Threads has attracted a lot of attention on the internet, possibly due to its complex nature. People have migrated from Twitter to Threads and compared the two platforms' functionalities, but the true reason for launching this text-based social media platform may be to gather data and build another large language model (LLM) called LLaMA or "Super LLaMA".
A month ago, in Lex Fridman's podcast, Meta founder Mark Zuckerberg mentioned that the company is developing another version of LLaMa, namely its open-source LLM. In the same podcast, Zuckerberg also talked about the Twitter alternative the company is developing under the codename P-92. A month later, Threads was released as a microblogging platform clearly inspired by Elon Musk's Twitter. Now, Musk is suing Meta for copying his platform.
Interestingly, Zuckerberg tweeted again after 11 years, using only an emoji.
Zuckerberg also praised Mastodon and Bluesky for their approach to building decentralized platforms for users. He further emphasized how communities can drive platforms like Wikipedia. Interestingly, he also mentioned that the next version of LLaMa may incorporate training from all services provided by Meta, including the text-based social media platform Threads.
Decentralization Triumphs
Twitter CEO Elon Musk has always wanted to create an alternative to OpenAI's ChatGPT. Some speculate that he may leverage Twitter data to make it more consistent. However, there has been no news about this since Musk mentioned recruiting a researcher from DeepMind to develop TruthGPT.
Now, Zuckerberg knows the power behind training chatbots with social network data. The goal of building an alternative to Twitter may have turned into building an alternative to OpenAI. The best thing Meta can do now is to keep the technology open-source, which Zuckerberg and even Meta's AI director Yann LeCun have advocated for.
With Threads, Meta now has the ability to collect and accumulate user data for building generative AI models. Interestingly, the bidding for open-source AI models is still ongoing, and this "Super LLaMa" will be decentralized. This sounds like Zuckerberg is returning control of data to users, but he may have some tricks up his sleeve.
To prove the legitimacy of OpenAI's GPT-3.5 and GPT-4, which were built by scraping the internet, including many Wikipedia pages, and made accessible to the public, Meta may decide to either fully open access to the data or restrict it to Meta only if Threads becomes decentralized.
In either case, Meta can avoid legal issues regarding public data or the training of this social media platform by simply stating that the platform is decentralized. The current version of LLaMa is trained on web scraping data, not on Meta's services. However, it seems that Meta may actually go in that direction.
Meta is Engaged in an AI Competition, Not a Social Media Competition
Instagram, WhatsApp, or Facebook provide a vast amount of data in text, video, or image formats. This may be advantageous for building multimodal AI in the future. However, the issue of data usage legitimacy always exists.
While OpenAI is trying to address the so-called "alignment problem," Meta is catching up. By heavily promoting open-source, Zuckerberg has also freed himself from the responsibility of aligning AI with everyone. The plan is bold, but it may benefit Meta.
On the other hand, Zuckerberg knows that even with the current version of LLaMa, it is still "an order of magnitude smaller than what OpenAI and Google are doing." Perhaps Musk has now given Zuckerberg an idea to build something on the text-based social media platform. Therefore, even though Meta may be four years behind OpenAI in terms of AI, Threads may help them catch up and create something to outperform ChatGPT before Musk does.
Interestingly, Zuckerberg has already started restricting accounts on Threads, which is one of the main reasons Musk expressed interest in acquiring Twitter and building a competitor to wake up ChatGPT. Therefore, even if Meta builds a chatbot, it won't be similar to what Musk wants to build.
Furthermore, compared to Twitter and ChatGPT, Meta is already ahead in attracting users. Threads surpassed 10 million users within seven hours of its release, while ChatGPT took seven days to reach 1 million users. The only questionable aspect is the privacy policy surrounding Threads. Currently, users cannot even delete their accounts without deleting their Instagram accounts. This seems to be a coercive way to keep users on the platform rather than offering something that Twitter doesn't have.
Source: https://analyticsindiamag.com/the-biggest-winner-from-threads-llama/
Alphabet Q1 paid subscribers reached 350M, fueled by YouTube & Google One. YouTube ad revenue slightly missed at $9.88B amid Premium growth. Total Q1 revenue hit $109.9B, beating forecasts.
Anthropic releases Claude connectors for Adobe, Blender, and Ableton, enabling AI workflow automation and app control. It also sponsors the Blender Development Foundation.
Apple introduces an App Store model splitting 12-month plans into monthly installments. On iOS 26.4+, it lowers upfront costs without reducing total price, adding billing tracking.
Google launches Aadhaar digital IDs in Wallet for India, enabling secure QR/NFC verification. A phased partner rollout targets key sectors, while passport-based credentials expand globally.
YouTube tests "Ask YouTube," an AI tool blending text and video for complex queries. Initially for US Premium users, it supports interactive follow-ups, expanding Google's AI search.
Google grants US DoD classified AI access, joining OpenAI and xAI in defense deals amid Anthropic's dispute. Contractual safeguards against surveillance and lethal weapons remain unenforceable.
Mistral AI launches Workflows, a Temporal-powered orchestration platform for secure, Python-driven enterprise AI automation featuring human-in-the-loop approvals.