ChatGPT most-popular AI chatbot in India, but it can get slippery at the top
Summary
- Sam Altman is back as CEO and has committed to continue building beneficial artificial general intelligence. But can OpenAI hold on to its first-mover advantage given the increasing competition and rapidly evolving technologies?
ChatGPT turned one on 30 November, remaining the most dominant AI-powered chatbot globally, particularly in India, even as the dust settles on the corporate turmoil at OpenAI. But the battle’s far from over for CEO Sam Altman and his team, with competition heating up not just from other tech companies, big and small, but also from rival technologies.
More on that later. First, a look into ChatGPT’s dominance and how most folks are engaging with it.
In India, as in the US and Brazil, the large language model-based chatbot is mostly used on mobile devices, per new reports by data.ai (formerly App Annie) and writerbuddy.ai.
In the AI tools sector, ChatGPT registered about 14.6 billion visits, averaging 1.2 billion visits every month, writerbuddy.ai, an AI-based writing platform, said in a 15 November study.
Of the top 50 AI platforms that writebuddy.ai analysed, including Google Bard, ChatGPT cornered a 60% share of the traffic.
Mobile devices are the preferred means of access for 58% of ChatGPT’s users, according to writerbuddy.ai, and about 74.16% of them are male, making for a significant gender disparity.
ChatGPT’s competitors including Perplexity AI and Character AI account for 18.69% of the chatbot market’s monthly visits, the report says. Despite the competition, ChatGPT dominates the AI tools sector with average usage duration of 30 minutes.
ChatGPT is the most-popular AI tool in the US, followed by India and Brazil, according to the writerbuddy.ai study.
But as per analytics platform data.ai, India accounted for the most ChatGPT app installations globally (18%), just ahead of the US (17.5%). Data.ai in its 1 December report added that ChatGPT’s mobile version had achieved 110 million downloads, and generated about $28.6 million in global consumer spending. Within one week of being launched on Google Play, the app hit a record weekly adoption of about 18 million new installations.
ChatGPT downloads on both Apple’s iOS and Google’s Android platforms combined were at above 4 million in the five weeks to 1 December, according to data.ai. The platform, it added, ranks first among generative AI apps in terms of total downloads, outpacing competitors such as Character AI and Ask AI.
But in terms of consumer spending, ChatGPT ranks second, outmatched by Ask AI from Codeway.
ChatGPT Plus is the sole in-app purchase for the ChatGPT app, and with its recurring subscription billed at $19.99 per month, it is “greater than many music and streaming services", according to data.ai.
But Achilles’ heel shows up
OpenAI’s one-year journey has been anything but smooth. While ChatGPT took the online world by storm garnering more than 100 million users in its first two months, OpenAI’s board shocked the world when it unceremoniously sacked CEO Sam Altman last month, only for him to be reinstated within a fortnight.
As the drama unfolded, OpenAI’s estimated $86-billion valuation came under a cloud. Microsoft chairman and CEO, Satya Nadella, worked at a frenetic pace to protect his company’s more than $11-billion investment in OpenAI. He almost got Altman to lead a new AI unit at Microsoft. The tech giant is now on the board of OpenAI as a “non-voting observer", which will give it visibility of the board’s decisions.
Nadella reportedly was informed just a minute before the world knew about Altman’s ouster, despite Microsoft having a 49% stake in the for-profit entity that the OpenAI nonprofit board controls.
Altman’s sacking on 17 November was apparently over concerns about the safety and speed of OpenAI’s tech deployment, which ostensibly unsettled some board members including Ilya Sutskever, a co-founder and OpenAI’s chief scientist. Some fear that ChatGPT could herald the onset of artificial general intelligence (AGI)–one of the reasons for Sutsekever’s differences with a techno-optimist like Altman.
The new initial board will comprise Bret Taylor (Chair), Larry Summers, and Adam D’Angelo. Sutsekever is no longer a board member but Altman said in a note that “... we hope to continue our working relationship and are discussing how he can continue his work at OpenAI".
Three priorities
According to Altman, OpenAI will now focus on three priorities:
advancing its research and investing more in “full-stack safety efforts, which have always been critical to our work";
improving and deploying more products; and
improving the company’s governance structure.
OpenAI was founded in 2015 by stalwarts including Altman, Sutskever, Greg Brockman, and Elon Musk, who served as a board member till 2018, when he quit over differences with Altman and Brockman. In 2019, OpenAI restructured to ensure that the company could raise capital, following which Microsoft invested $1 billion in 2019, following up with an additional $10 billion this year.
Altman said in his note after being reinstated as CEO that he was “looking forward to finishing the job of building beneficial AGI".
At the OpenAI Devday in San Francisco last month, the company announced major product releases–including a provision for anyone to build their own AI-based chatbot; GPT-4 Turbo with a 128K context window; GPT-4 with Vision; and Dall-E 3 API.
Meanwhile, though, ChatGPT has competition to reckon with from multiple open and closed source models, including Microsoft Bing, Google Bard, Anthropic Claude, Inflection Pi, and Perplexity.ai. Developers, too, have multiple options, including application programming interfaces (APIs) from Amazon Web Services (AWS), Azure, Cohere, Google Cloud, Hugging Face, and OpenAI.
Also, small language models too are evolving, implying that the future of AI will not be just about size but about smart, efficient choices (ChatGPT is a large language model-based chatbot). Training smaller models consumes less energy and is less expensive. SLMs can also perform specialised, domain-specific tasks relatively well after fine-tuning on only a handful of examples. Examples of popular SLMs include Llama 2 13B, Mistral 7 B, Falcon-7B, IBM Granite 13 B, and Microsoft Phi-2 2.7B.
Simply put, OpenAI has its work cut out for it. A year of astounding success will not guarantee a smooth path ahead.