Once I opened the Airtel app on my smartphone over the weekend to recharge my cellular plan, I used to be stunned to see a Perplexity Professional AI subscription value Rs 17,000 included with my plan. I double-checked whether or not I had by accident subscribed to Perplexity AI.
Earlier than I began to panic about being subscribed to a different service, I realised that the Perplexity AI subscription was free with my plan, and that the supply is legitimate till July 2026. However Perplexity isn’t the one firm providing its AI subscription plan without cost to hundreds of thousands in India. Reliance Jio, one other telecom operator, has an lively partnership with Google to offer choose customers a free Google AI Professional subscription for 18 months.
In the meantime, OpenAI is providing the annual ‘low-cost’ ChatGPT Go subscription, value Rs 399 per thirty days, free for one 12 months within the nation.
It’s not that AI corporations are averse to providing free companies to customers who purchase their merchandise.
Apple, for instance, is thought for providing free entry to Apple TV for 3 months with new merchandise just like the iPhone and iPad. Even my new Pixel 10 Professional Fold and 10 Professional XL include a free Google AI Professional plan, which incorporates premium AI options and 2TB of storage for a 12 months. In each instances, you pay a premium for his or her {hardware} merchandise, and these tech corporations often supply entry to new companies to extend reputation and lock customers into their ecosystems.
Nonetheless, Google, OpenAI, and Perplexity are providing long-term, free entry to premium AI companies to hundreds of thousands of customers in India. It makes one marvel why these tech corporations are so beneficiant at a time when synthetic intelligence is an costly playground, and the prices of infrastructure investments and coaching AI fashions run into the billions of {dollars}.
Properly, these tech corporations might not be exhibiting a candy gesture however moderately a calculated motive in providing premium AI instruments to hundreds of thousands of Indians without cost, given the massive long-term rewards.
Story continues beneath this advert
House to hundreds of thousands of smartphone and web customers
India is the world’s high shopper of cellular knowledge per consumer, and its web customers are set to surpass 900 million, creating huge market potential. The increase is essentially due to the provision of low-cost web, the penetration of smartphones even in rural areas, a digitally savvy youth aged 18 to 35, and the expansion of digital infrastructure and companies akin to digital funds. For this reason many corporations are pumping cash into India’s fast-growing web ecosystem, a prize many buyers discover too huge to disregard.
The function of telecom operators like Airtel and Jio is equally necessary in setting the stage for India’s AI increase. Collectively, Airtel and Jio have an enormous consumer base, and so they play a decisive function in guaranteeing efficient execution, from deployment to demand conversion, finally driving the expansion of consumer adoption for any new service.
….however the place do AI coaching knowledge come from?
No marvel corporations like Google and OpenAI see a chance to coach their most cutting-edge, superior algorithms and fashions in India. The factor is, each AI mannequin, from GPT to Gemini, wants huge quantities of human-labelled knowledge, and a rustic like India could also be simply good for changing into a spine of coaching knowledge. However the huge query is, the place does AI coaching knowledge come from?
To construct massive generative AI fashions, tech corporations usually flip to the public-facing web, however there isn’t any single place to obtain your entire internet. As an alternative, tech corporations choose their coaching units utilizing automated instruments that catalogue and extract knowledge from the web. In spite of everything, high-quality knowledge, primarily scraped from the online, is essential to the efficiency of AI fashions.
Story continues beneath this advert
These instruments embrace internet “crawlers,” nicknamed “spiders,” that are automated applications that systematically browse the World Large Net to index its pages. For instance, Alphabet, Google’s guardian firm, has already constructed internet crawlers to energy its search engine and may use its personal instruments to reap knowledge and prepare its AI fashions. Different corporations, nonetheless, depend on sources like Widespread Crawl, a significant supply of coaching knowledge that helps OpenAI’s GPT, which memorises large quantities of textual content to reply to consumer questions when given a immediate.
Along with publicly obtainable knowledge, AI corporations additionally use their very own knowledge for mannequin coaching. OpenAI, for instance, fine-tunes its fashions based mostly on consumer interactions with its chatbots. Meta AI is partially skilled on public Fb and Instagram posts. Amazon, too, says it makes use of some voice knowledge from clients’ Alexa conversations to coach its LLM. Nonetheless, normally, AI corporations have been secretive concerning the datasets used for coaching.
Now, comes the exhausting half
A scarcity of transparency about coaching knowledge is the largest crimson flag for why AI corporations are below scrutiny. The New York Occasions lately filed a lawsuit in opposition to Perplexity, alleging that the startup has illegally copied and distributed its copyrighted content material.
The swimsuit, filed within the Southern District of New York final week, accused Perplexity of unlawfully scraping The Occasions’ tales, movies, podcasts, and different content material to formulate responses to consumer queries. One other publication, The Chicago Tribune, filed an identical copyright lawsuit in opposition to Perplexity. The Tribune additionally argues that Perplexity scraped and distributed its content material with out authorisation.
Story continues beneath this advert
Perplexity, based by Indian-origin Aravind Srinivas, has been the topic of a number of lawsuits. Earlier this 12 months, Cloudflare, a number one digital infrastructure firm, accused Perplexity of hiding its web-crawling actions and scraping web sites with out permission. Perplexity denied the allegations.
