When OpenAI released ChatGPT to the public, some praised it, while others threw brickbats. 2 years on, it seems like many countries have adopted generative AI as a go-to tool. It’s said that India has the highest share of people using ChatGPT, probably due to a higher number of IT jobs for which the LLM could be leveraged.
But, could it be that AI may be seen more from a software perspective than from a product perspective? Could it be that AI has not yet come to the physical world? The thought itself might be a bit terrifying, especially if people have viewed “The Terminator” or “I, Robot” or other pieces of media. People have been talking about the robotics wave penetrating economies the way AI did, but is this still, kind of, sporadic? Are robots still a futuristic concept, despite how much technology has evolved?
So, what development could vindicate robotics a bit more? Here’s one: in November 2024, a robotics startup called Physical Intelligence raised about $400 million from a roster of investors that included OpenAI and Jeffrey Bezos, giving it a valuation of about $2 billion.
And what does Physical Intelligence do? It’s aiming to be a single generalist brain that could control any robot. So, traditionally, robots might have been designed to do specific tasks, like assembly in factories or cleaning floors or helping with surgery for medical procedures. But, with Physical Intelligence, there might be a unified AI system that could control different kinds of robots for different tasks without having to do reprogramming every single time. So, that could mean this might be able to interact with different robotic platforms, like a humanoid robot or just a robotic arm – the kind you might have seen on Iron Man – or other forms.
So, that could, maybe, reduce development costs and create a bit more collab. Imagine if a robotic arm is working along with a humanoid robot. The humanoid robot brings the raw materials to a robotic arm and the arm assembles components.
So, is this the “last mile” of AI: integrating AI into physical tasks? Then, it would have meant AI has penetrated all segments of society that way.
It was said that Physical Intelligence unveiled a model that could train robots to unload laundry, place food in containers, assemble cardboard boxes and even bus tables. This could be interesting, because according to the startup, no robot has been able to fold laundry, for the actions are too complex. Think about how you fold laundry. Because clothing is soft, it’s not rigid, so you have to manipulate it in certain ways to properly fold it by identifying its edges, adjusting the grip and moving fluidly. Even mundane tasks are actually super complicated: you just programmed yourself to do it in a way that becomes routine. Looks like Malcolm Gladwell was right: 10,000 hours of doing something won’t just make you an expert, it could make doing certain things boring.
Or maybe, a humanoid robot greets customers, takes orders and brings the food to the table, while the robotic arm cooks, cuts, assembles and plates. Wait, beyond an Amazon self-checkout store, could there be an automated restaurant as well?
There’s something called “transfer learning”, where models trained for one task could improve performance on a related task. So, maybe a robot, who has the ability to fold laundry could be your kung fu practice partner. Wax on, wax off.
Though, Physical Intelligence seems to have conceded that its robots, sometimes, falter, like dropping bowls when they’re stacking or not properly filling grocery bags. But, it’s still early ages, so it could improve and evolve over time.
For a place, like India, this could be pretty interesting, especially when a lot of stuff is labour-intensive. Would this come in to displace domestic workers? If you can picture workers harvesting tea leaves in Darjeeling, would that be robots doing that, now? Or could you see them sorting recyclables? There’s been talk about robotics automating manual scavenging, this could take this further.
Or maybe, this could change the narrative in terms of India’s software and hardware development. If India’s been seen as the back-office to the rest of the world through software, this could vindicate India to build an equally robust hardware ecosystem. Think sensors or chips or even facilities for testing in India to be exported to the rest of the world.
It’s said that revenue in the robotics market in India could hit about $445 million in 2024. Is that a lot? Maybe, it could be higher. Unfortunately, for VCs, robotics may not be some quick-growth SaaS play. It’s crazy capital-intensive, plus it requires patient capital. Maybe, VCs might find appeal in a platform that builds general-purpose robots rather than niche robots.
And there’s a very interesting survey that outlines that close to 20% of Indian women engage in paid work in a single day, while about 80% of them carry out domestic work. If there are robotics like this adopted en masse somehow, could this pave the way for gender equality in India and empower women to have professional lives? Could it bolster them to found startups to solve problems around them unencumbered by domestic duties? Or if being a housewife is what they consensually prefer, even then, the robots might be able to make things easier for them, maybe even giving women at home more time, energy and a mental breather.
Of course, for technology like this, the high upfront costs might mean adoption is only limited to rich households or big corporations. Even if this were, somehow, priced competitively, would the cost of maintenance or repair curb widespread adoption? Interestingly, would there be robotic service centres the way there are for phones? Would robotics follow the phone route to become cheaper over time with economies of scale?
And could a robot do the things human beings do? It’s one thing to perform tasks in a controlled and predictable environment, but what if there are unexpected obstacles or some kind of human interference?
So, will a Physical Intelligence-esque system make its way to India? And will India bot up or shut up?
