Josh Worth provides a critique for the use of the term artificial intelligence in his Stop Calling it Artificial Intelligence, but there are some caveats in the treatment of the term. (One is a typo in the URL path.)
It is not the term Artificial Intelligence that is the issue. It is what is collected under it in the media and sometimes in academic literature.
Is Artificial Unclear?
The term artificial is not particularly ambiguous or inaccurate. Artificial doesn't mean fake as Worth suggests. It simply means that it did not arise from natural processes. High end artificial flowers feel and smell like they are grown. Artificial flight is now called flying. We no longer see birds as the exclusive pilots of flight, so the word flying has changed to include artificial things called aircraft.
People don't normally make the mistake of placing the adjective artificial along side human capabilities, so there is no tendency to misnomers. There is a problem with the use of the term artificial intelligence though. Worth is correct in that overall statement. There may be two forks to the misuse of the term.
Defining Intelligence
One problem is with the term intelligence. The definitions we've seen are considerably problematic. Some are flat out wrong and can be proven so with heaps of counterexamples. Furthermore, definitions tend to be qualitative and difficult if not impossible to quantify.
Some propose the standardized testing of academia to quantify intelligence. If we use that definition of the word, from the staunch g-factor adherents, artificial intelligence is a target idea and nothing comes close to approximating it. No computer system has yet been admitted into a major university on the basis of high scores in college board testing.
If we use the ability to learn as a standard, then flies are intelligent because they frequently learn from one swat strategy changes to avoid getting squashed by the irritated person. There are a thousand other reasons why learning, by itself, is not an adequate characterization of human intelligence. A crack addict can learn how to make a purchase without holding a job. We wouldn't characterize that as intelligence, but rather dysfunction.
Excessively Inclusive Use of the Term
Another problem arises from avarice. To appear as an expert in what is perceived to be a lucrative expansion of technology, some who have no conception of the topic sometimes present conjecture as if it were peer reviewed fact. This has been typical of popular topics for centuries. There is often insufficient bandwidth of peer verification available to address even a small portion of publicly available information. Web publication has only augmented this problem.
Resulting from this coveting of expert reputation is collecting under the name artificial intelligence a number of things that are not intelligent.
- Multi-dimensional control systems
- Brute force searches of permutations
- Decisions made by statistically analyzing sample data
- Parameterized functional networks that converge to a defined optimal behavior
Those of the above that have no component of
- Cognition,
- Comprehension,
- Complex modelling and use of such models,
- Semantic mapping,
- Rational inference,
- Or some other clearly distinctive and broad form of adaptation
should not be included under the technology that exhibits authentic artificial intelligence. However, those that developed the theory of control systems, searches, parameterized function convergence, and statistics or developed working systems that use that theory are intelligent, just not artificial.
The above four might have terms that distinguish them from authentic intelligent systems outside the realm of biology.
- MDC — Multi-dimensional Control
- BSS — Brute force search
- NBL — Network Based Learning
- SDS — Statistical decisioning systems
It seems that terms that have three words and form distinct acronyms get further and last longer.
