What powers the targeting of ads on Meta?
In Meta’s commitment to being more transparent, it has revealed how its ad AI back-end systems work.
This allows users to understand the algorithms behind Meta and how ads are targeted to audiences.
Here are the three AI innovations in ads ranking:
- Meta GEM: The Super Brain
GEM catalogues, analyses, and links trillions of data points. By learning from vast amounts of data and recognising subtle patterns, GEM delivers the most relevant ads to the right audiences at the right time. Meta has reported that GEM increased overall ad conversions by 5%.
- Meta Lattice: The Unified Knowledge Base
Meta used to have many libraries related to each industry, but the Meta lattice has finally combined them all (essentially into one huge library). In ads, this means using a single, powerful system to deliver more relevant and effective ads, instead of relying on multiple smaller systems focused on specific products or placements. This approach allows for more efficient learning and faster delivery of relevant ads.
- Meta Andromeda: The Personal Concierge
Introduced in 2024, Meta Andromeda is a system that essentially learns users’ preferences through engagements and interactions and as a result, shows them ads that are more relevant and desired by users.
Together, these AI-driven innovations have enhanced ad engagement ensuring that users see more resonating content in their feeds.
Gemini 2.5 is available to all
Previously available only to paying Gemini Advanced and Google AI Studio customers, Google announced over the weekend that the experimental Gemini 2.5 model will now be available free of charge to all users.
The newest model holds the top spot on LMArena, a benchmark for evaluating human preferences in AI responses.
Gemini 2.5 Pro offers a huge upgrade in processing large datasets efficiently. Its one million token context window allows it to handle extensive information and manage long text inputs.
Plans to expand the AI model to two million tokens will significantly improve its ability to analyse lengthy conversations, documents and codebases.
However, there is speculation on how long the model will be available for free to the public.
Controversy as Musk seeks removal of Reddit posts
One of the primary motivations behind Musk’s acquisition of X was his stance against content moderation.
Musk took to X with a rant against Reddit, accusing certain users of ‘breaking the law’ by posting inappropriate comments about DOGE (The Department of Government Efficiency) employees.
Reports indicate that Musk personally reached out to Reddit CEO Steve Huffman, demonstrating his discomfort with content he disagrees with.
The X post from Musk was deleted shortly after the conversation was reported.
Musk’s frustration with X extends beyond the DOGE subreddits incident. It also stems from when he referred to it as “insane” that certain subreddits had banned users from posting links to X, further highlighting his growing dissatisfaction with content control on the Reddit platform.
Despite previously opposing content moderation, Musk appears to now seek greater control over the content on other platforms as well.
Read more here.
X introduces ‘reply with a video’ feature
As X shifts its social media platform towards being ‘video-first’, they have introduced a new feature to boost user engagement.
Now users can post a video response to an X update under a “react with video” option, making it easier to create reaction clips during live streams and posts.
This feature is similar to TikTok’s “reply to comment with video” function, showing that platforms are increasingly allowing users to engage with audiences in interactive ways.
This can be great for brands to reply to comments FAQs, tutorials and tease products.
Read more here.
Elon Musk’s xAI acquires X
Elon Musk announced that his AI company, xAI, has acquired his social media platform, X (formerly Twitter), in an all-stock deal that has shaped a powerful ecosystem under his control.
The deal values xAI at $80 billion and X at $33 billion, according to Musk’s post last Friday on X.
xAI owns Grok, an advanced AI chatbot tool that is specifically trained on data from X. A major advantage of Grok is its ability to access real-time data, allowing it to deliver highly relevant responses based on the latest content from X.
This real-time access enables Grok to stay in tune with current trends and conversations, while also engaging users in a more context-aware way, enhancing the overall interaction.
Musk stated that the futures of both companies are now intertwined, combining crucial resources like data, AI models, and talent.
This strategic merger is seen as a move to protect the investors who supported his $44 billion acquisition of X in 2022.
However, we expect the announcement of this acquisition to raise concerns about data privacy and the ethical use of AI within the X platform.
Read more here.