0. Introduction
The last remarkable innovation
is “artificial intelligence”. This innovation, like all the previous ones, is
subject to both, and sometimes at the same time, praise and criticism often out
of proportion. What seems to be new today is that it is a machine that does
intellectual work and no longer manual. But is it really new? And is this
machine really smart? We’ll see... We will also ask whether the problems raised
by this innovation are not, as with previous ones, more ethical and political
than technical.
1. Intelligence and
problem solving
“Intelligence” can mean
the ability to solve a problem: e.g., finding a way out, solving an equation,
correctly answering a question, understanding a speech, winning at chess (it is
a more pragmatic definition than the one found in the dictionary, such as
«mental function of organization of the real in thought»). “Intelligence” can
be that of a human, collective, animal or machine.
There are different types
of intelligences according to different beings and also for a same being. A
human being can have a relational, practical, creative, logical intelligence,
etc. Finally, intelligence is also the ability to use a variety of forms of
intelligences, that is to say, to alternate and combine these different forms
of intelligences.
Second, to say that
someone is “intelligent” is to make a positive value judgment. And if we talk about
an «intelligent» object (game, book, accessory, machine), this values the
object as well as its creator, its user or its owner. Then they get a power
(even if it can be misused).
2. Artifact as an
intentional product of Man
On the other hand, to say
that a being or thing is “artificial” can have a negative connotation and mean
“false”, “overdone”, “unauthentic”, “useless”. It is considered to be an
inferior copy of its model (e.g., a flower or artificial aroma).
But this connotation is
not necessary. The expression “artificial intelligence” generally refers to
machine-generated intelligence and nothing more (although the positive
connotation of “artificial intelligence” can be contrasted with the negative
connotation of “artificial intelligence”). The connotation of “artificial” can
even become positive if one recognizes, through this word, a feat of human
genius capable of creating an intelligent machine.
The term “artifact”
avoids the connotative interference attached to “artifice”. «Artifact» refers
to the products of human intelligence: tools, instruments, machines, works,
institutions. If the artifact is a product of human intelligence, and if this
artifact is also intelligent itself, then it can be considered as an extension
of human intelligence, that is to say an indirect way for humans to be
intelligent. If artificial intelligence were to be considered more intelligent
than human intelligence, it would constitute a remarkable feat of human
intelligence.
Unlike the artifact, the
natural product exists independently of man, and more precisely independently
of his conscious and voluntary action. E.g. a handprint on a wall is not an
artifact if it is not intentional. Natural products are animals, plants,
solids, liquids and all biological, chemical and physical processes that exist
by themselves and without our assistance. Man can also be counted in some way
among natural products.
What constitutes the
artifact is therefore the result of human intelligent action, that is to say,
Nature intentionally transformed by Man. Human side-effects such as pollution,
waste, accidents or “misalignments” (AI’s unintended behaviors) cannot then be
considered artifacts - even if it seems unintuitive to call them “natural”.
3. The difference between
human and artificial intelligence
The artifact becomes
“intelligent” when it can autonomously perform operations that recall human
intelligent activity: calculating, playing music or games, detecting and
alerting, pausing or running, translate, answer orders or questions, correct
his errors etc. Sometimes the artifice is much superior to man (speed of
calculation, memory storage). But it remains specialized in specific tasks,
where man performs an incalculable number of different operations.
Intelligent artifact can
be considered as a symbolic machine as opposed to an instrumental machine. Machine
represents a technological evolution based on use of a motor instead of muscle
strength. Machines are developments in instrumental techniques related to
manual activities (e.g., machine tools or vehicles). Symbolic techniques, such
as writing, have in turn evolved into mechanical forms (e.g., printing,
typewriters, telecommunications, calculators, etc.). Symbolic machines process,
as contents, graphic, sound, numerical and alphabetic symbols. But they could
not do so without, as instrumental extensions, their receiving or effecting
interfaces to interact with the outside world. The development of these interfaces
leads to robotics, when inputs and outputs of the processor become more
elaborate than a keyboard and a screen. We can say that robot is the body
development of computer.
