Sam Altman, the creator of ChatGPT, expressed his belief that India’s attempt to develop a foundational AI model similar to ChatGPT may not be worth pursuing.
CEO of OpenAI and creator of ChatGPT Sam Altman was in India for the last couple of days. His visit was concentrated broadly on the way forward and regulation in the area of artificial intelligence (AI).
Meeting with the students of Indraprastha Institute of Information Technology (IIIT) Delhi on Thursday, June 8, 2023, for a one-on-one session with Prime Minister (PM) Narendra Modi, the 38-year-old techie’s emphasis was on the AI tool.
During his visit to India, Sam Altman’s one statement captured the attention of Indians, coinciding with the country’s efforts to formulate AI regulations under the Digital India Bill. The bill, which has been in progress since the previous year, signifies the timing of Altman’s visit.
Sam Altman claimed that Indians would be “totally hopeless” if they attempted to develop something akin to ChatGPT.
This observation gained widespread attention, particularly considering the challenges that tech enthusiasts often face in the country.
Altman’s remarks about the difficulty of freely expressing opinions without taking responsibility for their accuracy among Indian techies sparked a viral response on social media soon after he made them.
On Thursday, while he was in India as part of his visit to Asia, when Altman made his remarks.
Altman was asked for advice and insight on how one could start to develop things like OpenAI and ChatGPT in India at an event hosted by the Economic Times. Rajan Anandan, a venture capitalist and former CEO of Google India, posed the question to Altman.
However, reports suggest that Altman was implying a lack of infrastructure in India, given that OpenAI has already created something like ChatGPT, it will always have the first-mover advantage leaving behind a huge learning curve for others who follow.
Altman argued that it would be impossible to create a tool like ChatGPT for a user in India since doing so would require the development of a specific type of infrastructure.
More importantly, Altman made it clear in his response that building a foundational AI model for a startup is not only challenging but also nearly impossible given that OpenAI has already created ChatGPT, which is driven by a massive LLM (large language model). Altman avoided discussing the skills Indians possess or lack in his response.
According to reports, Sam Altman’s visit to India also comes at a juncture when the Indian government is looking for conversational AI tools that could support the latter in its administrative tasks.
The government is looking to help the farmers in learning about different government schemes, understand customer grievances, and perform other citizen-centric duties.
Is it really impossible to build a foundational AI model like ChatGPT for India?
Ever since its debut in November 2022, ChatGPT has revolutionized the technological landscape as we know it. Developed by OpenAI, this natural language processing (NLP) chatbot has demonstrated the profound impact that artificial intelligence can have.
From excelling in university-level exams to writing compelling keynote speeches, from helping marketers to assisting programmers in writing and debugging code, ChatGPT has pervaded every industry and domain, leaving no sector unaffected by its transformative capabilities.
At its core, ChatGPT may appear to be a simple chatbot. However, upon tighter examination, it surpasses those boundaries and encompasses a far more extensive scope.
OpenAI has created an advanced artificial intelligence model that leverages the power of natural language processing, specifically built upon the foundation of ‘Generative Pre-trained Transformer 4’ (GPT-4) technology.
This innovative approach allows ChatGPT to go beyond traditional chatbot capabilities and achieve remarkable feats in understanding and generating human-like text.
The development costs of GPT-based applications depend on many factors. These factors include the model’s complexity, the intended use case, the dataset requirements, and the computational resources needed.
For example, when creating an AI app similar to ChatGPT, the cost can be significantly impacted by the volume and accessibility of the dataset. ChatGPT itself was trained on a substantial 570GB textual dataset.
Acquiring a large dataset can be expensive, particularly if there is a need to pay for proprietary data access or hire annotators to label the data.
Additionally, utilizing cloud-based resources can contribute to higher costs depending on the chosen resources and duration of usage.
However, this is not the cost that is the critical factor here, bringing together a team of experts, building the model, and training when you are following a player that has already built a foundational model is not easy.
In addition, when you reach a stage when you build a model similar to GPT4, the ChatGPT would probably move on to GPT8, four generations ahead with much more capabilities and your model may look outdated.
So, is it impossible as Aton calls it — no, but it is challenging.
