Google’s DeepMind has introduced GenCast, a cutting-edge AI weather prediction model that outperforms the European Centre for Medium-Range Weather Forecasts’ (ECMWF) ENS, widely regarded as the global leader in operational forecasting. A study in Nature highlighted GenCast’s superior accuracy, predicting weather more effectively 97.2% of the time during a comparative analysis of 2019 data.
Unlike earlier deterministic models, GenCast creates a complex probability distribution of potential weather scenarios by generating 50 or more forecasts per instance. This ensemble approach provides a nuanced understanding of weather trajectories, elevating predictive reliability.
Google is integrating GenCast into its platforms like Search and Maps, while also planning to make real-time and historical AI powered forecasts accessible for public and research use. With this advancement, the tech giant aims to revolutionise weather forecasting and its applications worldwide.
Google’s newest AI, the PaliGemma 2 model, has drawn attention for its ability to interpret emotions in images, a feature unveiled in a recent blog post. Unlike basic image recognition, PaliGemma 2 offers detailed captions and insights about people and scenes. However, its emotion detection capability has sparked heated debates about ethical implications and scientific validity.
Critics argue that emotion recognition is fundamentally flawed, relying on outdated psychological theories and subjective visual cues that fail to account for cultural and individual differences. Studies have shown that such systems often exhibit biases, with one report highlighting how similar models assign negative emotions more frequently to certain racial groups. Google says it performed extensive testing on PaliGemma 2 for demographic biases, but details of these evaluations remain sparse.
Experts also worry about the risks of releasing this AI technology to the public, citing potential misuse in areas like law enforcement, hiring, and border control. While Google emphasises its commitment to responsible innovation, critics like Oxford’s Sandra Wachter caution that without robust safeguards, tools like PaliGemma 2 could reinforce harmful stereotypes and discriminatory practices. The debate underscores the need for a careful balance between technological advancement and ethical responsibility
Meta Platforms has reported that generative AI had limited influence on misinformation campaigns across its platforms in 2023. According to Nick Clegg, Meta‘s president of global affairs, coordinated networks spreading propaganda struggled to gain traction on Facebook and Instagram, and AI-generated misinformation was promptly flagged or removed.
Clegg noted, however, that some of these operations have migrated to other platforms or standalone websites with fewer moderation systems. Meta dismantled around 20 covert influence campaigns this year. The company aims to refine content moderation while maintaining free expression.
Meta also reflected on its overly strict moderation during the COVID-19 pandemic, with CEO Mark Zuckerberg expressing regret over certain decisions influenced by external pressure. Looking forward, Zuckerberg intends to engage actively in policy debates on AI under President-elect Donald Trump‘s administration, underscoring AI’s critical role in US technological leadership.
World Labs, the startup co-founded by AI pioneer Fei-Fei Li, has introduced groundbreaking technology that transforms single images into interactive 3D environments. Unlike existing tools, these AI-generated scenes can be explored and modified directly within a browser, offering a dynamic and engaging experience.
The startup’s system leverages a category of AI known as ‘world models,’ which simulate 3D environments with improved consistency and physical realism. While the technology is still in its early stages, it aims to revolutionise industries like gaming, filmmaking, and design by providing accessible and cost-effective tools for creating virtual worlds.
Backed by $230M in funding from prominent investors, including Andreessen Horowitz and Intel Capital, World Labs is valued at over $1B. The company plans to refine its system further and release its first product in 2025, marking a significant step in the evolution of interactive AI applications.
The US Supreme Court has decided to allow a class-action lawsuit against Meta, Facebook’s parent company, to move forward. The case stems from the Cambridge Analytica scandal, where the political consulting firm accessed personal data from 87M Facebook users and used it for voter targeting in the 2016 US presidential election. Meta had sought to block the lawsuit, but the court dismissed its appeal.
Investors claim Meta failed to fully disclose the risks of data misuse, leading to two major drops in its stock price in 2018 when the extent of the privacy breach became public. Meta has already paid a $5.1B fine and a $725M settlement with users over related allegations.
The lawsuit is one of several legal challenges facing big tech firms. A separate case against Nvidia is under review, as investors allege the company misled them about its dependency on cryptocurrency mining.
OpenAI is under scrutiny after engineers accidentally erased key evidence in an ongoing copyright lawsuit filed by The New York Times and Daily News. The publishers accuse OpenAI of using their copyrighted content to train its AI models without authorisation.
The issue arose when OpenAI provided virtual machines for the plaintiffs to search its training datasets for infringed material. On 14 November 2024, OpenAI engineers deleted the search data stored on one of these machines. While most of the data was recovered, the loss of folder structures and file names rendered the information unusable for tracing specific sources in the training process.
