Authors: James Essinger
Jacquards' Web (36 page)
Furthermore, even when a scientific or technical breakthrough has facilitated an important initiative in the computer industry, experience shows that the subsequent path of that initiative is often unpredictable too. This is another lesson the story of Jacquard’s Web teaches us: that the progress of a new technology, far from following a clear, logical track, is generally a haphazard and even messy affair. Its success or failure itself will depend on a complicated web of dynamic, complex factors such 258
The future
as practical necessity, financial pressures, political considerations, and the personal needs and prejudices of the would-be inventor and of everybody else whose decisions affect the invention’s progress.
No account of the history of the computer, and its likely future, would be complete without a discussion of the concept of ‘artificial intelligence’ (AI). The notion of AI takes the evolution of the Jacquard loom to its logical conclusion. If it is possible to build a special kind of Jacquard loom that can weave information, why not build a special kind of information processing loom that can think entirely for itself? And if this is possible, could not an even more brilliant type of loom be built that could think for itself about
anything
?
Since the
1950
s, there has been considerable discussion of the possibilities for the practical realization of AI. The term appears to have been originated by the British computer pioneer Alan Turing. In its most ambitious form, AI means machine intelligence that could, at least in theory, seek to imitate the thought-processes of the human brain.
Naturally enough, the notion of an intelligent computer—
and in particular a computer that suddenly becomes self-aware and intelligent—offers many exciting dramatic possibilities, especially if the intelligence turns out to be of the malevolent kind. The idea has inspired countless novels and movies, most notably Arthur C. Clarke’s novel
2001: A Space Odyssey
and Stanley Kubrick’s
1969
movie of the same title.
One particular sequence in the movie features a spacecraft accommodating a revolutionary computer that exhibits self-awareness and a level of intelligence not only comparable with a human being’s, but in excess of it. During the mission, the computer makes an unaccountable mistake that triggers a spiral of further errors and disobedience. The captain ‘tells’ the computer he will disconnect it if it does not obey it. In response, the 259
Jacquard’s Web
computer, dreading disconnection above everything else, sets out to kill the crew and comes close to doing so. The story is effective because the computer’s motivation seems entirely plausible. Certainly, if a computer which could think for itself
were
ever built, it would be more than reasonable for us to suppose that one of its first concerns would be to take any action necessary to prevent itself ever from being switched off.
After
2001
was released, it was pointed out that the computer’s name, HAL, was simply the acronym IBM, with each letter being substituted for the one that precedes it in the alpha-bet. Had Arthur C. Clarke deliberately intended this to be the case? To try to clear the matter up, I contacted Clarke while researching this book. He e-mailed me by return to emphasize that the abbreviation HAL was simply taken from ‘Heuristic Algorithmic’: the technical term for the basis of HAL’s programming. The ‘connection’ with IBM is thus apparently a coincidence, albeit an intriguing and convenient one: perhaps even possibly an unconscious one on the part of the author. This ghostly link between the name of the world’s first—if fictional—
intelligent computer and the global corporation which, via the Hollerith link, can reasonably be said to owe its origins to the idea behind Jacquard’s loom, resonates in the history of computing like the fulfilment of a prophecy.
Fictional representations of artificial intelligence are so effective and thought-provoking that the fact of the minimal real-life progress made in the field can seem, by comparison, acutely disappointing. The simple fact is that today, no computer on Earth offers a level of performance that
remotely
resembles the reasoning and creative power of the human mind. Even the problem of mimicking the human brain’s remarkable skill at visual recognition—a task that initially seemed to offer the prospect of being achievable—has proved much more difficult than was originally anticipated.
Why can we imagine intelligent machines but not make them? Part of the answer appears to be that our technological 260
The future
imaginations always seem to run well ahead of our technological capabilities. This seems to be a consequence of the nature of our tool-maker’s brain. Very likely if it wasn’t the case, we would not have achieved a fraction of what we have achieved as toolmaking creatures.
At a technical level, a fundamental and overwhelming problem with developing true AI is that it is not known precisely how the human brain ‘digitizes’ information in order to make it available as the subject of thought carried out by brain cells.
There is a certain limited knowledge of the coding used by nerve fibres and even how some of the signal processing (for tasks such as visual processing) is performed, but otherwise our information about how our own brains actually work is woefully incomplete.
Until we have a much more comprehensive knowledge of this, it is difficult to imagine how a computer could be taught to understand even basic facts about our world; facts that the machine would need to know if it was to display any useful intelligence.
There has, however, been progress with AI in areas where the field in which the ‘intelligence’ is to be deployed is of its essence restricted and specialized and consequently does not require the computer to have any knowledge of the natural world. In particular, programs designed to play the game of chess have been notably successful.
