The most critical AI choices may not be about AI

Artificial intelligence (AI) is often perceived as a technology whose outcomes hinge on the decisions made today. Policymakers, enterprises and individuals confront significant questions regarding AI's development, regulation and application, ranging from workforce displacement and wealth disparity to misinformation, cybersecurity and the potential misuse of powerful systems.

However, some of the most significant choices may not pertain to AI itself. They revolve around the systems that accompany it. AI does not emerge in isolation; it infiltrates schools, hospitals, corporations and governments, power grids and data centers, labor markets and financial systems. These systems dictate not only who gains access to AI but also the extent of what can be accomplished with it.

Much of the enthusiasm surrounding AI is predicated on a compelling notion. Technologies once costly and scarce can now be broadly accessible. A student can inquire of an AI tutor about a challenging concept. A modest business can examine markets or automate operations without maintaining a substantial workforce. A junior employee can receive analytical assistance that previously demanded considerable expertise.

These are substantial advancements. Yet, access to increasingly capable software does not equate to spreading intelligence. Intelligence in reality is interconnected. It arises from people, knowledge, institutions, infrastructure, capital, experience, and the capacity to act. Access to increasingly capable software is not the same as disseminating intelligence.

One may encounter two businesses utilizing the same AI model. The first possesses skilled personnel, dependable infrastructure, clean data, financial resources, and managers adept at reshaping procedures around the technology. The second possesses the same model but fragmented systems, limited capital, and insufficient capacity to transform its outputs into action.

Despite having access to the same AI, they do not share identical possibilities. There is an additional assumption worth scrutinizing. We frequently discuss AI as though enhanced intelligence alone is the factor separating existing systems from improved results. South Africans should be especially cautious about this assumption.

A school might have access to an AI tutor, yet that does not resolve overcrowded classrooms, shortages of devices, inadequate connectivity, or the broader circumstances in which children are expected to learn. A clinic might have access to advanced diagnostic aid, yet knowledge does not generate medicines, specialists, functional referral systems, or additional beds.

A small business might have access to sophisticated AI tools, yet those tools do not provide working capital, stable electricity and water, or clientele. We have witnessed this gap between knowledge and action before. Climate change stands as a prime example. For decades, researchers have enhanced their comprehension of the origins, risks, and potential repercussions of a warming planet.

Sophisticated models, more data, and greater analytical capabilities have become increasingly available. Nevertheless, knowledge has not automatically translated into action. Energy systems, political interests, industrial reliance, public conduct, and economic trade-offs persist in shaping how societies choose to act. The lesson for AI is sobering yet crucial.

Increased intelligence does not guarantee superior outcomes. Societies can possess more knowledge yet continue to struggle with implementing it. These examples underscore a shared underlying truth. AI can alleviate certain information and analytical constraints, but it does not eliminate the physical, economic, and institutional obstacles within which individuals and organizations function.

This is one of the paradoxes of the AI surge. We are becoming proficient at generating knowledge, predictions, and recommendations, while many of our most pressing challenges remain issues of coordination, incentives, infrastructure, and collective action. The AI boom mirrors a similar phenomenon during the ERP era. Organizations invested significantly in integrated systems aimed at standardizing processes, enhancing coordination, and providing a unified view of the enterprise.

However, implementing an ERP system did not automatically result in an organization suddenly possessing better processes. Technology could impose structure and improve visibility, yet weak processes, inadequate incentives, and organizational complexity often persisted. The big data era presented a comparable lesson. Organizations amassed more data and developed more sophisticated analytics, yet better visibility did not necessarily lead to better decisions or results.

In numerous organizations, the issue was not merely a shortage of information. It was whether they had the capability and willingness to act on what that information revealed. Higher education provides another illustration. When massive open online courses emerged, they were often portrayed as a means to significantly broaden access to high-quality education.

They indeed expanded access to lectures, course materials, and expertise that had previously been challenging or costly to obtain. Yet, access to educational content did not automatically result in learning. Students still required time, motivation, prior knowledge, support, and the ability to persist through difficult material.

The surrounding educational system still played a role. This distinction bears particular significance for AI. Providing powerful tools widely may remove one barrier, but it does not eradicate all the other prerequisites for transforming access into capability. For South Africa, the strategic query is not solely whether our schools, businesses, and public institutions gain access to increasingly capable AI.

The more pressing question is what surrounds that access. AI will bestow people and organizations capabilities that were previously costly or challenging to acquire. This represents significant progress.