September 27, 2025

analysis  
Meta Launches Vibes, Anthropic $1.5B Deal, $800B AI Shortfall

Meta has introduced Vibes, a new feed on its Meta AI app that exclusively features AI-generated videos. This feed, available in the US and over 40 other countries but not the UK, allows users to create and remix AI videos, showcasing content from artists and creators. Vibes replaces a previous feed that focused on chatbot interactions and now displays videos alongside their creation prompts, with options to alter visuals and music. Meanwhile, the AI industry faces significant financial considerations, with a report suggesting that by 2030, companies might require an additional $2 trillion in AI revenue to remain profitable, facing an $800 billion shortfall. This intense investment in AI is seen as crucial for the U.S. economy. In legal news, a federal judge approved a $1.5 billion settlement where AI company Anthropic will compensate authors for allegedly using nearly half a million books to train its chatbots. Separately, LinkedIn will begin using member data for AI model development starting November 3, 2025, requiring users to opt out manually if they wish to prevent their information from being used. The broader implications of AI are also being explored in creative and ethical spheres. ArtPrize artists are incorporating AI as a tool in their work, while discussions around AI's impact on human intelligence and relationships continue, with some experts cautioning against AI becoming a substitute for human thought. The need for contextual intelligence in AI is also highlighted, as current models often lack the nuance for real-world applications. Innovations in AI privacy are emerging, with DESILO's THOR framework enabling large language models to run under homomorphic encryption, ensuring data privacy. Efforts to combat bias in AI are also underway, with Justice AI GPT designed to identify and rewrite biased language in documents.

Key Takeaways

  • Meta has launched Vibes, a new feed on its Meta AI app dedicated to AI-generated videos, allowing users to create and remix content.
  • A federal judge approved a $1.5 billion settlement between AI company Anthropic and authors over the use of books for AI training.
  • A report indicates that companies may need an additional $2 trillion in AI revenue by 2030, with a potential $800 billion shortfall, highlighting significant AI investment.
  • LinkedIn will use member data for AI model development starting November 3, 2025, requiring users to opt out to prevent data usage.
  • Innovations like DESILO's THOR framework are enabling large language models to operate under homomorphic encryption for enhanced privacy.
  • The development of contextual intelligence is crucial for AI to improve reasoning and accuracy in real-world applications.
  • Justice AI GPT is a framework designed to address and dismantle bias in AI systems by using decolonial datasets and rewriting inclusive language.
  • ArtPrize artists are exploring the role of AI in creativity, using it as a tool in their artistic processes.
  • Concerns are being raised that while AI may not cause job losses, it could lead to a reduction in human intelligence if used as a substitute for thinking.
  • The debate continues on the acceptable use of AI in education, with professors encouraged to set clear guidelines for students.

Meta launches Vibes AI video feed

Meta has launched Vibes, a new feed on its Meta AI app that features videos created entirely by artificial intelligence. The feed showcases AI-generated content from artists and creators, including videos of fluffy creatures and historical selfies. Users can create their own AI videos or remix existing ones with new visuals and music. Vibes uses video models from Midjourney and Black Forest and is available in the US and over 40 other countries, but not the UK.

Meta AI app gets new Vibes feed for AI videos

Meta AI has introduced Vibes, a new feed within its app dedicated to short-form, AI-generated videos. This feed encourages users to remix content from creators and communities. Users can also share their AI creations across Meta's platforms like Instagram and Facebook. Vibes replaces a previous feed that showcased user prompts and interactions with Meta AI, now focusing solely on video content. Videos in Vibes appear alongside the prompts used to create them, and users can remix them by changing music, visuals, or animation.

Meta's Vibes AI feed offers random AI videos

Meta's AI app now features a new feed called Vibes, which displays a continuous stream of random AI-generated video clips. This feed replaces a previous one that showed chatbot conversations. While the feed is currently random, Meta plans to personalize it over time. Online reactions to Vibes have been largely negative, with many describing the content as low-quality AI-generated material. The feed aims to showcase the capabilities of AI tools, allowing users to create and remix videos.

Judge approves $1.5 billion AI copyright settlement

A federal judge has approved a $1.5 billion settlement between AI company Anthropic and authors. The authors claimed Anthropic illegally used nearly half a million books to train its chatbots. The settlement will pay authors and publishers around $3,000 per book. U.S. District Judge William Alsup approved the deal in San Francisco federal court. He had concerns about the claims process to ensure all eligible parties were notified and received their share. The agreement does not cover future works.

