AI and automation: The real questions that only a few are asking

From inventing the wheel to the steam engine, from igniting the first fire to the modern-day computer, the history of human civilisation has been closely connected to technological inventions. However, for a really long time in this history, we never even considered technologies to be automatic, let alone intelligent. The reasons were manifold. Those technologies were made for a singular purpose; for example, the steam engine for power generation, and the hydraulic press for compressing things with immense force. However, completing a whole production process requires performing multiple tasks and making decisions, and hence always requires a human being, at least to operate the machines. 

With the rise of computers and algorithms, we thought that we could now maybe automate tasks by programming a system step by step on how to perform them. But computer coding requires a precise command based on a specific scenario and cannot dynamically adapt to the task at hand most of the time. A realistic production process, being very dynamic and complex, thus cannot always be pre-coded or commanded.  

The question remained and still does: "How do humans perform these tasks?" Artificial intelligence and advanced automation have tried to capture the essence of the answer to this question. A human being learns to perform a task by seeing a lot of guided samples of that task being performed. However, sometimes humans also perform tasks that are not guided, by utilising trial and error or logical reasoning. AI systems try to utilise just that— supervised or unsupervised learning based on existing datasets of actual tasks performed repeatedly by human agents. How successful it became is evident now to all of us. 

The revolution in our life: Threats of the Great Transition

AI came with a huge number of effects, challenges, and future possibilities for our society, politics, state, and even philosophy. In previous technological revolutions, we saw that the requirement for human labour and human skills in several sectors became obsolete. We may expect the same to happen now, and we are even seeing it happen in front of our eyes. 

By 2030, Goldman Sachs projects that the USA will have 8% of its ride-sharing services driven by driverless cars. China has made several significant leaps in the automation of industries. So, the question arises: if we keep replacing human labour with artificial intelligence, what will happen to the huge number of unemployed human labourers? Also, if there is no employment, people cannot buy products, even if machines are automatically making products.  Thus, the basic profit-driven economy is under threat. 

In our past industrial revolutions, we saw that people experienced a significant increase in their demand for different products. Many new products were invented, and many luxuries slowly became a necessity of some sort. So even though our productive powers increased, human labour still found new workplaces and shifted skill sets to maintain the increased demand. Something similar can be forecasted now. It is assumed that as people cannot rely fully on AI-based services, human labour and skills will shift significantly to service markets. Also, the supervised AI training and research sectors will employ a huge portion of labour. Developed countries already have a large service market and are working on ensuring that their industrial labourers get proper training to match this shift in skills. 

But for Bangladesh, which relies on a highly monolithic industrialisation focused on RMG, agriculture, and less-skilled labour, the challenges are very high. With our broken research and innovation infrastructure, as well as a lack of independent national industrial capital, we are going to become even more economically and industrially dependent on the more powerful states. Also, this very transitional phase of the new industrial system will cause a significant humanitarian crisis for the huge number of labourers we have. Many of them will have to pass through the most inhumane and uncertain life. 

Even in the cultural spectrum, AI will, and perhaps is, causing significant issues. AI, being a statistically trained tool, represents the cultural and ideological biases present in the training data. Sadly, most of the written documents and works used to train large language models are in English and represent cultural alignment of the West. Other types of artificial intelligence also carry such bias. For example, the agricultural data used to train a crop yield forecast will be taken from a foreign dataset that reflects the environment and the climate there. However, this bias is already widely discussed among researchers, and people are trying to resolve it as much as possible. But the real and complete solution will require being able to generate sufficient training data in our local context. 

Does Bangladesh have the capital and infrastructure to generate or gather these data? Probably not. So, Bangladesh, its people, and the culture will keep falling behind in the AI race until we give it a real thought.

The point of it all 

The AI revolution brings a significant philosophical question to light. Even though the increase in our needs and demands is less than the increase in productive capacity caused by technological development, we do not see any global reduction in our working hours. Instead, we see a rise in artificial demand for rather unnecessary products, huge wastage of production, and meaningless jobs. 

Some of the developed countries did reduce working hours and working days, but only at the cost of increasing working hours in the developing and underdeveloped countries. In fact, these developed countries have shifted most of their production lines to underdeveloped countries for bigger profit, which even triggered the unemployment crisis in those states. 

In Bangladesh, we often see that the working hours for both white-collar and physical jobs easily extend beyond eight hours without any proper overtime.  So, this brings up the question of human essence. In a society where we don't need to spend our labour producing our livelihood, what will humans do? Will they resort to arts and literature, caregiving, and social work?  And if the goal of our technologies was to ensure that we have more free time in our lives, are we not deviating from that even though we are so close to the goal? Without more free time, do the comforts and luxuries of our modern life matter at all? Is it perhaps better to live in a society that has a centralised production system providing for all, and human beings have the freedom and time to explore whatever they seek?

Some people go so far as to argue that we are moving to a new production system of the world, technofeudalism. Technofeudalism is the monopolistic control of technology leading to monopolistic control of the world. Surely, the audacious behavior of people like Elon Musk and Sam Altman may give us the feeling that they are the overlords of our society, but this is not the case. 

Even though technology has rapidly changed, our economic system has not. Our economy still relies on finance capital, and monopoly is not something that only came with Alphabet (Google), Meta (Facebook), and OpenAI (ChatGPT), etc. Monopolies are common and already highly systematised in sectors like oil, industrial raw materials, energy, etc. Also, finance capital-based investment groups eventually control the tech industries totally. We see how buyouts and other aggressive corporate strategies caused initiatives like WhatsApp and Instagram to be bought. Also, the power of advanced innovation mostly lies in the power to manage the capital for that research. So, it is not that we are moving to a technological monopoly because of a socio-economic transition. Rather, monopolies and almost feudal domination of oligarchs are the traits of the very economic system we are in now. Without a change in the economic model, this won't change.  

But all of these rapid developments do have a fundamental challenge that is very little talked about so far. Just like other advanced technologies, AI and automation technologies require a lot of natural resources. Making the required hardware requires raw materials extracted from the nature; running the servers requires water and air for cooling systems, and everything requires a great amount of energy. This is already putting significant stress on our environment, and concerns about large language model data centers are rising very rapidly. This is becoming even more asymmetric for developing countries like us, as many of the big corporations are looking to move their data centers to countries like Bangladesh where the environmental protection act is often bypassed for the sake of employment. 

What the future holds

A big question, therefore, arises on whether full automation is something that should even happen or not. The human brain is not only very effective but also efficient from the perspective of energy and other resource requirements. In order to mimic a simple task that can be done by humans, if we have to spend more natural resources, then it will be very problematic. So, the technological challenge now is to see that, under the constraint of limited natural resources, how much of production we can automate. Also, how we can make these technologies achieve an efficiency level comparable to that of biological beings should be a big question for the future. But the biggest question that hangs over us about AI and automation is how we are going to adapt it to our society and if the existing profit-driven mode of production is sustainable for a highly automated dream society of mankind. 

Tahmeed Hossain is a student of theoretical physics in University of Dhaka. He is also an organiser of Democratic Students Council, University of Dhaka.