IN the latest thriller to hit British screens, Keanu Reeves plays a man
with a digital memory built into his brain. Using a neuroconnection in the
back of his head, the eponymous hero, Johnny Mnemonic can plug into any
computer and store or download data. With this memory upgrade, his brain acts
like the hard disc on a computer and Johnny can use this ability to carry
secret digital information around the world.
And while you settle down to enjoy Hollywood鈥檚 take on the future, all over
the world researchers really are working out how to connect the brain directly
to digital devices such as computers, databases and video cameras. Their
mission is to develop thought-controlled computers and memory upgrades that
will make today鈥檚 keyboards, monitors and other interfaces obsolete. It may
even become possible to communicate over digital networks by thought
alone.
But while all this is far futures stuff aimed at 鈥渆nhancing鈥� people and
society as a whole sometime next century, the quest is proving positive in the
shorter term. So much so that if the researchers can perfect their ideas,
quite soon they may be able to help people with disabilities use their brain
waves to control wordprocessors and wheelchairs, or train pilots to fly
fighter aircraft using simulators controlled by their minds.
Advertisement
Progress, however, is limited by what little we know about the way the
brain works. There is one big problem facing researchers who want to connect
the brain directly to digital devices 鈥� or, more prosaically, design some sort
of brain-machine interface. Unlike a computer, the brain does not rely on a
steady stream of digital bits and bytes. Connecting them together is not
easy.
One possibility is to measure the night-and-day activity inside our brains.
This unceasing communication of billions of neurons produces bursts of
electricity that can be measured outside the skull using an electro-
encephalograph or EEG. Interpreting those signals is extremely difficult
because of the complex connections between neurons and the sheer volume of
signals that the brain handles. But it is these signals that some researchers
hope to use to control computers.
Neuroscientists have long known that the patterns of activity produced by
the brain depend on its state. For example, when we relax, our brains produce
signals with a frequency of between 8 and 13 hertz, known as alpha waves. But
when we get excited, we produce beta waves of between 15 and 30 hertz, while
other frequencies are associated with sleep or intellectual activity.
In theory, a computer monitoring these different EEG signals could be
programmed to turn a light on or move a cursor to the left if it spotted a
change from alpha to beta waves, for example. Controlling the computer would
be a relatively simple task of training your mind to relax or become excited 鈥�
a task already proven in experiments. Unfortunately, such a simple feedback
system is too limited to provide a useful interface between humans and
computers. So researchers have begun to look more closely at EEG signals to
see if whether there are more complex patterns that computers could respond
to.
In Taiwan, for example, Shiao-Lin Lin and colleagues in the Department of
Neurology at the National Taiwan University Hospital in Taipei, have begun to
uncover a deeper structure in the patterns. Shiao-Lin says that EEG signals
appear to be extraordinarily complex. In 1993, the team discovered that the
brain produces spikes of activity a fraction of a second after a mental event
such as looking at a series of characters on a display. Similar spikes occur
before an action takes place, when the brain is preparing to move the fingers
holding a computer joystick, for example.
It may even be possible to link certain patterns of brain waves with
specific mental tasks such as rotating an imaginary three-dimensional object.
Shiao-Lin believes that if these signals can be fully analysed, a computer
could be trained to recognise and act on characteristic patterns that
correspond to specific thoughts.
Last year, at the University of Tottori, near Osaka in Japan, a team of
computer scientists led by Michio lnoue took this idea further by analysing
the EEG signals that correspond to a subject concentrating on a specific word.
This research is designed to allow sufferers of degenerative nervous diseases
who cannot control their bodies to communicate using thought.
The system depends on a database of EEG patterns taken from a subject
concentrating on known words. To work out what the subject is thinking, the
computer attempts to match their EEG signals with the patterns in the
database. For the moment, the computer has a vocabulary of only five words and
takes 25 seconds to make its guess. In tests, lnoue claims a success rate of
80 per cent, but he is working on improvements that will make the system even
more accurate and hopes to have a version on the market within a year or
two.