This raises the question
of continuity between the instrumental and the symbolic. Is the symbolic
miniaturization of the instrumental? Can we reduce operations on symbols to
causal processes? Is the process between the keyboard and the calculator screen
only one continuous mechanical transmission? Can we say the same about what
happens between eye and hand when e.g. calculating on paper? If so, how do we
explain consciousness, freedom or creativity?
These questions refer to
analyses of philosophy of mind which will not be dealt with here. What we can
see now is that a technical evaluation of artificial or human intelligence
consists in measuring their performance from a mechanical point of view, in
terms of power, speed of calculation or intellectual quotient. In this respect,
machines can be much more efficient than humans (just like cars, trains or
planes when it comes to travel). For example, an automatic translator can also
be estimate, to determine whether it is translating correctly (even if it’s
done without skill and sensitivity of a human being), just as it can be used to
determine if an aircraft is flying properly.
4. Ethical review and
innovation neutrality
If the technical
assessment tells us whether something is doing an action effectively or
correctly, then the ethical assessment is about whether that action is good. A
scammer can act “well” technically, but it will not be said that he acts “well”
ethically. An action, we can say, is absolutely ethical if it is good for all
men and for the world in general. But this is no more precise than saying that
an action is good if it is good. Saying e.g. the medicine is good and the
poison is bad only clarifies their definition. This is no more instructive than
saying that health, joy, happiness, well-being and friendship are good in
themselves.
And if we are talking
about poisoning a dangerous tyrant, it is in principle only a question of
denying the negation, as «punishing the guilty», «eliminating the evil»,
«eradicating the disease», which is equally abstract. But if we approach ethics
from a concrete point of view, evaluation becomes relative. What is good for
one here may be bad for another there; what is good at that moment can be bad later
and vice versa.
From an abstract point of
view, regardless of the particular situations, the existence of most
innovations is not bad in itself. In other words, they are neutral. This is the case for e.g. automobile,
computer, telephone, television, refrigerator, backhoe, etc. Ethical issues
arise when it comes to qualifying inventions against facts. They become
sometimes good, sometimes bad. Automobile e.g. does not automatically guarantee
the happiness and well-being of humans.
A radical way to solve
ethical problem is to eliminate its object, as if one cured the patient by
simply making him disappear. If we get rid of certain technologies, then will
disappear, more or less quickly, their perverse effects (with their salutary
effects). This is comparable to what philosophers do who, instead of answering
a question, say that it has no meaning. We must then make sure that this is the
case, otherwise it would only be a rhetorical parade. There are many
innovations that look like problems without solutions, those that are negative
by definition: weapons of mass destruction, robots-killers, torture
instruments, pesticides, etc. But to reject computers, phones or cars, as
fundamentally bad, seems difficult to justify (but not impossible). Nevertheless,
just as it can be shown that a question may have been asked in the wrong way,
some innovations can be regarded as having been misplaced and needing to be
conceived differently. This is the case of products containing asbestos, lead,
DDT, radium e.g. etc.
5. The ethical response
of regulation
Most of the ethical
problems we face cannot be completely solved or eliminated. It seems
inconceivable that we would deliberately abandon most of the technologies we
use, nor that we could permanently eliminate the risks they pose. We then have
to try to regulate them as well as possible. For this, it is necessary to
intervene on several levels: upstream, as when the motor of a car is derated;
downstream, with retarders on roads, signs, radars, gendarmes and judges. The
regulation responds to a desire for safety in order to contain the overflow
(e.g. speeding) or prevent the occurrence of a feared phenomenon (e.g.
accident).
Regulation involves rules, which define limits, and means
of application: instruments, incentives, sanctions, controls, corrections etc.
