I did not bother to search any literature on the topic, which is immense. Therefore the ideas you might find here are, as usual, not guaranteed to be original.
Anyway I can think that most of the ideas sublimated from
In the beginning this text was meant only to be a short and sharp note but
it then grew into a bigger and vague text, I am afraid even inconsequential,
providing no answers but
half-baked ideas (possibly already chewed by somebody else some
hundred years ago).
¹logics, statistics, optimization
Artificial intelligence is a term indicating several mathematical tools¹
or obfuscative rebrandings
²E.g. neural networks vs optimization
of existing knowledge through biological analogies and implementation technicalities².
I want to explain
- what in my opinion means to judge something 'intelligent'
- how do we get the confidence that something is
intelligent
- why, to me, the
frenzy on artificial intelligence and the associated
millenaristic threats come from misunderstanding,
volountarily or not, the
meaning of the word 'intelligent', which seems to me to be used
as a shorthand
for 'efficient' and 'flexible'. Dr. Sepideh Pashami, whom I met
at the talk in Vetenskapsfestivalen added
'creative' to this when I told her. The three words together
still do not replace 'intelligent' in my opinion.
I will try to
show this in the
following.
Definition of intelligent
The word
intelligent comes from the Latin word
intelligens, present participle of
intellegere
which is the contraction of
inter (=between) and
legere (=choose). Hence something is intelligent when it can discriminate, choose
between different options, do something or something else or nothing.
There are lots of things that seem to be able to
discriminate: living beings, tools... which usually we don't deem
intelligent. I will
continue by refining
a definition for intelligent in the attempt to match
the usual and colloquial understanding of intelligent:
tentatively,
what
makes the best decision to
maximize his well being, of course depending upon the
interpretation of 'best'.
This
definition implies
the existence of a problem and a set of available actions that if
performed cause a change in the world's current state. The
presumed intelligent being must be aware of at least two such
actions, otherwise no choice would be possible. Not acting is
considered to be a choice.
Necessary conditions for intelligence
Hence a being is intelligent only with respect
to a problem it is trying to solve. A
problem
exists when you can imagine a state B of the world which
is
different from the current state A, and you want to reach B.
A problem probes the presumed intelligent being
Tentatively, let's say that the necessary condition to be
intelligent is to be aware of a problem. This assumption is
empirical in that I only know ways to measure intelligence by using
problems.
Awareness
To be aware that there is a problem, shorthanded 👁, means being aware
at once
- of a world, that is, the domain of the problem
- that there are 'states' of the world
- that there are actions
- that actions may have consequences upon the state of the
world
- that there is choice
So being aware of the above is a necessary
condition
to being
intelligent.
Judgement and will
I also think that it is necessary that the being has
- judgement, to tell whether the problem is solved or
not
- will
Judgement can be expressed as an action, and "acting upon
judgement" as well, so I will consider it being part of the item
⚡ and
not consider it separately.
What is will? Whatever makes a being want
to solve, or not solve, a problem. More on this later. Consider
implicitly 'will' as 'free will', unless I explicitly mark the
difference between the two.
Will and awareness are subjective necessary conditions to
intelligence
Hence
- awareness of a problem
- will
imply
intelligence : 👁 ䷀ ⇒ 💡
Remember, the above concepts are just the device I am using to
explain that
there is something we can separate and analyze by verbal means.
Rephrasing: I do not mean that something presumed intelligent
must have a verbal formalization of
'problem' or 'state of the
world' as I did explaining 🌍, ⚛, ⚡, ↷, ⚄
and ䷀.
Before using this definition to investigate reality and look for
other intelligent beings, let's note that
👁 ䷀ ⇒ 💡 is acceptable as long as one applies it to
oneself. After all what said above, some readers may agree
that this implication is true but just because, by means of the
shared meaning we have in common for my words, they
can match
with their own inner experience. It is, in other word, not a
'discovery' but an axiom, a definition, a convention that me
and some readers might accept to have a shared interpretation
of reality. It is easily acceptable because we are used to
attribute awareness and will to our fellow human being or pet,
without further inquiry; actively from analogy (it looks and
acts like me, so it must be like me),
passively from education (golden
rule). No problems nor states.
In other words, awareness and will are qualia and
they are lifted to being objective by the shaky grounds of analogy
and/or convention or...who knows what.
