Author Note: This article is a recap and summary of a very interesting panel discussion on how tech leaders are thinking about the AI revolution and its social impact in light of Rajiv Malhotra’s book “Artificial Intelligence and the Future of Power” published in January 2021.
- M. Vidyasagar, Fellow of the Royal Society, Former Director, Centre for Artificial Intelligence and Robotics, DRDO
- Prof P. R. Kumar, Regents Professor, University of Texas A&M
- Shailesh Kumar, Chief Data Scientist, Reliance Jio
The discussion was moderated by Gopinath Kanchi, Computer Scientist, Indian Institute of Science, Bangalore
Gopinath Kanchi introduces the participants, after which Rajiv Malhotra starts with an introduction of the book.
First, he says that while the positive aspects of AI are many and these are also much talked about, that’s not all there is to it. The impact of AI will be unequal between individuals, industries and even countries, leading to digital colonization. Many people will be caught in the middle of their careers, they’ll lose their jobs and won’t be able to retrain. The previously established equilibrium will be destroyed, as the new technology empowers some, while others are discarded either because they can’t access it or can’t afford it. And until a new equilibrium is reached, maybe 25-50 years later, there will be social unrest that could turn violent.
Next he compares the AI disruption with the Industrial revolution in a couple of ways. He says that the industrial revolution transferred jobs from India to Britain, leading to our impoverishment and colonization. This unequal impact will also be there in the AI era. Additionally, because capital was a limiting factor and communication slower, the farm to factory transfer of labor during the industrial revolution was spread across generations. So people weren’t disrupted in the middle of their careers. But technology adoption is very quick now and the AI disruption has already started, so jobs will increasingly be at stake.
Third, the reports discussing AI’s impact on India (Niti Aayog, FICCI, Pricewaterhouse, McKinsey, Ernst & Young) take a top-down approach, presenting the corporate view. But 90% of Indian workers are small and medium industries (SMBs), or self-employed. So there’s a concern that Indian issues at the grassroots, at the panchayat level, at the state level, aren’t being highlighted. Even country level issues aren’t discussed, with most of the reports presenting the western viewpoint.
Then Gopinath asks a question about the massive job losses, especially at the mid-level managerial level. He wonders if Indian technology workers will be a casualty of AI.
Rajiv says there are two issues with respect to the AI opportunity for Indian workers: first, we are only considering AI’s impact on those who will be in the AI business, having taken an AI course and being trained in it. But this will represent less than 25% of the total population. It is the bottom 500 million people like the local store or street vendor that will be disrupted by the Amazons of the world. And then, even with respect to AI workers, the general goal of India’s thinkers seems to be that we supply the brains (labor) to make American, British, Chinese or any other intellectual property (IP), but never become knowledge producers creating our own IP.
Shailesh Kumar joins the conversation here.
He says that a technology isn’t good or bad on its own, but it’s what we as human beings make of it. AI is positive in the sense that it solves problems. He offers 3 examples of this-
First, a platform like Ola helps travelers and drivers meet, so drivers find travelers and riders find taxis more easily. It doesn’t disrupt jobs. Second, a model like Jio’s doesn’t take the kirana stores out of the loop, but instead includes, digitizes and thus helps them manage their inventory, pricing and supply chain better. Third, he talks about the positive implications for agriculture. For example, if we knew what is growing where across the country and when it would be available, we could increase the efficiency of the system.
On the negative side, he talked about the impact of IT labor, which is today being used to do low-end jobs like labeling the data for use by AI machines. There’s no future in this work.
M. Vidyasagar says that today we are a point and click culture and that part of jobs is not going to change, but what happens behind the screen as a result of the click is the change that AI will bring. He’s also confident that since AI is open source, we will continue to have access to the tools that can be used to write these software programs. Even connectivity, which is an important requirement, is not such a big issue today (and Jio had a big hand in this). He says that some people were using the free tools to build applications, but they wouldn’t be the ones being interviewed .by organizations like McKinsey’s. He thinks the Indian government shouldn’t frame an AI policy but should instead provide the tools, networks and interfaces in all Indian languages to make it happen.
