Telegram will integrate a large language model like ChatGPT.
Telegram is quietly dominating the world of messaging. They will soon integrate LLM and this is why:
While all eyes are on bing-chatgpt-gate or Google, I think Telegram will be the one to show the first real success integrating large language models at scale.
Telegram has been learning conversational interfaces user experience since 2013, through data of 500+ million users, steadily growing Y2Y. They have been creating features, copied by the whole industry, gathering trust despite accusations of being a Russian spy, drug dealer's nest, or simply an app used by terrorists. Who can afford that today?
Let’s see why I think they will be a strong competitor for Google and Meta.
Reason 1:
Telegram is an innovator in messaging space. They were the first to integrate reactions, the first to let you make interactive stickers, edit messages, send auto-destructive messages, etc. They don’t have better engineers than Meta, but they probably have better data and data-driven product teams, who understand users’ pain points faster.
Reason 2:
They have large groups used by media and brands (some as big as 7 million members vs 1024 in Whatsapp until November 2022 when WhatsApp finally launched communities with a lag of 2 years on Telegram).
Addressing large audiences, Telegram has already access to billions of contextualized conversations, links, videos, images, and articles with comments from real humans (Putin’s bots excluded) on each post, and emoji reactions (discreet classifications).
Gold for LLM training sets.
Reason 3:
They are quietly eating SaaS verticals: group video calls (Zoom), large file storage and transfer (Wetransfer/Box), embedded e-commerce (Shopify), and more to come. Unlike Twitter, Telegram may soon become The Western WeChat.
Reason 4:
Telegram is developer-friendly like Stripe was and has created a real ecosystem. The Telegram API is letting anyone create their own version of Telegram. It means endless possibilities for building business applications using conversational interfaces while relying on the Telegram messaging backbone.
Look at the following 127.000+ VS 57.000+ Telegram VS WhatsApp GitHub repositories:
Conclusion
Telegram is already active in the AI space, rolling out the Audio-to-text feature. They could easily continue with summaries and auto-replies.
Then let anyone map their own business API (aka Booking.com, Amadeus, OpenTable…) to Telegram’s conversational API (custom LLM model trained behind) and dominate the world (if you could book your plane tickets, Airbnb, or book a table in a single app, it would be the forgotten promise of Magic). They will do in the B2C space what Teams is doing in the B2B one.