Will and awareness are tentative necessary conditions to
intelligence
The shortcoming becomes apparent when we study the dung beetle:
the beetle forms a dung ball and
carries to its nest. If you remove the dung ball while the beetle
is
rolling it to the nest, the beetle will anyway go all the way to
the nest, unable to stop the algorithm that we expressed to
describe the beetle's behaviour. This
is not intelligent because it cannot discriminate when we do: the
beetle lacks in awareness compared to us studying it. If
we test
the sight and nothing else of a
blind from birth human then we
could
conclude that it is a non-intelligent being. We don't because we
will be more keen to find other grounds to assess the being's
awareness, since we already granted it by analogy and/or
convention. And for awareness it is certainly the presence of a
nervous system that motivates further attempts of
defining
problems to measure this being's intelligence (or of the beetle).
Can something
without a nervous system be intelligent, or aware? I don't know, but it
is a more objective ground for lifting awareness from a
more subjective judgement, given that we agree on
- what is a nervous system
- the ways to experiment/operate it
As you can see, this agreement still comes from verbal agreement
between human beings, but the "leap of faith" seems shorter (i.e.
more people would share this interpretation of reality). Well, more
famous and clever people than me wrote about this, I hope you get the
idea.
This does not anyway tell which are the problems that a being solves,
Where do problems come from?
nor which are
not being solved by deliberate choice... When
lifting
👁 ䷀ ⇒ 💡 from subjective
experience to an objective definition, without
a form of abstract communication (e.g. verbal) with the individual
being
tested
(as would happen by talking
to the aforementioned blind human),
this meta-problem is
insourmontable. We
might look for physical clues, such as appearance or electric
signals in the body of the being under investigation, but this might
be a
limiting assumption because we might be missing the right problems with
which
test the awareness of a being. Could you tell that human beings fall in
love, have suicidal thoughts or enjoy playing games by looking at
just their physical and electrical activity, and nothing else? Yet
subjectively it
is even easy to conjure an explanation for which such behaviours are solutions to
problems.
What is will?
What generates such problems? Will? If will is what prompts to
solve such problems, is formulating such problems
a cause or an effect of having a will? A solution to another problem
itself? Who comes first, intelligence or will? Can we therefore
revert the implication? I don't
know.
Inquisitive
implication
Back to the implication. Its meaning
becomes in practice: if I
can convince myself that this thing is aware and has will, then I can
be more confident that it is intelligent. To mark explicitly this
different
meaning, I will write it as
👁 ䷀ ⤜ 💡
It becomes therefore a tool for investigation.
In
Some examples
the following I will test 👁 ䷀ ⤜ 💡 with some examples. My
goal is to show that, at least in such examples, we reason in a way
that is more or less captured by such inquisitive implication.
Human beings created since long ago machines that automatically solve
problems.
A flush toilet
The flush mechanism of a flush toilet allows you to get rid
of fæces by discharging the water contained in the water tank; and
the float in the tank solves the problem of guaranteeing that you can
flush again
shortly. This system solves at least one problem. Is the
float intelligent? No, because it has no choice ⚄, hence it is not
aware.
Google maps finding the best route
The program that finds the best route on whatever input like that
in Google Maps has choice. We can even attribute it awareness 👁. But given that we know it is a program, this
is not a choice coming from free will; that is, it is
somebody else that decided how Google Maps should behave.
Cargo cult
People
practising cargo
cults exhibit awareness 👁 but their
knowledge of the world is fallacious. But we would still tell
that they are intelligent, because we would see their
behaviour as moved by will.
Extra-terrestrial visit
If extraterrestrials would visit us and had no biological
features, would we attribute will or awareness to them? or rather
would we
look for evidences for them being tools of some other
willed, intelligent being? Would we behave the same if they were
talking cats instead, or would we, in this latter case, rather
assume first that
they are aware and willed?
There are even cases in practice where
we can, even for just a
moment,
grant will
and awareness, therefore intelligence, to a being whose behaviour is
highly
unpredictable or
catches
us unprepared while focused on something else.
yelling at your computer because
it is not working as you expect
For instance you
might get angry at your computer or car when it is not working
properly and you would react in a
violent way as if the malfunction was made voluntarily and
with bad intentions.
Mother nature's wrath
Weather is highly unpredictable;
and in fact some environmentalists
anthrophomorphize nature by granting will
to it. Is nature (i.e. the world minus mankind) intelligent?