Professor P R Kumar is next. He praises the book, recognizes the call to action and appreciates the loose, inclusive definition of AI in the book before moving on to the topic of ethics in AI. He says that India’s long tradition of thinking deeply about the ethics of politics, warfare, administration, etc positions it well to consider the uses and ramifications of these new technologies. He suggests that when researching or writing on these topics, the negative ramifications should also be mentioned in order to increase awareness. Alternatively, there could be an oath for engineers, programmers and economists. He stresses the need for broad awareness in society, among school teachers and children in middle school, elementary school, high school etc.
Next, Gopinath asks about the possibility of a technology denial regime like was done in the case of nuclear energy or the possibility of using quantum computing for adversarial learning.
Professor Vidyasagar says that since AI tools are open source, they would be available for us to use, which is actually better than the government creating a framework and thereby turning us into a surveillance state.
Rajiv comments that IBM Watson isn’t open source and we don’t also have a Jio, TCS or Infosys equivalent. However, the sheer scale of the thing, with thousands of man years of work that went into it, as well as its testing and perfection in different industries like medical, banking, etc made it hard to replicate.
Elon Musk-backed GPT 3 harnesses natural language processing to understand what’s being said in messages, emails and videos using natural speech. It can also be used to compose or translate things. So it’s a very useful tool in artificial intelligence. GPT 3 was free until recently, when it got exclusively licensed to Microsoft. So technically, it’s available, but Microsoft will make money from it that we’d have to pay in order to use it.
Another aspect is big data that social media companies like Facebook, Twitter and Google have. Companies like Facebook and Google get to decide which posts will be more popular/viral, or which people will be shadow banned or openly banned. When you are wronged (as Rajiv has been many times), you have to take your case to these companies. Then they become adjudicators, like the British East India Company was with the Rajas. Rajiv’s book refers to this role as Google Devata, Facebook Devata, Twitter Devata etc. The power of these companies comes from big data because so many millions of people are using the platforms and have invested in building their followings, so it’s very difficult for them to move away and risk losing that.
These companies can’t be taken to an Indian court because they’ll keep escalating the matter until it reaches a U.S. court. In the U.S., they blame it on the algorithm (algorithm devata) and you can’t argue with it. You really have no recourse except to learn the agamas that make these devatas happy (like the agamas governing your deities and the resultant rituals).
So while in theory, AI tools are open source (and positive media coverage has led to lot of people buying into this theory), in reality it is not. And this kind of thing will not lead to technology denial but technology imperialism. These tech companies (whether controlling vertical market experience, or natural language processing or data and algorithms, as seen above) could let others use the technology for a price, but they remain in the driver’s seat. This is a producer-consumer relationship, which is not one of equality.
Facial recognition is another example. These companies have a head start of hundreds of millions of faces that they have already used to train their systems. So anybody building their algorithm and training it today has to make up a huge amount of lost ground.
In this asymmetry of power, we are the ones at a disadvantage with the foreigners being at an advantage. So under make-in-India, we should be making our own platforms, building our own big data, Watsons, GPT 3s etc.
Professor P R Kumar responds.
He says that there are certain things that are possible for India to do today and others that are difficult. For instance, India doesn’t have a good semiconductor wafer fabrication plant, which he calls the steel mill of the future. Because of the technologies involved, this is very hard for India to do. In that sense, machine learning, facial recognition technology, building a good database, etc. India will be able to do.
Shailesh Kumar says that technology always creates asymmetries because the one who has the technology is at an advantage. So just as in the past, the ones that discovered the routes became the colonizers. Today, digital colonization is happening because we have not tried to build our own technology companies at any level. For example, the data centers that these digital giants are being forced to build in India now (because of government mandate) are 2-3 generations behind the latest available technology. And there’s not much we can do about it.
In the first layer, there are the chips (we haven’t produced any Intels or AMDs), next layer the computers (we haven’t produced any Dells), then the software (we haven’t produced a Microsoft), then data (we haven’t produced anything like the FANG companies (Facebook, Apple, Netflix, Google). And the last one is AI. This is an entire value chain that is currently entirely coming from foreign companies. So if we start at the top we can be cut off by any of the suppliers. So just like food sovereignty, we need to have end to end sovereignty in technology.
