Q4A: to continue from my last post:
*how will India compete with USA and China in ai*
Typically the first round in any new cycle goes to the early movers. The final leaders emerge later on. Toyota and not Ford leads in automobiles. Apple and not Sony leads in smartphones. Android and not Symbian leads in mobile OS. And Byd and not Tesla leads in EVs.
Conceptually I expect ai to follow the trend. Finally ai is what ai does.
And for ai to deliver it needs data, it needs training and it needs use cases. There is no second source to India for this. China can do parts of it but not most of it. USA is a lawyer country and their courts will delay their progress by decades. Take reinforcement factor human training – India is by far the global leader. Or gene fasta file interpretation. Finally we may not have a google maps but both Google and Apple use Genesys International to produce their maps.
India doesn’t have to be everything to everyone. We have different metrics. The ai cost of replacing a blue collar labourer for mental work is Rs 28,000 pm. The cost of replacing him with a humanoid robot is Rs 80,000. These numbers spell the death knell for workers in the developed countries but not in India. In fact by using ai and working in hybrid mode by using frugal ai systems – the cost of Indian labour can further reduce substantially making it even harder for ai to catch up. But population scale unemployment in the developed world is a reality they cannot run away from.
In fact I often joke that there will be so much unemployment and mental health issues in the developed world that people will have to come to India for wellness. And wellness will become way larger for dollar exports than IT. Wellness is also high value add, addictive and accretive on forex. There is hardly any import content. So wellness is very ai resilient
The other area is improving productivity of humans and enhancing their capabilities. I spoke about a plumber assistant from Amroha who is now a global authority on neuromorphic ai. His company was just “me and a phone”. And he had used this phone to do everything- build a foundational model, train the model and run inference. His model doesn’t need a data centre, network or cloud. This is anecdotal. But indicative.
The other issue is the role and route of technology itself. Sure you can use a machine gun for everything from shooting your enemy to shooting a mosquito. But this may not be the best way to shoot a mosquito. And not the ideal way to eliminate your enemy either.
A GW of data centre is going to cost USD 50 billion. You build a thousand of these u will need a TW of power which is no big deal given that China produces that much of solar annually. But how will you service the usd 50 trillion. One way is to say the USD itself is a Ponzi scheme and milk it till it dries out. I agree with the Ponzi scheme part but don’t see how you will milk it.
To be contd.
Blog.satishmehta.in