EEG signals could be used in even more complex situations. Earlier this
year at Imperial College, London, Stephen Roberts, a lecturer in neural
computing, started to look at ways in which EEG signals can help people with
disabilities control their wheelchairs. Roberts is concentrating on brain
activity generated by the intention to move a limb 鈥� signals which are still
produced even when movement does not take place.
But computer-based analyses of EEG signals are never 100 per cent accurate.
鈥淔or control of a wheelchair, even 90 per cent accuracy is not good enough,鈥�
says Roberts. To double-check someone鈥檚 intention to move in a particular
direction, his system also monitors eye movements. If the line of sight
doesn鈥檛 match the computer鈥檚 EEG prediction, something is wrong. 鈥淚t is better
to stop the wheelchair for one second rather than make a false move,鈥� he
says.
Elsewhere, at labs such as the Alternative Control Technology Laboratory at
the Wright-Patterson Air Force base in Dayton, Ohio, researchers are
investigating ways of controlling flight simulators using EEG signals.
Despite some successes, work like this has highlighted the obstacles that
stand in the way of thought-controlled computers. Most people do not usually
concentrate on the way their brain works and at the Wright-Patterson base,
many subjects had difficulty consciously controlling the type of EEG signals
that the brain produces 鈥� a very different action to normal thought processes.
In Japan, the sheer complexity of EEG signals is still proving a major
headache to analyse 鈥� even using sophisticated computer-based systems such as
neural networks.
Both problems limit the range of useful tasks, suggesting that EEG signals
may not be an effective way to perform the complex tasks that are routine
using other existing interfaces such as a keyboard and mouse. Would it be
possible to use an EEG interface to compose a letter? The ideal solution
would be for the letter 鈥渆鈥� to appear on a computer display when we think it.
But detecting the appropriate EEG pattern or training people to produce other
more recognisable patterns that would correspond to 鈥渆鈥� (never mind the rest
of the alphabet and the many other tasks that would be necessary) appears
almost impossible. Is there a better way of achieving 鈥渕ind over
肠辞尘辫耻迟别谤鈥�?
Which brings us back to the far futures and the researchers who believe
that a far better way to connect the mechanisms of our brain to the bits
flowing around a computer is to connect them directly. Not surprisingly.
research into 鈥渘eurocompatible interfaces鈥� is in its infancy although the
basic techniques and technology are developing rapidly.
At the department of Bioengineering at the University of Utah, Richard
Normann鈥檚 team has come closest to 鈥渏acking鈥� into the brain. They have been
developing ways of supplying video images directly to the brains of people who
have lost their sight. The problems of blindness provide an ideal test
application for neuroprosthetics because most forms of blindness are due to
defects or damage to the eyes. This leaves the complex neural machinery of
vision in the brain known as the visual cortex still working.
The visual system is especially useful because it is highly adaptive,
intelligent, and self-regulating. In 1974, William Dobelle, also at The
University of Utah, found that direct stimulation of the visual cortex in
blind people evokes 鈥減hosphenes鈥� 鈥� points of light which are similar to those
created by signals passing directly from a visual system which is working
properly. The results were encouraging: people were able to 鈥渞ead鈥� Braille
characters made up of phosphene patterns created by direct stimulation faster
than they could read them with their fingertips.
More recent research in 1992 and 1993 at the National Institutes of Health
and Johns Hopkins University in the US has centred on improving the quality of
the artificial vision experienced by blind subjects (鈥淪ight for sore eyes鈥�,
New 杏吧原创, 19 August 1995). All of this suggests that blind subjects may
be able to adapt well to new forms of prosthetic visual systems.
Direct data
Normann鈥檚 group has been developing devices that can be implanted directly
into the brain. They consist of an array of 100 鈥渘eedle鈥� electrodes resembling
a tiny hairbrush. Each needle is less than 2 millimetres long, isolated from
its neighbour by a glass sheath and mounted on a silicon base about 4
millimetres square. The idea is to capture images using a video encoder,
transform them into electrical signals and excite neurons in the visual cortex
of the brain using the electrode array to produce an image directly in the
mind.
