Many entrepreneurs are focusing on creating multilingual ‘platform’ apps to cater to Bharat, but theirs is not a business model that will last
By Siddharth Pai
I am in New York City, meeting with founders of firms from the ‘Silicon Alley’. The range of firms working on hyper-focused B2C technology niches is astounding. Today I met with a firm that is trying to create a market in specialised chefs and wait-staff at city restaurants that are beginning to open their doors again.
New York, as we all know, is a melting pot. Just a short walk takes one past restaurants that serve cuisine from Peru, Brazil, India, Mexico, Greece, France, Italy, Thailand, China, Malaysia … you get the drift. The pandemic put many of these establishments out of business. As the US begins to return to normal, many of the specialised staff who once worked at these restaurants are still free agents. They are now open to ‘casual employment’, much as drivers for Uber, Lyft, Zomato and Swiggy are.
Such hyper-specialised niches will only grow. ‘Marketplaces’ for every imaginable sort of skill have been enabled by two factors: One, a modern-age breakdown in Nobel laureate Ronald Coase’s economic reasoning in 1937’s ‘The Nature of the Firm’.
Two, technology has drastically reduced the transaction costs for individual buyers and sellers, and allows them to directly work with one another. Earlier, both sides had no alternative but to go through a monolithic firm of the type that Coase described.
Start-ups in India don’t seem to have yet caught on to the hyper-specialisation of the marketplace. Many Indian tech entrepreneurs have instead focused on ‘Bharat’ a catch-all term for the hundreds of millions of newly-online Indians. Many entrepreneurs are focusing on creating multilingual ‘platform’ apps that they hope can be used ‘as a service’ by businesses catering to this vast, underserved market.
While this is laudable, and while some such firms have even received significant funding, in my opinion theirs is not a business model that will last. Apart from the fact that plug-ins are already available to behemoths such as Google’s ‘Translate’ application, the very nature of how these programs are taught by using machine learning has also been stood on its head.
During the middle of the pandemic last year, an artificial intelligence lab in San Francisco called OpenAI revealed a technology that has been in the making for some time. This new system, called Generative Pre-trained Transformer 3 (GPT-3), learnt the nuances of natural language over several months—language as spoken and written by humans. It analysed thousands of digital books, and nearly a trillion words posted on the internet. GPT-3 is the output of several years of work done by the world’s leading AI labs, including OpenAI, which is an independent organisation backed by $1 billion in funding from Microsoft.
Meanwhile, at Google, a system called Bert (short for Bidirectional Encoder Representations from Transformers) was also trained on a large selection of online words. It can guess missing words in millions of sentences, such as “I am going to see a man about a…” or “I am going to… a man about a dog.” These systems can manage many interfaces, from chatbots and voice commands to Amazon’s Alexa or Google. But GPT-3, which learnt language from a far larger set of online text than previous models, opens up many more possibilities.
For most of today’s ‘machine learning’ or ‘deep learning’ programs, including image recognition tools for self-driving vehicles, we think of thousands of people in India or Sri Lanka labelling every picture so that an AI program can refer to those labels each time it attempts a task, such as recognising a traffic sign accurately, or a pedestrian instead of a bicyclist.
Unlike these, GPT-3 and Bert can be primed for specific tasks using just a few examples, as opposed to the thousands of examples and several hours of additional training required by its ‘deep learning’ predecessors. Computer scientists call this ‘few-shot learning’ and believe that GPT-3 is the first real example of what could be a powerful change in the way humankind trains its machines. The real surprise from GPT-3 is that Systems Architects have been able to provide just a few simple instructions to let it even write its own computer programs.
At a basic level, computer programs are English-like commands given to a computer in a logical sequence such that the commands produce a certain outcome after a computer acts on them. As such, GPT-3’s mathematical descriptions of the way we piece English together works whether we are writing books or coding software programs. Using these maps, GPT-3 can perform tasks it was not originally built to do.
So far, OpenAI has shared GPT-3 with a small number of testers, since many kinks, including biases and profanities, need to be sorted out. That said, OpenAI scientists have claimed that “any lay person can take this model and provide these examples in about five minutes and get useful behaviour out of it.” Unsurprisingly, Microsoft has licensed exclusive use of the source code, but others can use a public interface to generate output.
As such new-age methods become more generally available, ‘vernacular platform as a service’ start-ups in India run the risk of becoming quickly obsolete. I would daresay that the smart way around the inevitable would be for such firms to focus their vernacular use cases on hyper-specialised marketplaces such as those that are springing up here in New York.
The author is Technology consultant and venture capitalist
(By invitation from New York)