OpenAI and the Network Effect (ft. Md Rafi and Ola Krutrim)

"Who is the greatest Bollywood singer of all times?" I typed into chat.krutrim.com

It listed seven, but missed Mohammad Rafi. 

Horrified, I followed up, "Why is Mohammad Rafi not in this list?" 

And it missed the context, replying, "Mohammad Rafi is not in the list because the list you are referring to is not provided."

With a deep sigh, it reminded me of Altman's India visit June last year. Someone asked him if India should invest in building a Foundational model (assuming funding and talent is not as issue). And he replied, "it would be hopeless to compete with us on training foundation models.. you shouldn’t try”.

Try they will, and they should. The world's fourth(?) largest economy has pockets of deep pockets that can sustain the demands of developing a resource hungry technology such at Foundational LLMs. But distribution, diffusion and monetisation remains challenging, when chatGPT, Copilot and Gemini in Indic languages are just an App away on the same device. 

Network Effect is the game of volume and velocity. Targeting a niche market and offering a unique value proposition for that market helps gain the critical mass of adoption. Incentivising the users for adoption has paid off for building the user base in a shorter time. but gaining customers and expanding the user-base is as critical as retaining users. Building differentiation and continuous innovation are mandated for maintaining the cutting edge and enables a quick pivot in response to market trends. Vertical integration with interoperability and strategic partnership are the necessities for expanding into untapped markets.

All of this is easier said than done. Google, the AI giant of the planet, and the original innovator and patent holder for Transformers (the technology that underpins GenAI), is struggling against an upstart OpenAI. Google Bard has already been axed by the new marketing strategy of Google Gemini, which had its own share of bad PR after inception. Apparently, Microsoft has build a more effective Network Effect strategy by integrating and building interoperability with OpenAI: It is able to drive mass user-adoption through its browser and app ecosystem, and incentivising them by providing free (though limited) access to best GenAI technology; is able to take the Android-advantage and first-mover-advantage away from Google; is able to keep Apple and Amazon at bay; and is able to derive the competitive advantage for winning in the GetAI market. 

For how long this competitive advantage will last is hard to tell at the moment. The pace and quantum of the change in the industry is very high. It suffices to say, however, that Microsoft and OpenAI have a clear head-start. 

But sometimes, a head-start is all that you need. The underlying phenomena here is that Network Effect will ensure that chatGPT will continue to become bigger, at the cost of others, with every incremental user adding to the value of the product and the LLM learning from every interaction. (Not to be confused with Brand loyalty, which is sans any direct feedback loop with the user except for an emotional connect with the product. Google is a much bigger brand and OpenAI is not even a fraction of it, but that hasn't helped Google gain in terms of Network effect.)   

With barriers of entry so high in terms of resource intensiveness and costs, competition and rivalry so intense, substitution costs being so low, and an absence of curated, high quality, India-specific, dataset, reduces the overall industry attractiveness for a firm that wants to invest in building India-specific Foundational LLM from ground up, and make it a thriving, profitable, endeavor.

What does that mean for BharatGPT? A list of 60+ LLMs is here. Foundational LLMs are only a handful, including Krutrim, Bhashini, OpenHathi, Project Indus. For them, the challenges are enormous. For the rest of the crowd that is largely startup-led, their models are based on 'american' open-source Foundational LLMs. They have a much smaller parameter base, and are targeted to solving defined use-cases. Some of them may survive the test of time, and a round of consolidation in a maturing industry would see two or three mid-size players appearing down the lane, by the end of the decade.

Other, smaller, countries and societies are likely to break the ice first - especially in Europe, where local language driven use-cases would start coming into the market. Likewise, there are plenty of opportunities of carving out a niche in Indian context. But whoudl all of them require its own foundational models is not clear.

• What do you think about the future of such LLMs in India? 
• Are there any specialised or nuanced use-cases for India that you know of? 
• What do you think of the overall GenAI "bubble", and its capability of writing code? 
• What impact would it have on IT/ITeS/BPO industry that is USD100Bn+ today? 

For Ola, this Krutrim tastes rather "unnatural" and raw, at least for now.

"Krutrim" is an India-centricFoundational LLM,
that seems to fall short on multiple accounts
including - bias in dataset, and Context limitations.