In October, social media firm Reddit additionally sued Perplexity in New York federal court docket, accusing it and three different corporations of unlawfully scraping its knowledge to coach Perplexity’s AI-based search engine.
Information websites and quite a few publications have accused AI corporations of utilizing copyrighted content material with out authorisation to construct and function their AI methods. In 2023, The New York Occasions blocked OpenAI’s internet crawler, GPTBot, from utilizing its content material to coach AI fashions. Quickly, AI corporations realised they wanted to strike offers with publications, as knowledge can’t be correctly used with out permission.
This led OpenAI to start signing agreements with main worldwide media corporations to make use of their copyrighted content material as coaching knowledge. Axel Springer, France’s Le Monde, and Spain’s Prisa Media inked offers with the ChatGPT makers to offer materials for coaching its AI fashions, adopted by the Monetary Occasions, which additionally lower a deal permitting ChatGPT customers to obtain summaries, quotes, and hyperlinks to FT articles.
Story continues beneath this advert
Later, Reuters and the Related Press signed offers with OpenAI, as did Hearst, The Guardian, Condé Nast, Vox, TIME, and The Atlantic. Microsoft signed a cope with USA At this time. In the meantime, Perplexity gained entry to the work of AdWeek, Fortune, Stern, The Impartial, and the Los Angeles Occasions. Axios, a number one know-how publication, additionally signed a licensing cope with OpenAI.
Nonetheless, publishers haven’t any subject with search engines like google like Google utilizing internet crawlers to entry their web sites. That manner, the search corporations can, in return, get direct visitors to their content material.
That mentioned, the bitterness between content material creators, publications, musicians, artists, and AI corporations persists, with stakeholders turning to the courts to forestall what they see as AI companies infringing on inventive rights. Take Disney and Common, for instance, who lately sued the bogus intelligence agency Midjourney over its picture generator. The 2 Hollywood studios allege that it’s a “bottomless pit of plagiarism”.
They declare Midjourney’s device makes “innumerable” copies of characters, together with Darth Vader from Star Wars, Elsa from Frozen, and the Minions from Despicable Me. Transparency about knowledge sources needs to be given precedence, in spite of everything. Even when artists and musicians ink a cope with AI corporations, there may be all the time the query of the extent to which AI makes an attempt to recreate their model with just some keystrokes. There’s nonetheless no concrete reply.
Story continues beneath this advert
Can Indians shield their knowledge from AI?
Gifting away AI subscriptions value hundreds of rupees without cost just isn’t a brand new tactic. Google and others have proven that this technique has labored up to now and will work once more if entry to AI companies is supplied freed from cost. Actually, corporations like Google gained entry to lots of their clients by providing companies without cost. Take its Google search engine, for instance, which is actually free however shows adverts on the outcomes web page and collects consumer knowledge – the supply of most of its income.
However there may be all the time a “catch”: a price to free on-line companies, and we as customers finally pay the worth. A startup like Perplexity solely wants one factor: your consideration. Its objective is to construct a considerable consumer base, and if we’re efficiently lured in, it will possibly safe funding and develop even larger.
Perplexity’s valuation has elevated to $20 billion in simply three years. The identical could be mentioned of OpenAI, which has amassed 800 million weekly ChatGPT customers, boosting its valuation to $500 billion. That’s how capitalism works.
All main AI corporations are eyeing India, and for good purpose. India not solely has a big shopper base however can also be a hub for the outsourcing IT trade. On one hand, world AI corporations are gaining inroads into extremely diversified shopper markets the place customers communicate a number of languages and every area has its personal distinctive tradition and dialect, particularly in rural cities.
Story continues beneath this advert
On the similar time, they’re having access to a big pool of customers from startups and SMBs. For giant tech corporations, the extra customers who have interaction with their AI companies, whether or not a scholar, a company skilled, or a warehouse employee managing methods, the higher it’s for coaching their AI fashions. And no market is best than India in the meanwhile.
If the AI ecosystem develops, it may additionally create a marketplace for cloud farming in smaller centres throughout India, which can be utilized for constructing datasets to coach AI and for content material moderation.
One elementary query that can’t be ignored is whether or not delicate private knowledge could be refrained from AI coaching. Presently, India doesn’t have a regulation particularly governing synthetic intelligence. Whereas the Digital Private Knowledge Safety Act (DPDP) 2023 gives broad protections for private knowledge, it has but to be enacted. Furthermore, the Act doesn’t tackle AI methods or algorithmic accountability.
In states like California within the US, digital privateness legal guidelines do give customers the suitable to request that corporations delete their private knowledge. Within the European Union, the Synthetic Intelligence Act imposes controls on “high-risk methods” utilized in areas akin to training, healthcare, regulation enforcement, and elections. It bans some AI use altogether.
Story continues beneath this advert
Nonetheless, there may be presently no clear technique to make an AI “neglect” beforehand discovered knowledge; utterly eradicating copyrighted or delicate data would require retraining the mannequin from scratch, which may value tens of hundreds of thousands of {dollars}.