AI spending may exceed revenue, report says

Tech companies are investing heavily in artificial intelligence, but this spending might outpace potential revenue. A Bain & Company report suggests that by 2030, companies may need an additional $2 trillion in AI revenue to be profitable, with an estimated shortfall of $800 billion. The demand for AI is growing rapidly, doubling the pace of Moore's Law, which itself is facing physical limitations. Without AI infrastructure spending, the U.S. economy might be in a recession. The report highlights the economic challenges of keeping up with AI's computational demands.

Contextual intelligence is key for AI's future

The rapid growth of Generative AI (GenAI) highlights the need for contextual intelligence. While current AI tools are impressive, they often lack the nuance and accuracy required for real-world applications because they miss context. Human minds process information instinctively using deep knowledge and local insights, something current AI struggles to replicate. Contextual intelligence aims to give AI access to deep, specific knowledge, improving its reasoning for tasks like creating reports and presentations. This is often achieved through retrieval augmented generation (RAG), which indexes data from various sources to train AI models.

DESILO's privacy-safe AI breakthrough recognized

DESILO Inc. has developed a new framework called THOR that allows large language models (LLMs) to run fully under homomorphic encryption, ensuring privacy. Their research, accepted at ACM CCS 2025, demonstrates running an existing LLM without retraining, keeping inputs and outputs encrypted. The system achieves near practical speeds, with performance improvements in matrix multiplication operations. THOR is presented as the first framework to run an entire LLM under homomorphic encryption at near-practical speeds. This innovation supports DESILO's upcoming solutions like Harvest, designed for secure analysis across multiple institutions.

LinkedIn to use user data for AI training

Starting November 3, 2025, LinkedIn will use member data for AI model development, a policy already active in the US and other regions. Users aged 18 and over must manually opt out through their account settings to prevent their information from being used. The data may include account details, email, payment information, and service data like IP addresses. LinkedIn states this practice complies with privacy rules, though previously collected data may remain. Users can request the removal of past data via a Data Processing Objection Form.

ArtPrize artists explore AI's role in creativity

This year's ArtPrize competition features artists exploring artificial intelligence, with some using it as a tool and others making it a topic. While robots are not allowed as artists, human artists can use AI in their creative process. Examples include using AI for accessibility, like helping a dyslexic artist write statements, or incorporating AI-generated elements into applications while submitting hand-created pieces. Some artists use AI directly in creating new media art. The event encourages visitors to consider the role and extent of AI in art.

AI won't cause job losses but could reduce intelligence, CEO says

Will Wilson, CEO of Antithesis, told a panel at Catholic University of America that artificial intelligence will not lead to mass unemployment because human desires are infinite. However, he expressed concern that people might use AI as a substitute for their own intelligence, potentially making them less intelligent. Father Michael Baggot also warned about 'artificial intimacy' distracting from human relationships and called for AI guardrails. The discussion touched on the ethical considerations of AI, including its potential impact on human minds and affections, and the church's role in addressing these issues.

AI's place in the classroom debated

The role of artificial intelligence in education is a subject of ongoing debate, with professors holding differing views on its use. While using AI to write entire essays is considered cheating, the permissibility of using it for drafting outlines or brainstorming ideas is less clear. The article argues that the learning process itself, including tasks like outlining and brainstorming, is crucial for student development. It suggests that professors should determine the acceptable use of AI in their classrooms, advocating for a mindful and cautious approach from both students and educators.

Justice AI GPT aims to fix bias in AI systems

Christian Ortiz has developed Justice AI GPT, a framework designed to address bias in artificial intelligence. Ortiz argues that bias is inherent in AI design, often stemming from Eurocentric datasets. Justice AI GPT counters this by pairing large datasets with a 'decolonial dataset' created with global experts, aiming to dismantle biased information at its source. The tool identifies and rewrites biased language in areas like job postings and HR documents to be more inclusive. Currently, 112 organizations use Justice AI GPT for bias audits and DEI coaching, with plans to expand its reach.

Sources

AI video generation Meta AI Vibes feed AI copyright settlement Anthropic AI spending AI revenue Contextual intelligence Generative AI Retrieval Augmented Generation (RAG) Homomorphic encryption DESILO Privacy-safe AI Large Language Models (LLMs) LinkedIn AI training AI in art ArtPrize AI and employment AI and intelligence AI in education AI bias Justice AI GPT Decolonial dataset DEI coaching