(These are excerpts from my book "Intelligence is not Artificial")
In Defense of Progress: Augmented Intelligence
Enough bashing computers. The computer might be the only major appliance invented since television, but it is qualitatively different than all the previous appliances. What can the dishwasher do other than wash dishes? The computer, instead, can do a lot of things, from delivering mail to displaying pictures. A computer is many machines in one. That was, in fact, the whole point of the Universal Turing Machine: a universal problem solver. Little did he know that its applications would range from phone conversations to social media.
The secret is the software:
In fact, there has been little progress in the physical world but a lot in the virtual world created by computers. Just witness the explosion of online services in the 1990s and of smartphone applications since 2007.
Perhaps even more importantly, the law of entropy does not apply to software: everything in this universe is bound to decay and die because of the second law of Thermodynamics (that entropy can never decrease). That does not apply to software. Software will never decay. Software can create worlds in which the second law of Thermodynamics does not apply: software never ages, never decays, never dies. (Unfortunately, software needs hardware to run, and that hardware does decay).
The catch is that software does not have a body and therefore cannot do anything unless it is attached to a machine. Software cannot cook and cannot start a car unless we drop it inside a computer and attach the computer to the appropriate machine. Software cannot even give answers without a printer, a screen or a speaker.
Disembodied software is like disembodied thought: it is an abstraction that doesn't actually exist.
Software has to be incorporated into a processor in order to truly exist (to "run"). In turn the processor, that ultimately only does binary algebra, has to be attached to another machine in order to perform an action, whether cooking an omelette or starting a car.
De facto, we attach a universal problem solver to a specific problem solver. However, there is a way to maximize the usefulness of a universal problem solver: attach it to another universal problem solver, the human mind.o
One could argue that, so far, Artificial Intelligence has failed to deliver, but "Augmented Intelligence" has been successful beyond the hopes of its founding fathers. In the 1960s in Silicon Valley there were two schools of thought. One, usually associated with John McCarthy's Stanford Artificial Intelligence Lab (SAIL), claimed that machines would soon replace humans. The other one, mainly associated with Doug Engelbart at the nearby Stanford Research Institute (now SRI Intl), argued that machines would "augment" human intelligence rather than replace it. Engelbart's school went on to invent the graphic user interface, the personal computer, the Internet, and virtual personal assistants like Siri; all things that "augmented" human intelligence. This program did not necessarily increase human intelligence and it did not create a non-human intelligence: the combination of human intelligence plus these devices can achieve "more" than human intelligence can alone.
The search engine is a good example of "amazing" augmented intelligence and "disappointing" artificial intelligence. It must be terribly difficult for search engines to keep up with the exponential growth of user-provided content. The ranking algorithm has to become exponentially smarter in order for the search engine to keep providing relevant answers. It's something that the user doesn't see (unlike, say, a new button on the microwave oven), but it's something vital to make sure that the World-wide Web does not become unsearchable, i.e. that the World-wide Web does not become the "World-wide Mess".
Back to the Table of Contents
Purchase "Intelligence is not Artificial"