In May 2023, Pakistan’s Ministry of Information Technology & Telecommunication (MoITT) floated a consultative draft of the country’s first National Artificial Intelligence Policy. The Ministry further claimed that a Policy Committee would be formed to solicit multi-stakeholder input from members of the academia, industry, government etc., and the draft would be finalised thereafter.
Before delving into the key features of this draft policy, it is pertinent to highlight that the MoITT, contrary to international norms, has solicited public feedback after the initial draft had already been floated. Usually, multiple stakeholders are involved while the document is being prepared. The bureaucratic machinery, in an effort to appear “technocratic”, has left little room for proposed modifications to be incorporated. This can be gleaned from similar exercises while promulgating contentious cyber security and social media rules and guidelines that promoted draconian clauses against freedom of speech and personal privacy; some of these attempts even invited criticism and resistance from the higher courts of law.
Evolution of AI’s Significance in Pakistan
The hype around Artificial Intelligence (AI) has increased over the past decade, but in Pakistan, this began gaining momentum around 2017 onward. It began with a few opinion pieces in institutional publications calling for the securitisation of AI against “hybrid war” to proper governmental initiatives by two different political governments. Near the very end of its tenure in mid-2018, the Pakistan Muslim League-Nawaz (PMLN) government led then by Prime Minister Shahid Khaqan Abbasi, inaugurated a National Centre for Artificial Intelligence (NCAI) at the National University of Sciences & Technology (NUST), followed by a Rs 1.1bn budgetary allocation for select universities (mostly in Punjab and Islamabad, one in Khyber Pakhtunkhwa (KP) and Sindh each); most importantly, NUST was declared as the headquarters from where these research and development (R&D) efforts on AI would be coordinated.
A month later (May 2018), the succeeding federal government of Pakistan Tehreek-e-Insaf (PTI), led by then Prime Minister Imran Khan, approved the Digital Pakistan Policy. This was the first high-level government policy to lay out a plan to set up innovation centres in different thematic areas across the provincial capitals and minor/auxiliary cities, which included AI as a special focus area. The year concluded with the President of Pakistan Dr Arif Alvi, himself a former PTI leader, ambitiously declaring his own Presidential Initiative for Artificial Intelligence & Computing (PIAIC).
On the practical side, it is a rudderless policy driven more by utopian ideals instead of factual appreciation of strengths and weaknesses.
Two years later (during the PTI government) in 2020, Pakistan Air Force (PAF) took the lead in setting up a Centre of Artificial Intelligence and Computing (CENTAIC). The next year (2021), PAF also inaugurated a Cyber Security Academy within Air University, during which the Air Force’s C4I lead also announced the intent to set up an Air Force Cyber Command.
Shortly after the deposition of the PTI government by the incumbent Pakistan Democratic Movement (PDM) alliance in the first half of 2022, the budget was approved to set up a Sino-Pak Centre for Artificial Intelligence (SPCAI) at the Pak-Austria Fachhochschule: Institute of Applied Sciences and Technology (PAF-IAST) in Haripur, which purportedly collaborates through linkages with academia and industries in Austria and China. Also, in the same year, the Pakistan Army announced the inauguration of its Cyber Command, which reportedly consists of two divisions, one of which (the Army Centre of Emerging Technologies) is reasonably believed to include AI in its focus areas.
The incumbent PDM government, through the Ministry of Planning, Development and Special Initiatives, had reportedly constituted a 15-member National Task Force (NTF) on Artificial Intelligence with the purported objective of supporting national development, even before the draft policy was published. The dichotomy is mind-boggling since MoITT has the primary mandate of supervising ICT-related initiatives.
Ignoring the Elephants in the Room
The authors of the draft National AI Policy are surprisingly oblivious or intentionally ignorant of major obstacles to its proper appreciation and implementation (adoption).
Firstly, there is no nuanced conceptual understanding of the AI paradigm. The debates around the actual limits of “intelligentisation” or even “machine consciousness” are ongoing. One camp, which includes longtermists such as the Swedish-American physicist Max Tegmark presents extreme hypothetical scenarios in which intelligent machines will not only come at par with intellectual humans but outgrow the latter’s potentials through leaps and bounds. On the other hand, we have the rather pragmatic views of renowned German psychologist Gerd Gigerenzer who argues through compelling case studies that human intelligence, comprising of common sense, intuition, consciousness etc., will most likely remain outside the scope of adoption within complex algorithms mostly because of limitations in Machine Learning and Natural Language Processing (NLP) methods.
Whatever our subjective views toward each camp may be, it is true that the hype around AI has increased public confusion around the issue. That is precisely why the Government of Pakistan needs to highlight whether it views AI as simply an automation complex that would render so-called “low-skilled” jobs obsolete or whether it actually foresees the likelihood of national stability and livelihood being up-ended by super intelligent machines in the long-term future (autonomous or hybrid); the former approach keeps economic and social security interests at its core while the latter is evidently security-centric and has the potential to absorb paranoia.
Secondly, the entire policy is seemingly predicated on the unwise assumption that a framework for AI development can be implemented within the prevalent bureaucratic monolith. The rustic civil service machinery in Pakistan is rooted firmly in colonial-era nostalgia and mentally stuck in the industrial age, where orders, directives, notifications etc., are still executed through a “corps” of redundant clerks and processed on paper. More importantly, by design or plain convenience, successive generations of bureaucrats have ensured that very few departmental records are digitised. The culture of digitisation has failed to trickle down as it seldom burdens those at the top. Consider, for example, two important actual scenarios:
- Despite the issuance of Computerised National Identity Cards (CNICs), government and private sector offices (especially banks) do not value the data stored in microchips embedded within the ID cards and insist on receiving photocopies on paper as hard proof of identity.