Who can tell? It is likely that it is the high stakes (=
risk of annihilation)
that
make this attribution possible, without further inquiries.
The simplest life forms, microbes (including viruses), are
continuously surprising us when they develop antimicrobial
resistance; but no doubt we won't grant them will nor intelligence,
because we would assume they do not have choice and their countering
does not come from their will. As
individuals suffering the consequences of being the unfortunate
host of such beings, anyway, it is not uncommon in desperation to
attribute them a malevolent will. Same applies for other kind of
sickness such as cancer.
Again on will
The case of microbes is a good example to stress how the concept of
'will' probably does not belong to the physical world, but exists in
the abstraction created by our being human, and, to a lesser but
growing extent, with our fellow mammals. We can see
that the microbes want to survive, and thrive, at their
host's expenses. We also see them reacting to countermeasures: it
really looks like they do have awareness of problems (their survival)
and will, but we don't grant them intelligence because we don't know
how communicate with them as we do with
fellow humans and animals. We know, though, other things about their
internal workings which, similarly to the
the
Google Maps example let us decide that their choices do not come
from free will.
Note, then, that with fellow humans and animals we conclude that they
have will and awareness more for lack of such additional information
(or voluntarily closing our eyes upon what is known).
So what are such intelligent machines if not very complicated things
that deceive us to the point that we give up our suspects?
Intelligent is not just efficient+flexible+creative
As I mentioned before, I suspect that when media talk about AI it
uses the following definition:
X is intelligent ⇔ X is efficient, flexible and creative in solving
problems
I want to show that this definition is not enough to define what we
call intelligent.
⇏ :
Cargo cult people are intelligent because we assume they have
will and they surely exhibit awareness, but they do not solve
successfully their problem (getting food from the cargo).
intelligent being's will
⇍ : The point here is subtler and the following reasoning might be
wrong.
Will an artificial being
which is
efficient,
flexible and creative in serving its solely, own purpose be
considered intelligent? Here comes will again as a subjective judgement.
If
Google Maps,
showing creativity, plays a song it wrote instead of telling me how to
get to
the grocery store, it is likely that I won't congratulate it. This is
simply a machine that it is not working, no matter how well its
circuits or software mimick our model of brain. It feels like the
truth of the
above equivalence depends on
the
purpose this
machine has,
and
whether this purpose comes from its own will or not.
If Google Maps was running on an
android working at a tourist office, would it change anything?
Yes, we
could be deceived if we
did not know it is an android; we might think we are dealing with
a stupid intelligent being as it sings its song; but as
soon as we would know that it is an artificial being, then we
would know the android's purpose was given by its maker; and the
purpose is, primarily, to deceive us. Hence, the android is not
intelligent because it has
no will. This is why I think intelligent =
efficient+flexible+creative does
not work.
The "successful"
instances of the
Turing
test
which I read about in the last years do not tell that
machines are becoming intelligent, but rather that we
- have
the means today to camouflage an artificial tool
- consider
as reasonably real experiences that are, in fact, wildly
simplified from those in which we decide whether something is
intelligent or not: e.g. using a
chat computer program versus talking face to face
to fool someone into thinking that he is interacting with
something intelligent and endowed with will for a certain amount
of time.
I hope at this point I convinced you that:
-
If we use the etymological definition of intelligent alone,
then intelligent machines are already among us, because there
are machines that can already discriminate;
- that we deem intelligent something if we can believe it has
awareness 👁 and (free) will ䷀;
- If we use 👁 ䷀ ⤜ 💡
then
- we are not saying anything about other causes of
intelligence
- we attribute will to the being under investigation on
subjective grounds
-
Artificial intelligence that you can buy does not really tell
anything about what intelligence is. It is just something
that has been built to be efficient, flexible and creative,
with 'creative' being, I guess, a more concise
salespeople expression for sparing
all the "it's not a bug, it's a feature" that the
service people would otherwise have to utter during the
product's life.
And if not, I hope at least it was pleasant.
Finally: is there risk of mankind being wiped out by AI? Yes, as
there is risk of mankind being annihilated by atom bombs, cars,
lack of exercise, drugs, wars, fat food, floods, meteors and meteorism.
Thanks for reading. If something was unclear
or could be expressed more clearly
please
contact
me.