You are on page 1of 3

Format: Blog Post.

Ozone and the Robot Zone: lessons from the Montreal Protocol.

Like others have previously stated, lessons can be drawn from history. In this short
summary, I will explain the success of a treaty itself on scientific cooperation and its
possible use for A.I regulatory matters.

Disruptive technologies have undoubtedly become part of human life, and, while they
propose unique challenges and characteristics, we can draw conclusions from past
experiences and try to emulate successes and avoid their pitfalls.

I thoroughly present a somewhat similar situation in which science, international


cooperation and diplomacy managed to reach a consensus on policy and eventually reach
an agreement on what was feasible and possible.

The Montreal Protocol on Substances that Deplete the Ozone Layer is regarded as complete
diplomatic success in terms of environmental policy. Although different in terms of its
nature the reasons of its success could be applied to A.I, should a need for a treaty arise.

Act early, act fast:

One of the successes of the Montreal Protocol can be explained by the relatively quick
speed in which different governments decided to tackle the ozone layer issue. This principle
is called the precautionary principle enshrined in article 15 of the 1992 Rio Declaration.
which mandates that countries take action even if the science is not yet conclusive.

Although A.I has been in development for several years, it has really taken off in bounds
and leaps in the recent two to three years, reaching the general public with societal
consequences unknown as of yet.1

Currently, there are no binding international agreements in terms of A.I regulations. The
only instrument is a declaration of intent by UNESCO, signed by 193 countries. While
clear in intent, it is not a formal treaty, and thus, it is not legally binding.

If necessary, future international obligations regarding A.I could follow a similar path as
the 1993 declaration by using the same precautionary measures that guided countries in
1987. In case A.I development reaches a dangerous threshold for humanity, the same
principles that guided countries in 1992 would prove useful once more.

Sometimes less is more.

1
https://www.unesco.org/en/articles/unesco-member-states-adopt-first-ever-global-agreement-ethics-
artificial-intelligence
The Montreal protocol wasn’t trying to regulate an all-encompassing environmental treaty
like the COP21 in 2015, which, despite being a complete hallmark in diplomatic history, it
was signed too late when environmental damage had already been done.

The protocol just regulates a very strict and specific type of chlorofluorocarbons (CFC),
those that deplete and damage the ozone layer without extending to other types of
pollutants. This allows flexibility in its adoption and it allows a very clear and
unambiguous commitment to eradicate a common danger to all parties of the treaty.

A.I technologies posses similar characteristics in terms of scope compared to environmental


issues. It is a vast, multi- domain revolutionary technology that will undoubtedly affect
generations to come. As such, any kind of treaty regarding A.I could implement a similar
approach in terms of a specific objective such as regulating ethics, the definition of A.I, the
concept of A.I itself and its application in specific domains such as healthcare, inequality,
development, data protection and privacy. By targeting specific issues, A.I. regulations can
ensure a clear and achievable commitment to eradicating harmful applications of the
technology.

Keep your friends close.

The Montreal Protocol established clear and achievable production limits for CFC products
in articles 2 and 3. These limits were accountable and could be enforced in the short term.
In contrast, the Nationally Determined Contributions (NDCs) outlined by the Paris
Agreement have not succeeded in curbing emissions to 1.5°C and have remained more
often than not in gestures of good will instead of a real commitment towards global
emissions. However, it would be unfair to say that NDCs have failed, as they represent a
significant effort and success in providing a framework for countries to reduce their
emissions to the best of their abilities. Nevertheless, NDCs have remained a long-term,
ambitious goal that has proven insufficient to meet the Paris Agreement's objectives.

Any kind of future A.I regulatory framework must be not only achievable, but enforceable.
Just like the Montreal Protocol, it has to walk the middle ground between being impactful
enough and providing enough leeway to make significant contributions for the betterment
of humanity.

A friend in need is a friend indeed.

Nevertheless, the Montreal Protocol was also flexible enough to accommodate the needs
and realities of Less Developed Countries (LDC)as it shows in article 5. Recognizing that
the success of any agreement regarding CFC’s would have to take into consideration the
socio-economic structure of poorer countries, it also contains a special article regarding the
limits of use and production of CFC’s for them. While flexible, it also provided an
achievable and enforceable objective in terms of emissions.
We could also draw a parallel between LDC and the future regulation of A.I in terms of
breakthrough evolution of technology. As it is, AI language models have the potential to
create a new revolution in terms of productivity, however, revolutionary as it might be, AI
can also be very expensive to maintain and retrain. Any kind of future internationally-
binding instrument will have to undoubtedly take into consideration the benefits for
mankind as a whole, and prevent it from being restricted to a few select nations.

Future challenges and lessons learned:

We cannot assume that replicating every measure of the Montreal Protocol will
automatically create a functioning legal framework for AI. Nonetheless, we can use our
past experiences as a guide to lead us toward a more equitable and optimistic future, in
which our fears over A.I will have been addressed and will remain just that, fears product
of our own prejudice.

You might also like