Plaintiffs are now forced to restart the time-intensive search, leading to concerns over OpenAI’s ability to manage its own datasets. Although the deletion is not suspected to be intentional, lawyers argue that OpenAI is best equipped to perform searches and verify its use of copyrighted material. OpenAI maintains that training AI on publicly available data falls under fair use, but it has also struck licensing deals with major publishers like the Associated Press and News Corp. The company has neither confirmed nor denied using specific copyrighted works for its AI training.
Elon Musk’s social media platform X is testing a free version of its AI chatbot, Grok, which was previously exclusive to premium subscribers. Over the weekend, reports surfaced from users and researchers indicating that some free accounts in regions like New Zealand now have access to the AI tool. While usage is capped to 10 queries every two hours for the Grok-2 model, this marks a significant expansion of the technology’s reach.
Grok, developed by Musk’s company xAI, launched earlier this year with advanced features like image generation and understanding, powered by Black Forest Labs’ FLUX.1 model. Previously available only to paying users, the decision to extend limited access to free users may reflect xAI’s strategy to grow its user base and improve feedback for refining its technology.
To use Grok for free, accounts must be at least seven days old and linked to a phone number. This move positions xAI to compete with AI giants like OpenAI’s ChatGPT and Google’s Gemini, while also potentially bolstering its valuation, which reportedly reached $40B in recent funding discussions. This test of free access could accelerate Grok’s development cycle and further establish xAI in the competitive AI market.
Apple has introduced ChatGPT integration with the release of iOS 18.2, iPadOS 18.2, and macOS Sequoia 15.2, allowing developers to explore new features tied to its Apple Intelligence system. The integration enables ChatGPT to enhance Siri’s knowledge and power new writing tools, along with other features like image generation and cleanup tools.
Users who opt into both Apple Intelligence and ChatGPT will be able to leverage OpenAI’s models without needing a separate ChatGPT account, though non-premium users will face limitations on the number of queries. Siri can now call on ChatGPT for certain tasks, such as generating recipes or helping with travel plans, making the virtual assistant more versatile.
Apple Intelligence also includes ‘Compose,’ which lets users generate text based on prompts in supported apps. Alongside this, users can experiment with OpenAI’s image generation or create customised emojis through Apple’s Genmoji tool, offering a more creative and intuitive user experience.
The competition between Elon Musk and Mukesh Ambani is intensifying as they vie for dominance in India’s emerging satellite broadband market. After India’s government decided to allocate satellite spectrum administratively, rather than through auction, the stage is set for a fierce battle. Musk’s Starlink, which uses low-Earth orbit (LEO) satellites, is poised to enter the Indian market, while Ambani’s Reliance Jio has already partnered with Luxembourg-based SES, utilising medium-Earth orbit (MEO) satellites.
The stakes are high as satellite broadband promises to bring internet access to remote areas of India, helping to bridge the country’s digital divide. Both billionaires have taken opposing views on how the spectrum should be allocated, with Ambani pushing for an auction, while Musk argues for the administrative model, aligning with international standards. India’s telecom regulator has yet to announce spectrum pricing, but projections indicate that satellite internet could reach two million subscribers by 2025.
This rivalry underscores the vast potential of the Indian market, where nearly 40% of the population still lacks internet access. Both Musk and Ambani are vying to capture this untapped segment, but pricing will be critical, especially in a country where mobile data is among the cheapest globally. Analysts predict a price war, with Musk’s deep pockets potentially giving Starlink a competitive edge, though challenges remain due to Starlink’s higher costs compared to local providers.
The US Department of the Treasury’s Office of Foreign Assets Control (OFAC) has recently updated its Russia General License (GL) 25E, maintaining authorisation for essential and incidental transactions to telecommunications involving the Russian Federation. That license facilitates various internet-based services, including instant messaging, social networking, and e-learning platforms.
It supports the ongoing exchange of communications and allows for the export or reexport of related software, hardware, and technology, provided such transactions comply with the Department of Commerce’s Export Administration Regulations. However, it is important to note that transactions involving significant Russian telecommunications companies designated by OFAC remain unauthorised under this license and must be carefully analysed.
The Department of the Treasury’s Office of Foreign Assets Control has also issued a critical alert regarding Russia’s attempts to evade sanctions by establishing new overseas branches and subsidiaries of Russian financial institutions. That alert warns that these efforts to open new international branches or subsidiaries should be considered potential red flags for sanction evasion.
Financial institutions and foreign regulators are advised to exercise caution when engaging with these entities, as activities such as maintaining accounts, transferring funds, or providing financial services may carry significant risks of facilitating Russia’s attempts to bypass sanctions.