A significant milestone was reached in
1996
. During a match played in Philadelphia that year between the then World Chess Champion Gary Kasparov—widely regarded as the best chess player of all time—and an IBM-sponsored chess-playing computer program named Deep Blue, Kasparov suffered a loss to the computer. This was the first loss ever by a reigning World Chess Champion to a chess program at normal tournament time limits.
Subsequently, in May
1997
, an updated Deep Blue with a capacity for considering
200
million positions per second—twice its normal speed—beat Kasparov in a six-game match, with the computer scoring two wins, three draws, and one loss.
261
Jacquard’s Web
Strictly speaking chess-playing programs do not consciously
‘think’ in any meaningful sense. What they are able to do is to assess tens of millions, or hundreds of millions, of possible chess positions within a fraction of a second. Nonetheless, the program does display a form of artificial intelligence, because with approximately twenty possible moves available, on average, to each player for each move, the number of potential positions in a game of chess after only ten moves by each side is already astronomical. It is totally impossible for any computer to be programmed with details of all the positions that could ever arise on a chessboard. Instead, the programmers have no choice but to equip the computer with a set of rules that it can use in any given position. In other words, the computer is obliged to make its own decisions having been programmed with initial instructions.
This is not unlike how the human mind behaves, at least when it is solving problems.
All the same, we are entitled to ask ourselves just how intelligent computer chess programs really are if we consider that such a program, operating inside a burning building, would simply go on playing chess rather than trying to escape. But strictly speaking that would be our fault for not giving the computer any way of knowing that the building was on fire.
In any event, chess-playing computer programs nowadays routinely compete against human chess grandmasters (and against other chess-playing programs) in international tournaments.
Many human expert players have developed successful strategies for dealing with chess-playing programs by keeping the position blocked and avoiding the kind of tactical complications at which computers, which don’t usually make mistakes in tactical calculations, excel. Whether this kind of mental struggle for supremacy between man and machine intelligence in the limited arena of the chessboard will one day be enacted in the greater world is surely one of the most terrifying questions we face as we move ahead into the twenty-first century and beyond.
262
The future
Meanwhile, we have little choice but to face the fact that it is extremely risky, if not downright foolhardy, to set down specific predictions for how computers—those modern incarnations of Jacquard’s loom—are likely to develop in the future. But we can be certain that the same qualities of ingenuity, passion to make a dream come true, and sheer desire to solve a pressing practical problem that have played such an important part in driving forward the story told in
Jacquard’s Web
will fuel developments in the future, developments that are sure to surprise and delight us.
The Jacquard loom has never stopped weaving since that first day two hundred years ago. How can we imagine, as we move on into the future, that it ever will?
THE END
263
This page intentionally left blank
Charles Babbage’s belief that the Difference Engine could indeed be built and made to work properly was given a triumphant vindication in
1991
. In that year a team of engineers at the Science Museum in South Kensington, London completed an astonishing six-year project to build Babbage’s second Difference Engine design: a modified and improved version of the Difference Engine prototype known by Babbage as Difference Engine No. 2.
The machine built in
1991
has worked perfectly from the outset.
The Science Museum once offered regular opportunities to view the completed Difference Engine in action. However, at present these have been suspended to allow the Museum’s engineers to concentrate their efforts on building the replica of the Difference Engine No.
2
mentioned below. Yet ultimately, the only way of appreciating the true brilliance and the beauty of Babbage’s concept of a Difference Engine—and especially the carriage mechanism—is to view the completed Difference Engine No.
2
in action at the Museum.
The one place in the world where you can see a completed Babbage Engine in operation is, at the time of writing, the Science Museum in London, by special arrangement. Only a trained curator can operate the Difference Engine and make the machine perform calculations. Unlike many of the exhibits in the Museum, the Difference Engine cannot be activated by members of the public directly. Turning the Engine’s handle requires expert training in order to make the machine operate at the right speed and with the correctly-paced pauses between each turn.
265
266
Appendix 1
The modern Difference Engine No. 2 completed in 1991 by a team at the London Science Museum to Babbage’s plans. It works perfectly.
I have seen the Difference Engine operating on two occasions. Both times the operator has been the Science Museum’s then assistant director and head of collections, Doron Swade, the inspired and dedicated computer scientist and historian who led the modern building of Difference Engine No.
2
and who has done so much to make an entirely new generation aware of Babbage’s foresight and genius.
When the Difference Engine is operating, the levers of the carriage mechanism—regularly spaced in successive segments of the Figure Wheels one above each other in the columns—create an entirely unexpected phenomenon: a beautiful oscillation that has the appearance of endlessly changing and rippling sine waves flowing around at the back of the machine.
(
left
) Babbage’s plans for Difference Engine No. 2.