- Cyber Security advisories, apart from routine government circulars and notifications, are not published as electronic documents on departmental web portals; instead, scanned copies of printed documents are uploaded.
It is not improper to suggest that the primary hurdle in digitisation is the bureaucracy itself. Secondly, less paperwork and more automated workflows would make the “clerical corps” redundant, thus placing a “burden” on officers to take direct charge of their departmental affairs. These measures have far-reaching political and economic consequences, which the government of the day would reasonably have no appetite for.
Thirdly, the policy remains oblivious (or perhaps intentionally avoids acknowledging) to the rising cases of arbitrary Internet restrictions across the country. This policy of high-handedness by state authorities on the pretext of law and order disrupts market confidence by causing losses in the billions, shakes the foundations of a knowledge-based economy and ultimately threatens the integrity of the so-called National AI ecosystem itself. Given that the policy prioritises healthcare, mobility, and energy as target sectors for the adoption of AI, such disruptions can literally prove fatal.
Fourth, the policy sets out many lofty targets that are unrealistic and impractical. As Mutaher Khan rightly notes, “For the sake of brevity, a lot of details from the policies have been left out, specifically those relating to the modalities and processes. This is what the whole thing really seems to be about ticking checkboxes. AI is all the hype these days, and countries around the world have produced their strategies for dealing with it, so how can we stay behind? Even if the paper it was printed on will be used to serve samosas”.
Fifth, one of the greatest drawbacks to this policy is its proposal to set up yet another bureaucratic (or “technocratic”) behemoth named as AI Directorate or alternatively called AI Regulatory Directorate (ARD), which would serve as the pivot for AI regulation, coordination, and enforcement. It is ironic that the drafters envision this Directorate to function under a National Commission for Personal Data Protection (NCPDP) since the Act of Parliament necessary to set up such a commission is yet to be approved, i.e., the Personal Data Protection Act (PDPA). The policy seems to be content with distributed and divergent streams of AI R&D being led by the President, Armed Forces, NUST, etc. Should these initiatives be allowed to operate in silos? This aspect needs deliberation.
Assessment
As discussed above, the draft policy suffers from five central/broad flaws:
- Absence of a National Outlook on AI: The draft should have been prepared through inclusive, multi-stakeholder dialogue with all key stakeholders, primarily from the industry and academia, so a cohesive approach toward AI can be developed. This should have happened before the draft was prepared through exclusive coordination with handpicked consultants.
- Bureaucratic Aversion to Digitisation: This inherent malaise is very difficult to be treated and can only be overcome through strict political supervision by the federal and provincial governments working in tandem. It is also important to start preparing for the eventual phasing out of the “clerical corps” only if the state actually intends to adopt AI properly.
- Internet Access and Connectivity: The state should desist from restricting Internet access under any pretext and should find novel ways of targeted shutdowns of certain web links through direct back-channel coordination with the platform hosts.
- Pointless Goal-Setting: The adoption of AI should not be turned into routine departmental tasks since its intricacies and protocols keep shapeshifting each day. It is okay to highlight ambitions/aspirations but absurd to set fixed deadlines or cut-off dates as these are simply impractical in a country like Pakistan, where general awareness of AI itself is uncommon. Such irrational deadlines can conversely encourage the use of improper means to achieve end goals, such as quality control lapses etc.
- Distributed Efforts in AI R&D: An integrated and holistic approach to AI should be adopted for which synergy can be achieved by the appointment of a National Coordinator (or similar designate) operating autonomously and whose office would function with the bare minimum resources simply to coordinate efforts among existing channels instead of forging a new (supra) one which would, in any case, generate further middle channels.
Despite multiple academia-centred initiatives, there is very little public information through which their implementation and projects could be evaluated, let alone analysed. Most of the labs that were haphazardly set up in 2018 have zero or incomplete information. Broadly, the discussion around AI and its significance appears to be more rhetorical than substantial.
Neither the PML-N nor PTI federal governments initiated any inclusive multi-stakeholder discussions from which a shared and truly “national” understanding of the AI phenomena could be derived. As the preceding paragraphs reveal, institutes or centres patronised by Pakistan Armed Forces have been the major recipients of funding for AI-related R&D. There is, thus, an element of securitisation that risks dominating the overall national orientation toward the adoption of AI technologies.
The federal government of PTI slept on the PDPA draft since 2021, and so far, neither of the houses of the Parliament has accorded it the seriousness it deserves. The PDM government is focused more on its survival and internal strengthening than paying heed to legislation that is actually at the core of national security in personal data protection.
Conclusion
The draft policy is a clever hodgepodge of technospeak, with the right jargon and buzzwords (“marginalised women”, “PWD”, etc.) to make for juicy headlines and earn laurels from the powers that be, not to ignore that certain paragraphs were found generated through ChatGPT as mere “fillers” (the more meat, the better). On the practical side, it is a rudderless policy driven more by utopian ideals instead of factual appreciation of strengths and weaknesses; a basic SWOT analysis of existing initiatives would also have sufficed as a foundation for informed policy making.
Pakistan’s bureaucracy is known for its unusual but rare efficiency, but in this instance, it seems to have outdone itself. It would be no less than a miracle if academic members (not those patronised by certain institutions) and industry veterans are able to bring politicians out of oblivion.