(some rules appear to regulate the previous rules, such as those that limit
abuse of power). Failure to apply a rule is equivalent to an unrealized
architect’s plan or the composer’s score not performed.
6. Freedom as regulation
A naive or dishonest
conception of freedom is to equate it with the absence of rules, instead of
distinguishing between rules that are approved and those that are disapproved
and experienced as oppressive. There are e.g. grammar rules without which we could
not speak and social rules that can be perceived as a barrier to freedom of
expression.
At the political level, a
rule legitimately interferes with our action when it guarantees protection
against arbitrary interference. A rule is legitimate if it can be recognized by
all as good, which means that we must be able to recognize that it can be
useful to anyone. The rule applies to any person, universal and possible.
The motivation of the search for a balance of interests
and skills is the possibility of being harmed by an imbalance. The aim is to
set aside what makes me or would make me suffer if it happened to me. Morality
is generally described as selfless. But in a sense it is hyper-interested: to
act as if what happens to others could happen to me too.
7. The democratic
regulation of human and non-human
The ethical rules must
make it possible to minimize or even abolish domination, exploitation and
ignorance; this in order to reduce or even eliminate suffering. They are
democratic in essence insofar as they take into account the expertise and
involvement of each person, rather than a paternalistic authority. This is what
differentiates the absence of suffering, through the free development of each
one, from anaesthesia as elimination of sensitivity. Democracy means power for
and by all, and should therefore extend to all levels of decision-making, in
infrastructure and superstructure, in society and the state. It is a matter of
balancing evaluation and decision everywhere in an optimal way.
Democratic activity is
about the relationship between humans and their relationship to their natural
and artificial environment. In other words, the decision-making circuit
integrates questions of practices to be adopted vis-à-vis humans, animals,
machines, plants and land. And it is through human speech that the non-human
expresses itself, as vulnerable or threatening.
8. Conclusions
We invent tools that increase our problem-solving
abilities and thus our intelligence. To this end, symbolic machines and
intelligent artifacts imitate and multiply certain aspects of human
intelligence. It is a virtuous circle: human intelligence produces instrumental
intelligence to develop itself (e.g., calculi, abaque, pascaline, etc.). These
techniques and machines designed to solve problems are therefore good by
definition. Judging artificiality pejoratively is a matter of prejudice. A
human without artifice is not a human. But attribution of human qualities to
machines, such as consciousness, freedom or creativity, is fiction.
Regulation is a human
activity of solving technical and ethical problems. It uses instrumental and
symbolic techniques that need to be regulated too. The purpose of these
regulations is to direct innovations towards the good. Thus, human intelligence
allows both innovation and regulation to avoid the dangers associated with
these innovations.
Regulation is presented
as a therapy, following a critical diagnosis of innovations and their harms.
These emerge from social practices linked to the uses of innovations and the
vicious circles in which they are caught. But the important question is who makes
the diagnosis and the therapy. Because innovation criticism can respond to
different underlying strategies. For example, some criticism of innovations,
especially when it comes from bodies curiously interested in their development,
can be used to demonstrate their effectiveness. Or criticism of innovations may
be aimed at masking the human responsibility of some authorities. These are all
reasons to defend democratic approaches to the diagnosis and treatment of
innovations.
More generally, democratic procedures
must be able to deal with the complexity of phenomena and to accommodate the
diversity of situations and their changes in time and space. Democratic
regulation should be carried out on various scales: first, at the individual
reflexive level, with self-government; in the family and community, with daily
discipline; in the organization of work and collectives or associations; in
institutions and civil service; and in international bodies. It is the
combination of these different strata that should make it possible to contain
the negative externalities and counter-objectives related to innovations. Let
us add, as virtues inherent in democratic regulation, the fact that decisions
taken democratically have a legitimate character, and that democratic practices
develop intelligence. Technological change alone cannot contribute to the
construction of viable societies if it is not accompanied by the development of
democratic practices.