The age of uncertAInty: discussing AI’s effect on jobs and policy | The Express Tribune

AI is coming for your jobs…but which jobs,and how quickly?

Associate Professor of Economics at Lahore University of Management Sciences Ali Hasanain held a session on Tuesday at the Overseas Investors Chamber of Commerce & Industry about the Economics of AI.

He explained the current state of LLMs, the potential effects of advanced models on labour markets and implications for jobs, productivity and policy.

With Pakistan’s expanding digital economy and the desire to implement AI on a national scale, these discussions are essential as they will shape new business strategies and guide the workforce to effectively work with AI models.

The sprint of progress

AI can be described as “intelligence everywhere, at a low cost,” according to Hasanain. It recognises specific patterns and learns how to act on them.

AI models have been doubling the datasets they’re trained on every six months since 2010. “The trouble is that we know very little, and the world is moving very fast,” he said, pointing out that there is no theoretical upper limit to how much AI can advance in complexity.

According to Hasanain, these results lead experts to say that AI is approaching the level of complexity in problem-solving that a human has.

However, with this complexity comes the limiting factors: energy costs have increased exponentially alongside the complexity of AI models, there is not enough good data available for training, and models need to undergo scrutiny to ensure they behave safely.

Machine, over man

Artificial General Intelligence (AGI), which Hasanain likens to a computer that can sense and act on any pattern, is the purported evolution of current AI models.

There is contention in the AI community on when, or even if, this will happen, as Chief Executive Officer of OpenAI Sam Altman is reported to have said in 2024 that “AGI will be a reality in 5 years, give or take,” while Vice President of Meta Yann LeCun said, “we are not going to get to human level AI by simply scaling up LLMs.”

Hasanain said that, if realised, AGIs would create both opportunities and risks at large scales. Robots equipped with AGI would be able to do most jobs a human can do.

If the limiting factors of making more complex AI are overcome, the effects on economic structures would be vast. From a decline in overall wages to a larger population because of the replacement of humans in jobs, to more growth in jobs that pertain to managing and making AI-related systems.

Reproducible factors such as making AI chips would become cheaper over time, while irreproducible factors like land, energy and raw materials and “original” goods would likely increase in value, he added.

On the other hand, the time taken to produce and adopt AI and robots, the training and practice required to use them effectively, and the social consequences of replacing certain roles —therapists, counsellors, faith leaders, AI regulatory bodies— with AI counterparts are all barriers AI faces to widespread adoption.

Bracing for AI

Hasanain emphasised that policymakers must start preparing for this reality, regardless of how far we are from it, and detailed key aspects that he felt deserved the most attention.

Inequality of income distribution under AGI due to loss of jobs would have to be answered, to which he proposed work-independent income solutions, like the Universal Basic Income.

The need for education systems to be built around prevalent AI capabilities would be paramount. Many skills would become obsolete, necessitating a re-evaluation of curricula, educational goals and delivery methods.

Mitigating widespread social discontent stemming from the loss of jobs would require the distribution of benefits and strengthening of institutions to withstand rapid change

Changes to fiscal policy, antitrust laws and market regulation would need consideration to prevent a large amount of power being concentrated in the hands of the few, i.e companies with the most efficient models would find it easy to dominate across all sectors.

Environmental protection is a large concern, as running models is resource-intensive. AGI will only exacerbate this. Emphasising alternative energy solutions to fossil fuels, such as nuclear energy, is critical for long-term sustainability.

When asked about the balance between the ethical use or even the non-adoption of AI versus keeping up with the “arms race” of making advanced models, Hasanain said that while AI might be inevitable, the type of AI developed would depend on government regulations. “The kind of laws that will be developed, that story is still to be written”.

“But the history of how technologies get created and operationalised into the economy doesn’t give me confidence,” he added, likening it to the prisoners’ dilemma: In the race to revolutionise the industry, it would be too easy to cut ethical and moral corners in the hopes of staying competitive. Hasanain calls for policymakers to start thinking proactively about our future with AI.