Meta Platforms is building a search engine, planning to leave Google and Microsoft in the dust. Until now, the company’s AI chatbot has depended on Microsoft’s Bing for data on sports scores, news updates, and stock information. Now, they want out, aiming to run everything in-house.
This week, Meta’s latest power move sent Alphabet shares down by 0.8% — no surprise there. Meanwhile, Meta’s stock climbed 0.3%. Looks like Wall Street likes Meta’s bold attitude.
NotebookLlama: Meta’s answer to Google’s Podcast AI
Meta’s also launched NotebookLlama, a direct shot at Google’s NotebookLM. This AI feature takes documents (like PDFs or articles) and spits out podcast-style summaries.
Meta uses its own Llama model here, with NotebookLlama adding “dramatization” and back-and-forth banter to make the content sound more like a real podcast. Problem is, it still sounds like a robot. Voices cut each other off awkwardly, and it’s a far cry from any human conversation.
Meta knows the system’s janky. Their team even admitted that the text-to-speech model makes it sound unnatural, saying, “The text-to-speech model is the limitation of how natural this will sound.” They’re considering adding two AI agents to debate the topic instead of the single model they currently use.
But let’s be real, no AI podcast—including Google’s—has nailed it yet. AI-generated podcasts still make stuff up (hallucinate, as they like to say) and dump random “facts” that don’t exist.
Meta’s makes AI deal with Reuters
Last Friday, Meta signed a deal with Reuters to supply its AI chatbot with trusted news content. This move comes after years of Meta cutting back on news, especially after the constant backlash over misinformation and profit-sharing disputes.
zThe deal is simple: they get real-time news; Reuters gets paid. Details of the financials? Not disclosed. But it’s likely a multi-year deal, according to Axios. Through this setup, Meta’s AI will summarize and link to Reuters articles when users ask news-related questions.
Meta’s still vague on whether Reuters content will train its language models. Reuters, meanwhile, confirmed that they partnered up to provide “fact-based” content to power the AI systems, without sharing specifics.
Other AI players, like OpenAI and Perplexity (backed by Jeff Bezos), have similar deals. Meta’s partnership with Reuters isn’t exactly new; they’ve been working together since 2020 on fact-checking.
Meta’s escalating AI spending and revenue growth
Meta is not skimping on AI. Wall Street estimates that Meta’s total spending this year alone will run up to $40 billion, with a lot of that going toward AI.
Analysts forecast Meta’s Q3 earnings per share to hit $5.21, translating to $13.49 billion in profit—up 16% from last year. Revenue is also expected to surge 18% to $40.19 billion.
Meta’s so-called “hands-off” approach on politics translates to algorithmic suppression of anything remotely political on its social media platforms, especially around election cycles. Try mentioning “vote,” and your post might vanish.
Internal inconsistencies are hard to ignore too. A report by The Intercept revealed that Meta’s Israel policy chief, Jordana Cutler, targeted pro-Palestine groups on Instagram, labeling them as “blacklisted” without basis. Meta called the report “irresponsible,” but it’s not new.
Documents from the Facebook Papers show Meta’s repeated compliance with government demands to censor posts, especially in India, where it bent to pressure from Modi’s administration.
At the same time, misinformation has a free pass, especially during crises. AI-generated conspiracy posts flood Meta’s platforms after every major event—from hurricanes to health crises.
Zuck even admitted to Congress he regretted taking down some COVID-19 misinformation, as “Plandemic” fans know too well. Despite a growing demand for accountability, Meta’s solution is fewer moderators, fewer policies, and even more spending on AI.
