go forwards from there. On the other hand, these
systems have to interface with a human - and that is often a
clumsy and slow bottleneck in the resulting system.
Despite the problems, intelligence augmentation has had an
enormous impact on the world so far - arguably much greater
than machine intelligence strategies that omit humans.
This video will look briefly at the current state, and then consider
the future prospects of the field.
Augmented intelligence
Some of the original cybernetic pioneers who originally conceived
of the idea of intelligence augmentation envisaged it as an alternative
to the pure machine intelligence projects of the time.
However, from the point of view of the machines, intelligence
augmentation looks more like a way to make use of existing human
brains as stepping stones, rather than competing directly with them.
I'll illustrate the issue with some diagrams.
Here's a cybernetic diagram of the nervous system of an unmodifed human:
You can see that the sensory system is in green, and the motor system
is in blue.
Next, here's the corresponding cybernetic diagram of a pure-machine
intelligence:
Considering a human brain augmented with some machine sensors produces a
diagram that looks like this:
The machine sensors preprocess part of the human's sensory data, and
post-process the human's motor output.
The machine sensors do not need to passively process the data. They
may themselves be intelligent. Just as the human retina highlights
motion and edges, so machine sensors can highlight correct answers
and potential risks.
Similarly with the motor outputs - they too can benefit from being
intelligent - so they can identify incorrect actions and correct them.
Now consider this cybernetic diagram of an augmented human intelligence.
As time passes, we can expect the machine intelligence component to
grow and grow - gradually replacing the functions of the human brain
one domain at a time - until the original brain is redundant and can
be discarded.
For the most part, intelligence augmentation looks like a route to
pure-machine intelligence - rather than an alternative to it.
However, the relationship between pure-machine intelligence projects
and ones which seek to augment human intelligence is not
entirely benign. For one thing, the former has a greater need for
parallelism. The sorry state of commercial computing hardware in this
respect is - in part - a reflection of the fact that augmentation
projects already have the resource of a huge parallel
machine to draw on.
Sensors and actuators
One of the problems with augmented intelligence projects is that they
require a specialised range of sensors and actuators devoted to the
task of interfacing with humans.
Computer input devices are the biggest problem. Humans have terrible
output bandwidth. Some of the most important input devices so far have
been the computer keyboard and mouse. Cameras, scanners and
microphones are becoming increasingly important.
Computer output devices generally face fewer problems. Computer
monitors, computer printers and loud speakers have been some of the
most significant output devices so far.
Future prospects
Looking to the future, input devices can be improved by getting
speech recognition into a more usable state, and by training a camera
on the user to track their head and eye movements - and to respond to
gestures. There's also quite a lot which can be done with additional
buttons, toggles and wheels - as my own set-up demonstrates.
Output devices can also be improved. In the short term, monitors will
grow, and people will increasingly tile monitors, to produce monitor
walls. On a longer timeline, the advent of 3D printing,
programmable matter, and robots will result in more output channels.
Cost and portability are other areas with considerable scope for
improvement. As costs come down, and portability improves, computers
will become increasingly ubiquitous in wealthy countries - and will
increase their penetration in less affluent ones.
Head-mounted systems will analyse the environment, whisper advice into
our ears, and use lasers to project meta-information onto the external
world.
Rewards
Finally, we come to the issue of finance and rewards.
Pure-machine intelligence projects typically start off facing a
funding gap - since considerable work needs to be done initially
before the resulting products can be competitive with human labour.
There is no such gap for intelligence augmentation projects. They
start out being competitive with humans - since a human is the base
component - and every step forwards from there brings immediate
rewards.
As a result, intelligence augmentation projects often enjoy better
funding, since they bring quicker returns. This in turn raises the bar
which pure-machine intelligence projects must rise above in order to
be viable.
So, in summary, intelligence augmentation is an extremely
important area. Those wishing to contribute to the progress of science,
technology, or civilisation in general should consider this field
closely - efforts in this area seems to have a history of paying off on
a large scale.