At the moment, results seem to show that the approach is capable of
creating artificial vision, even though it may appear to the subject like a
grainy version of reality 鈥� similar to looking at the large scoreboards at
football stadiums. Of course, Normann鈥檚 work is confined to people who have no
other hope of seeing again and he finds it difficult to imagine healthy,
novelty-seekers risking such surgery.
Other research is approaching the great link-up from a different angle 鈥�
using the body鈥檚 existing information channels to the brain such as the eyes
and ears. Thad Starner, a researcher at the Media Lab at the Massachusetts
Institute of Technology, believes that the surgically invasive technology
needed to jack in to the brain has yet to be demonstrated. Instead, he is
working on 鈥渨earable computers鈥�, tiny microprocessors worn on the body that
are in continuous communication and even connect to the Internet.
The basic building blocks for wearable computers are already around.
Lowpower, credit-card sized computers with the power of 486 PCs are already on
the market. The US Army has developed wearable computing and communications
devices such as head-mounted displays, cameras and personal communicators to
receive and transmit information on the battlefield. And Starner says the army
has already tested 鈥渁ugmented soldiers鈥� in the field.
According to Starner, the killer application will be augmented memory,
rather like Johnny Mnemonic鈥檚 except that the hard disc will be outside the
brain. The idea is that a total recall system will selectively record a user鈥檚
life using face recognition, voice recognition, and some sort of global
positioning system to track location (鈥淒on鈥檛 forget your memory aide鈥�, New
杏吧原创, 5 February 1994). Starner says that when greeting colleagues, your
remembrance agent鈥� will recognise them and suggest the top five pieces of
information relevant to the conversation.
Wearable computers provide several advantages. One is that the computer is
always switched on and always immediately accessible. By contrast, current
palmtop computers must be opened, turned on, attention must be focused on the
screen, and then both hands must be used to operate it. 鈥淐ompare this to a
wearable computer with a head-up display and a one-handed keyboard,鈥� Starner
suggests. 鈥淲ith my wearable I can store names and interesting snippets of
conversations while shaking hands and maintaining eye contact during
professional meetings.鈥�
The second advantage of wearable computers is consistency. Most
interactions with computers using keyboards, pens, mice or whatever require
training each time a new system is developed. Wearable computing promises a
single consistent interface. 鈥淪ince so much time is spent with the wearable
interface, users tend to get very proficient with it and customise it to their
needs. Indeed, it is worth their effort to customise the interfaces since
these devices are designed for long-term, intimate relationships,鈥� says
Starner.
Starner鈥檚 approach 鈥� enhancing human abilities with computing technology 鈥�
is in marked contrast to Normann鈥檚 work 鈥� largely centred round replacing lost
or missing functions. Somewhere in the middle is Chip Maguire. Based at the
Royal Institute of Technology in Sweden, Maguire believes that wearable
computers will be limited in the same way as EEG-controlled computers and the
only way to get a usable human-computer interface is by using direct
connections with the brain. Maguire suggests that these would allow us to have
a system installed inside our heads which would provide voice communications
and an 鈥渆yes-up鈥� display which would superimpose text and pictures on our
normal vision.
The first group of people to use the devices will be the disabled, says
Maguire, but eventually he believes that other people will also
willingly undergo the surgery that neurocompatible computers require.
Robosoldier.
One of the first groups of able bodied volunteers will be the professional
military, he thinks. After them, might come those involved in information-
intensive business such as foreign exchange dealers. Maguire expects that the
first prototypes will be around in about five years and that military systems
will appear within ten years. But other users may have to wait for two or
three decades before the technology becomes acceptable.
If Maguire is right, then life in the 21st-century could become more
complex than any Hollywood thriller. Aside from any moral and ethical
objections, the concept of beings that are part-human, part-machine raises
many practical issues.
For example, if we have software embedded in our brains, how do we ensure
its quality and reliability? What happens when there is a new hardware upgrade
or a new software release? What if somebody discovers a software bug or a
design error? Even a Hollywood script writer would be hard-pressed to picture
the consequences.