Current public anxiety surrounding the establishment of AI data centres in T&T appears to be emerging from multiple phenomena.
There is first - and not to be ignored -
a general, irrational mystique surrounding the value and quality of generative
AI and its constituent components. Somebody messaged me last week with a simple
query: “Cell tower debate again?”
There is however an accompanying absence
of public trust in the conduct of state business. There are questions regarding
where, precisely, such projects are situated on the country’s development
agenda, especially given their prior absence. And there are also broad
socio-economic questions to be addressed.
There is now a recognisable convergence
of these discrete elements of the current AI data centre challenge.
The increasingly popular Syam Nath
petition quite rationally addresses the specific issue of water usage. AI data
centres are not, you see, ordinary warehouses for data. They are
resource-intensive facilities that require comparably far more energy and
cooling than other plants.
The petition also significantly includes
a general appeal for “responsible planning, transparency, and evidence-based
decision-making” when projects with implications for environmental instability
– the management of water resources in this case - are being considered.
Yes, the petition has attracted some
exaggerated fears over AI itself, but it also tangentially addresses the key
questions of trust and the way orderly economic progress needs to be engaged,
albeit with a primary focus on ecological viability. There are also broader
socio-economic considerations that deserve attention.
We are no newcomers to such questions.
For instance, our energy sector has over recent decades confronted questions of
the contribution of industry-generated carbon emissions to global climate
change.
This country is currently part of a
wider, hugely challenging, effort to balance economic benefit against what
scientists conclude is a need to reduce the carbon footprints of all countries.
Few periods have presented comparable developmental dilemmas for nations such
as ours.
They certainly did not arise when the
first oilwell was excavated over 100 years ago, nor were there such questions
in the boom years of the 1970s.
To emphasise this, the energy sector, in
Guyana is indispensably assisting in the achievement of an unprecedented level
of socio-economic development.
That neighbour’s language on the
achievement of global carbon targets has thus become far more temperate than
previous non-oil years, even when T&T was both tacitly and openly giving
expression to the dilemma.
The question has become one of “net
benefit” rather than of absolute hazard. The concept of “net-zero (carbon)
emissions” and accompanying measures in part capture the quandary through
tabulation of a notion of global “balance.” The activities of AI data centres
have also recently joined the emissions discussion.
The focus for us, I would suggest,
should be on net potential benefits based on science and transparency. None of
the larger producers has completely abandoned their energy industries, even
those who are benefactors in the framing of a new way forward.
Raising related questions on AI data
centres should not, however, be dismissed through condescension or insult.
These issues arise partly because of a sense that development is being engaged
by “vaps” and improvisation.
Although there is now a Ministry of
Artificial Intelligence - which is largely concerned with integrating digital
tools into public administration - there is informed opinion suggesting that
merely hosting AI infrastructure is not the same as pursuing national AI
capacity.
To address this would require an answer
to the question of whether the proposed projects can serve as the foundation of
a sovereign, nationally controlled language model serving governance and,
eventually, the public.
The difference between the latter and
the mere hosting of remote facilities is enormous, and the country cannot
meaningfully debate scope, risk, or net benefit until it knows which objective
is to be pursued and the scientific metrics involved.
There is, meanwhile, no reason to
needlessly hammer prospective partners. Hummingbird AI, for instance, has
openly declared its interest as a corporate entity with a capacity to
“originate, structure, finance and deliver nationally significant AI infrastructure
projects.” Fair enough.
There is little doubt that the
significant investment involved can and will, at least in the short term,
generate the benefits of foreign currency inflows, jobs, skills acquisition, and
infrastructure development.
There remain technical, contractual, and
operational details, however, that should be disclosed, under conditions of
full transparency to the public so that a rational assessment of net benefit
can be made. There are undeniable pluses and minuses.
We are not there yet. Until we are,
there is nothing unreasonable about identifying and debating the principal
areas of concern.
