Ramón López de Mántaras |
It has been said that
man is the
only animal that stumbles twice on the same stone. In other
words, it is difficult to learn from history (especially when history is not
taught) and it is difficult to learn from our own mistakes. This is happening
in relation to the field of Artificial Intelligence. In the
previous article I mentioned that the creators of the name of this
discipline predicted that in 10 years spectacular results would be obtained.
Twenty years later, something similar happened when expert systems were
invented. In 1990, Ray Kurzweil predicted in his book The Age of Intelligent Machines
that strong artificial intelligence would come by the year 2000. In 1999, when
he saw that this prediction was not going to be fulfilled, he moved it to 2010
in his new book The age of spiritual machines. As this
prediction was not fulfilled either, between 2009 and 2014 he delayed
it until 2029. It seems that now he is making less optimistic predictions
in this field, and more about immortality, as I mentioned in
another article.
Lately the media are
announcing the coming of strong artificial intelligence, the real one, in just three
years, or at most ten. What do the true experts say about this, those who are
doing research on Artificial Intelligence? Let us look at the opinion of Ramón
López de Mántaras, director of the Institute for Research in Artificial
Intelligence (IIIA, of the Spanish Higher Council for Scientific Research, CSIC).
He has received the Donald E. Walker Award for Artificial Intelligence in 2017;
the EurAI Distinguished Service Award in 2016; the Spanish National Computing
Award in 2012; and the Robert S. Engelmore Award from the Association for the
Advancement of Artificial Intelligence (AAAI) in 2011. This is what he thinks
about all these announcements:
- An article published in 2009
reads as follows:
Artificial intelligence
guru Raymond Kurzweil has predicted that in 2029 a computer will first pass the
Turing test, designed to discover whether a machine is intelligent. From that moment, in theory, it would not be possible to
distinguish a human from a robot. López de Mántaras is not so optimistic.
- At a conference organized in
2013 by the CiViCa association, López de Mántaras
said this:
The other side, general
artificial intelligence, seeks to develop artificial intelligences that have
[the] versatility and general ability to know many things. This does not mean
that their intelligence must be similar to human intelligence; that’s the mistake
of some of these approaches. In fact, this is impossible, in my opinion.
However sophisticated artificial intelligences may be in the future, in 100,000
or 200,000 years, they will still be different from human intelligences.
- In another conference organized
in 2017 by InnovaSpain.com, he said the following:
In Artificial
Intelligence there are many advances, but not as many as the media give to
understand. Basically we have software capable of solving complex problems that
require perception, but there are several examples that show its deficiencies.
Ray Kurzweil |
In a debate organized in 2017 by
the Biocat Foundation in Barcelona, several experts on Artificial Intelligence
recommended caution,
because there are many obstacles to overcome, especially the problem of common sense: all those things that
we know about the world, but that the machines don’t know. I’ll give a few examples:
- If a piece of chalk is split in two, the result is two pieces of chalk; but if you split a table in two, you
won’t get two tables.
- To eat you must be awake.
- You cannot remember the future.
In
a study conducted in 2017 by the Universities of Yale and Oxford among 352
experts in Artificial Intelligence, the average response to the following
question: when
will machines do all our tasks better than we do? was 2140. Speaking
about this, López de Mántaras made this comment: Beyond 10 or 15 years, there is no serious
basis, scientifically speaking, to predict anything. In any
case, it is clear that the experts are much less optimistic than the media.
Another day we’ll speak
about the Turing test.
Manuel Alfonseca
No comments:
Post a Comment