If this is something their users actually wanted, they'd be paying xAI, not the other way around.
This is a data sale, pure and simple. xAI certainly isn’t making their money off their LLM/inference. OpenRouter shows Grok at about a billion tokens a day while the top 20 account for 2.5 trillion per day.
I’ve suspected Elon expanded his failing strategy to include data brokering when he saw the opportunity to get access to everyone’s data via Doge. Hence the reason Elon is ready to step away now that Doge (many of which are xAI employees) has finished gaining access to the data from every government system. Quietly offering access to corporate and political clients to query the data of every single person via Grok seems like an easy way to generate some revenue when no one wants your AI.
Grok is indeed trailing the race and could use extra numbers. It doesn't matter to their bottom line (or their ad campaigns) if these numbers come from a flat rate agreement with Telegram.
If anything, Telegram should have bargained for more.
Admittedly, some of the uses in the article do sound useful. Time will tell
But this is why we must champion open protocols and open source software.
I absolutely believe both LLMs and social media platforms should be federated and controlled by the user.
So, the question is whether they should pay to generate demand in a new market. Then, who to pay and how much?
I'll also note that OpenAI took the market by offering an expensive service for free (ChatGPT). Then, they offered a monthly plan that may or may not have been profitable. One could argue that OpenAI has been paying people to use its service for a long time.
I also wonder if xAI gets something out of it. For instance, they might get all the conversations with the AI. I believe ChatGPT similarly put free, user conversations to use internally in ways that boosted their paid apps. xAI might have some plays like that.
https://www.paypal.com/uk/webapps/mpp/invite/terms
It is a common way for businesses to acquire customers.
And to collect firefox user data. That't why firefox is always connected to e100.net.
Users wouldn't specifically ask for Grok, but they might like to have access to an AI assistant. When you are competing against OpenAI etc. it makes sense to incentivise big, PR generating customers such as Telegram.
For xAI to pay Telegram... where does xAI get value back out of that? I guess we'll be seeing AI with ads soon?
It not exactly unrealistic to imagine that people will ask LLMs about flights, hotels, restaurants or even insurance.
All that needs to happen is for LLMs to add a “buy now” button, and for the provider to take a 15% commission (still a lower take rate than Expedia FWIW).
The "dream" is that the LLM/AI does all the work for me and just magically gets me to "this is the perfect flight for you and here are the reasons" but I have to tell it the things I care about (price, time, etc.). A lot of the time it's not exactly clear, even to me, which of those parameters matter the most so as a consumer it's actually nice to see, for instance, the options with times/costs/stops/etc. and for me to be able to look those over and make a decision. The LLM could provide those options, but then what has it done that I wasn't able to do with Google Flights or another ota? Is it just using more natural language in the request? Or do most people really want to wholesale handoff the decision and just go with the "trustworthy" LLM/assistant without any of the rationale?
I suppose actual personal assistants do that type of thing all the time for wealthy people, but that doesn't seem like it would be applicable to the masses who want/need to comparison shop for the best deal that meets a bunch of criteria.
I say this as someone who gets major value out of LLMs already, but for buying things in particular I'm struggling to understand why you'd want to hand off the "browsing" and just fast forward to the "buying."
The first are power users, in the sense of people who have complex requests. Like those who’d ask for the cheapest flights from Manchester to Turkey during the summer school holidays for without layovers unless those layovers were in Paris, for a holiday to be roughly 10 days long, excluding very early morning flights. Such a request could be made with existing OTAs, but would be painfully time consuming.
The other type of user for whom LLMs might be useful are the opposite, those with very loose requirements. Think “get me a flight to a warm place tomorrow”.
Skyscanner etc would still have value for the people in the middle.
https://google.github.io/adk-docs/tools/built-in-tools/#goog...
> When you use grounding with Google Search, and you receive Search suggestions in your response, you must display the Search suggestions in production and in your applications. For more information on grounding with Google Search, see Grounding with Google Search documentation for Google AI Studio or Vertex AI. The UI code (HTML) is returned in the Gemini response as renderedContent, and you will need to show the HTML in your app, in accordance with the policy.
There's no reason for Grok to pay this much for the deal unless either Telegram sees it as a net negative that needs reimbursement or they're competing with other bids.