ARTIFICIAL INTELLIGENCE
IT IS 2047, in a remote village. A farmer wakes up and hears the tractor in the field. Sowing, harvesting, or replanting as needed. He walks around with his smart device that continuously assesses his crop health and places orders for fertilisers and plant treatment automatically. It also recommends where to ship the next harvest and what to plant the coming season. The farmer then discusses the recommendations with the device and decides the final course of action.
Later in the day, the children in the house sit with their personal tutor―a smart device that is tuned to their individual learning needs and style. The classes are created on the fly by the tutor for the stage of the learning, that day’s level of alertness and engagement, and performance in the past few sessions. The classes are periodically reviewed by a parent or an expert. When the farmer feels unusually tired later in the day, he puts on his medical consultant―a device that evaluates his vital signs and recommends the course of action. If he needs attention, the device will set up a consultation with a medical expert on a remote platform, which can also escalate to a human doctor. Any medicine is delivered autonomously by a drone from the nearest health centre.
Artificial intelligence is a generational technology. All the above scenarios and more will be made possible, in part, by advances in AI. These examples not only require AI but related platforms and technologies, such as drones and autonomous vehicles. AI acts as multiplier and allows us to achieve far greater impact with these technologies. AI started in the 1950s with the goal of understanding human intelligence well enough so that you can write programs to mimic it.
The goal was not to create thinking machines that would compete with humans. One of the pioneers of AI, Alan Turing, wondered whether a machine could act indistinguishably from the way a thinker acts. This side-stepped the question of what constitutes intelligence and gave a behavioural target for AI. Over time, the field has evolved. The technology used today by ChatGPT is quite different from what was used to build Deep Blue that beat chess champions in the 1990s, though both are called AI. Whenever a significant success of AI happens, tremendous buzz starts building up. This is true especially in popular imagination―recall the Terminator movies and fears of killer robots. Often, these predictions of what AI can do far exceed their capabilities due to a lack of understanding of the limitations of the technology. As we understand their shortcomings better, we typically scale back our expectations. This has led to many seasonal variations in the AI hype cycles and we have experienced at least two AI winters (a period of reduced interest and funding) in the past.
While the current upswing of interest in AI started about a decade ago, the advent of ChatGPT and similar models has boosted expectations significantly. Even seasoned veterans are claiming that the AI singularity, or the point where AI systems exceed human intelligence, is only a few years away. In reality, we are still trying to understand the limitations of this generation of AI systems. The correction to the expectation will come in the next few years.
Given the changing nature of the technology and our understanding of it, it is hard to predict what the future will bring in 20 years. The scenarios described at the beginning of this article are achievable in the next few years, at most a decade. The current state of technology is sufficiently advanced and can support these and more truly futuristic scenarios.
AI is going to make our jobs easy and increase productivity, but is far from replacing humans entirely. AI assistants/avatars/co-pilots will become commonplace―these would be agents built to help with various functions that a human is called on to perform. AI will then become an augmenting intelligence. The cases described at the start are merely scratching the surface. AI will scale almost every aspect of life in ways unimaginable currently. Language, for instance, will no longer be a barrier for communication, and Babel fishes (a creature from The Hitchhiker’s Guide to the Galaxy that allows one to understand any dialect) will become a reality. What will take time is the productisation of these technologies to the extent that they can be used widely and marketed in a profitable manner. Especially in the Indian context. There are many AI services that are being provided in India already. But we need to improve the digital infrastructure and access to technology to see impactful results.
For India to participate and lead the AI revolution, the government needs to improve investment in AI infrastructure and fundamental research. The private sector, too, should invest much more. The government should provide the right commercial and regulatory environment for this to happen. Many powerful tools are being built. They probably would change the way society operates, both in good ways and bad. Skills that were valued a while ago, that needed tremendous practice to get right, might fade away. Creativity would still be prized, but the ability to translate imagination into a song or into a painting might be much easier with AI tools. The capacity to produce very realistic mimicry will lead to erosion of trust in interactions over social media/telecommunications. Even a video call might not be trusted, and one might have to resort to more complex methods to verify identity. If your mother calls, put the phone down and call her again to ensure that is truly your mother calling! Law enforcement will become very challenging as will maintaining security of borders. More powerful forms of identity theft will evolve, as will verification methods. Like nuclear technology, both benefits and misuse of AI will abound. We need to develop nuanced regulatory structures that permit innovation to thrive while minimising the ill-effects. Given the versatile nature of AI, these will impact all walks of life. Just as a person from the 18th century would struggle to adapt to today’s society, we might struggle to adapt to 2047!
Prof B. Ravindran, Heads the department of data science and artificial intelligence at IIT Madras.