AI and Employment: Are They in Conflict?

In the run up to the 2024 general elections, an important issues was the question of growing income (consumption) inequality observed particularly post the Covid epidemic. The epidemic itself led to accelerated use of digital technology for delivery and generated new methods of production. More recently, concerns have been expressed in the media and the government about the use of Artificial Intelligence (AI) in creating "deep fake" videos. In academic and other related circles there have been concerns about AI (like ChatGpt etc) used as "unfair means" in exams, essay writing etc. These two areas of concern have been highlighted in both print and visual media.

AI is also now recognised as a major area of study in academic institutions where courses like Datamatics are now routine and teach the use of AI in handling big data in forecasting etc. as part of standard course curriculum.

What is now emerging is that AI is being used in companies in cost cutting mostly in routine labour using areas like telephone answering services, assembly operations in manufacturing, inventory control, etc. Though clear data is not yet there, there is no doubt that AI (like "capital" in economics) is aiding efficiency even at higher skill levels and hence displacing labour. While the start may be with unskilled labour (used, for example, in telephone answering services), even computer programmers will find themselves redundant in the near future.

In one sense, the hoopla over AI reminds one of the hoopla over the "Y2K" scare at the turn of the last millennium. The need to make computers ready for a switch over to the new year 2000 from 1999 was a great godsend for India. Since hours were expected to be spent in just "keyboard" pounding to ensure smooth switchover to the new system, the "silicon valley" in the USA transferred many hours of work to India and was the basis of the growth of the Indian IT industry. Many years later the "Y2K" issue turned out be a red herring as it created no disruptions in national and international operations. So, the current "AI or nothing" episode may well go the same way.

What exactly does AI do? After much discussion with "experts" in AI here is what I have gleaned. AI, and in particular search engines like chatGPT, Bard etc or their advanced regenerative counterparts are like "super googles". So, if you wanted to generate last 25 years hourly data on temperatures in Delhi and calculate the average peaks and troughs, variance etc. and find average of these averages after correcting for seasonal variation, the AI search engines (paid ones) would do so in a jiffy. No need to get together a battery of data crunchers, econometricians etc. But here is the catch. The AI tools give you a single answer. But you have no idea how that number is generated, whether using charting, econometrics, non-linear methods etc. In fact, if someone was to challenge your answer with a new number there is no way of deciding the correct answer except by the gut feeling of the researcher. However, the use of programmers, econometricians is done away with.

So, unlike traditional quantitative methods, AI gives an answer but cannot guarantee it is a unique answer. This can often be a drawback where issues of optimal efficiency arise.

But the greatest drawback of the current obsession with AI is that it does away with analytical ability. Let me illustrate. While AI may actually generate the desired answer, it is still the individual who needs to determine why you need that answer. The focus on the technology (and large power resources) used to generate an answer in these new search engines diverts attention from the need for that answer at all (shades of the Y2K story?). All the new technologies like genAI, block chain systems (crypto currencies) are all highly energy intensive. Yet, today the means seem to justify the ends.

But, in one sense, it is unlikely that AI tools will ever make labour redundant. For AI is still a production side initiative geared to creating a new product or service at presumably low production cost. But while it may displace labour on the production side it cannot do so in consumption. In general, all supply side technologies displace labour in production but, while creating new products, they do not create new consumers. So robots may become efficient waiters in restaurants using AI tools, but only labour can "consume" their services. Can we see robots serving robots?

In the end, as newer and newer technologies create newer and newer products while reducing the need for labour, the latter will still need to be around to "consume" the products created by new technologies.

So the fears of labour displacement by new technologies like AI are unfounded. While production methods scale higher and higher technological mountain peaks, in consumption the standard labour has no competition.

Has AI run its course? Is the recent slide in shares of Nvidea an indication? Only time will tell. But labour need not worry!

Manoj Pant
Visiting Prof., Shiv Nadar University