He explains why open source is a façade. He says that a company open sources something when it knows that you can’t do much with the technology, you can’t become their competitor. For example, Google open sourced TensorFlow but it isn’t useful without all the data Google has. You can download the free version but can’t go far with it. Similarly, Twitter at one point allowed people to download their data and do analytics with it. But they stopped as soon as they realized the value of the data.
Rajiv says that even in the book, he has questioned why companies like Jio weren’t doing more. But it’s clear from Shailesh’s comments that the challenges are huge and this will take time. We also need to add quantum computing to the chips-computing-software-data-AI list because China is working hard at it and they’ve had a few breakthroughs. If they are able to crack it, they would be able to hack into all the networks, with encryption or without. And military supremacy depends on it. India has begun to understand this and recently hired somebody from Cambridge to start quantum computing work in India. So to bring India on par with the U.S. and China will take years of work and billions of dollars in investment.
We need something like Bhabha Atomic Research in AI, on the same scale and with huge funding, and collaborating with industry and academics, the way they have the military-industrial-academic complex in the U.S. that China has adopted.
Professor P R Kumar clarifies that whereas a group of people here could build a computer operating system for handhelds way back, before Palm (that didn’t take off for various reasons) it isn’t the same thing as building a semiconductor fab, which involves a $5 billion investment to begin with. There are also other challenges. For example, there’s a lot of confusion about how India should do 5G. Until recently, it was not even part of the 3GPP, the organization developing 5G standards. If you’re part of it, you get veto rights. Then recently, we formed a standards organization here called Telecom Standard Development Society of India (TSDSI). But because they invited all companies registered in India to be members, a lot of foreign companies also joined, making it difficult to get started with the whole thing. So there are many challenges, but the most worrying to his mind is that we still don’t have a wafer fab of any significance. In this respect, he suggests starting with gallium nitride because it could enable us to leapfrog the silicon based chip designs. Another area that he thinks we will be able to tackle is lighting sources (used in semiconductor manufacturing) because it requires less funding.
Gopinath has a question. He says that India doesn’t have many platforms except at the government level where we have India Stack; GST; jandhan, aadhar and mobile (JAM); and Open Credit Enablement Network (OCEN). So he wonders whether these are usable and what is the future for platforms in India. And he wonders whether international powers and the big technology companies could change international laws to their advantage and to India’s detriment.
Shailesh Kumar says that we need to become platform thinkers rather than tactical thinkers. He praises the government’s digital payments platform, Adhar and national education policy, calling them amazing and far-reaching but thinks it unfortunate that these platforms have come from the government. We need many more such platforms if we’re to digitize India on every level.
Platform development takes time, research and resources to build and an ecosystem that can grow it. It also involves the coming together of political, industrial, technological and academic will. A startup can’t do this.
He also says that while India will undoubtedly benefit from AI, AI will benefit hugely from India because it is so rich in languages, genomics, consumer behavior, etc as well as in problem statements. We could do GPT 3 or a speech to text understanding system in Indian languages, or translation systems across Indian languages, etc. The AI opportunity in India is huge. The largest diversity of data is in India and the biggest range of problems (that the AI algorithm will solve) is also there in India. The next generation has to understand this potential.
Rajiv talks about funding for AI startups, which is coming largely from foreign VCs including Chinese. NASCOM, which represents software people has more or less conceded that Indians will supply labor to build foreign IP.
Professor P R Kumar says that there’s a lack of broad awareness of these issues in India. He gives some examples: In 1987, C-DOT developed a telephone exchange for rural India but because of some political/economic pressures it folded, and the reasons are shrouded in mystery. Second, we developed 4G technology in India, which could have been tested out as a part of the Bharat net program that’s connecting fiber to all Indian villages. But nobody is doing it. Third, there are a huge number of IIT graduates who dream about becoming a clerk at a foreign bank. Without broad awareness of issues, there won’t be any policy change in the country.
Rajiv talks about the lack of awareness among social scientists, psychologists, economists, gurus, etc. He talks about the lack of discussion at lit fests.