Socially construct the right AI

…and achieve societal benefit, not harm

The rush of headlines spelling gloom and doom in employment because of artificial intelligence (AI) seems to have gained speed recently: “Within a year, India’s call centres will lose most of their business thanks to AI,” says one headline, quoting the head of one of India’s largest software services companies. Oh god! I thought, the call centre industry employs more than 300,000 Indians, mostly under 30, unmarried, college-educated, and consisting of both men and women. If they lose their jobs, it will be calamitous. Then came a statement from Goldman Sachs saying that more than 300 million white-collar jobs could be lost soon with the rise of AI. And then came an International Monetary Fund blog officially stating that 40 per cent of all jobs worldwide will be lost to AI soon.

Seeing all these gloom-inducing headlines makes one wonder whether employers are embracing AI for this very reason: To help them make more profits by cutting down employee headcounts by getting computer programmes to substitute for them. The reason I am starting to question the direction AI is currently encouraged to develop, namely that of achieving headcount reduction, is because a new line of thinking called “Social Construction of Technology” (SCOT) has started making a strong case that the direction and speed of technological innovation is not self-driven but driven by social forces.
 

Here is an example of how social forces shaped the evolution of an object that we are all familiar with, the bicycle. The earliest version of the bicycle appeared in the early 1800s in Europe and had no pedals. Users had to move forward by pushing their feet against the ground. In the 1860s, pedals and cranks were added to the front wheel of bicycles, and this design significantly improved efficiency and speed. And it was a replacement for horseback riding. By the late 19th century, the wheels were made equally sized and a chain drive connecting the pedals to the rear wheel was created, making the bicycle have better stability and comfort. This was called the “safety bicycle” and gained widespread adoption, particularly among women, who saw bicycle riding as a sign of modernity, and it allowed them to travel to work. This saw the US Army replace their horse-based regiments with bicycle-based ones. Then, in the 20th century things like pneumatic tyres, use of light-weight material like aluminum, and better braking systems led to the bicycle that we know of today. Adoption became even more widespread with the emerging health and fitness movement. And today, the bicycle is seen as a clean and eco-friendly mode of transportation.
 

This study of the evolution of the bicycle has led many scholars to come together under the SCOT movement to study and understand how different technologies have evolved under the influence of social forces. (For the interested, there is a book with a collection of articles describing these efforts called The Social Construction of Technological Systems, edited by Wiebe Bijker, et al).
 

The recent hoopla about ChatGPT and its likes tries to attribute almost magical qualities to such AI technology. However, ChatGPT itself is quite humble in its answers. For example, I posed this question last week: “What is the latest version of Python?” … and I got this answer: “As of my last update in January 2022, the latest stable version of Python was 3.10. However, since then, newer versions may have been released. You can check the official Python website or the Python Package Index (PyPI) for the most current version” …owning up to the facts that (a) its knowledge is dependent on when it was last updated (says that was two-odd years ago!), and (b) that it’s a search engine.
 

One clear plus point about ChatGPT and its ilk (Google’s Gemini, to name one other) is that their creators have made great progress in taking our human conversational input (and not just a keyword or two entered into a “search box”) and giving results back in a friendly conversational style. So, is it this human conversational style that makes folks happily accept such answers as “artificial intelligence”? In other words, the programmers who put it together have deliberately “socially constructed” their software service to use a conversational style to make them human-like. This is an evolution from the “punched card” of the 1970s, then large “keyboards”, the mouse, and the touch screen appeared, then a conversational style of input and output but still via the keyboard, and now seems to be heading in the direction of voice-input and soon voice output in a language of your choice.
 

This brings us to the topic of which direction we should now socially construct AI: To replace clerical/bureaucratic labour (as computers have for the past few decades been trained to do) or create genuine “value added” to businesses? When Jeff Bezos, founder and head of Amazon, was asked this question recently (please note he refers to AI as machine learning, a more pragmatic view), his answer was: “I would say a lot of the value that we’re getting from machine learning is happening beneath the surface. It is things like improved search results, improved product recommendations for customers, improved forecasting for inventory management …”
 

And this has led to continued high growth for Amazon and its employee count, not reduction of employees!
 

Maybe our government authorities, as part of the effort to socially construct the right AI/machine learning tools, should issue a warning to businesses in that sphere as America’s Federal Trade Commission recently did: “Are you exaggerating what your AI product can do?” the Commission asks, warning businesses that such claims could be charged as “deceptive” if they lack scientific evidence, or only apply to extremely specific users and case conditions. Companies are also strongly encouraged to refrain from presenting AI as a means to potentially justify higher product costs or labour decisions and take extreme risk-assessment precautions before rolling out products to the public.

The writer is an internet entrepreneur; ajitb@rediffmail.com