Artificial Intelligence News Live: Venture capital sees select firms embrace AI-driven 'quant trading' strategies

Venture capital sees select firms embrace AI-driven 'quant trading' strategies

Artificial intelligence is increasingly influencing venture capital, with some firms starting to fully implement 'quant trading' methodologies—using data science and machine learning for investment decisions rather than rapid trading. Most VC firms are not yet abandoning human experience, but a few are pioneering this AI-driven approach, as per a report in Financial Times. This shift is occurring amidst the rise of AI-focused venture funds and the growing availability of alternative data crucial for training AI models to evaluate private companies. The term 'smarter' machines in this context refers to their capacity for efficient data processing and pattern identification, not human-level intelligence.

Data bill opposed by Sir Elton John and Dua Lipa over AI concerns finally passes in UK

A data bill in the UK, identified in wider reporting as likely the Data Protection and Digital Information Bill, has passed despite significant concerns from prominent artists including Sir Elton John and Dua Lipa, according to a report in BBC. These artists, and the broader creative industries, opposed the bill primarily due to fears that a proposed Text and Data Mining (TDM) exception would allow Artificial Intelligence companies to train models on their copyrighted works without permission or compensation. The final version of the bill reportedly narrowed this TDM exception compared to earlier drafts, now requiring 'lawful access' to content and allowing rightsholders to opt-out, suggesting some creative industry concerns were addressed.

Google launches new AI model and website for tropical storm forecasting

Google has developed a new AI model for forecasting tropical cyclones and, on June 12, 2025, launched a website called Weather Lab to share its AI weather models, The Verge says in a report. This experimental model can generate 50 different scenarios for a storm’s track, size, and intensity up to 15 days in advance, and the US National Hurricane Center (NHC) is collaborating with Google to evaluate its effectiveness. Google released a research paper on the model, which is not yet peer-reviewed, claiming its predictions are at least as accurate as traditional physics-based models, and states its five-day North Atlantic and East Pacific track predictions were, on average, 87 miles closer to actual tracks than European Centre for Medium-Range Weather Forecasts (ECMWF) predictions in 2023 and 2024. The model is trained on Europe’s ERA5 archive, similar to Google's previous GenCast AI weather model, and Google emphasizes Weather Lab is currently a research tool, with AI intended to assist, not replace, traditional forecasting. Google is also working with other research institutions, including the Cooperative Institute for Research in the Atmosphere at Colorado State University and researchers in the UK and Japan, to improve its AI weather models.

AI to assist with earlier offside detection at FIFA Club World Cup 2025

Artificial intelligence will be utilized at the FIFA Club World Cup 2025 to help achieve earlier detection of offsides, Al Jazeera said in a report. This AI-driven system, likely building on existing Semi-Automated Offside Technology (SAOT), is designed to process data from tracking cameras and provide rapid alerts to Video Assistant Referees (VAR), who then make recommendations to the on-field referee. The implementation of AI for offside calls is a key technological innovation for the tournament, which will also see the introduction of other technologies such as referee body cameras.

AI sales startup Landbase secures $30M Series A led by Ashton Kutcher’s Sound Ventures

AI sales startup Landbase has secured $30 million in a Series A funding round led by Ashton Kutcher’s Sound Ventures, TechCrunch said in a report. The company, which emphasizes a concept its CEO calls 'digital trust' and employs what is described as a 'vibe GTM' (go-to-market) strategy to foster authentic connections in digital sales interactions, reportedly attracted pursuit from nearly 130 venture capital firms for this investment.

Libraries including Harvard to provide historical texts and documents for AI training

Tech companies are turning to library collections to enhance AI's understanding of humanity, moving beyond internet-sourced data, as Harvard University is releasing nearly one million books to AI researchers—some dating back to the 15th century and spanning 254 languages—this Thursday, AP says in a report. Additionally, Boston's public library is also preparing to offer its troves of old newspapers and government documents for AI training, aiming to provide AI models with more comprehensive and historically grounded knowledge.

AI and deepfakes used in sophisticated Facebook ad scams featuring celebrity endorsements

Facebook ad scams are increasingly utilizing artificial intelligence and deepfakes to produce highly convincing fake celebrity endorsements, enabling sophisticated fraudulent schemes that target specific user demographics, Fox News said in a report. This advanced use of AI makes these scams significantly more difficult for average users to identify, as the technology allows for the creation of highly realistic, personalized, and scalable fraudulent content. Social media platforms reportedly face ongoing challenges in effectively detecting and moderating this surge in sophisticated AI-generated scam content at scale.

Researchers develop AI tool to restore age-damaged artworks in hours

An AI-based tool has been developed by researchers that can restore age-damaged artworks, including oil paintings suffering from cracks, discoloration, and flaked pigment, in a matter of hours, The Guardian says in a report. This new technique aims to slash the substantial time and expense typically required for art restoration, potentially enabling the digital restoration of paintings previously not considered valuable enough to warrant the cost of traditional conservation efforts.

Oak Ridge site to bolster US AI energy infrastructure amid competition with China

A former Manhattan Project site at Oak Ridge is being developed to support United States' Artificial Intelligence ambitions by addressing the significant energy demands of AI data centers, Fox News says in a report. This initiative, occurring under the current Donald Trump administration, aims to counter concerns that China's rapid expansion of nuclear power could provide it an advantage in powering its AI infrastructure. The effort at Oak Ridge likely involves enhancing energy infrastructure and utilizing the capabilities of Oak Ridge National Laboratory for advanced computing, rather than repurposing historical enrichment facilities for direct power generation for AI.

Aspen fire departments use AI for faster wildfire detection and response

Aspen fire departments are turning to Artificial Intelligence (AI) technology to achieve faster wildfire detection and response, Fox News says in a report. Officials in Aspen aim to have emergency crews on the scene of a wildfire within five minutes of detection, a goal they state would be nearly impossible without the precise location data provided by AI systems. This technology typically involves machine learning algorithms analyzing imagery from strategically placed cameras to automatically identify early signs of wildfires, such as smoke or flames, and then uses the camera's known geographical position and orientation to pinpoint the fire's origin, enabling quicker and more targeted deployment of resources.

Engineer's experiment shows ChatGPT defeated in chess by 1970s Atari 2600 program

An experiment conducted by an engineer resulted in OpenAI's ChatGPT, a large language model, being defeated at chess by a program running on a 1970s-era Atari 2600, CNET says in a report. This outcome underscores that while powerful in language processing, ChatGPT is not designed for strategic, rule-based games like chess and lacks a dedicated chess engine. The Atari 2600 chess program itself was extremely basic, meaning the result primarily illustrates ChatGPT's unsuitability for the task rather than the Atari program's strength or a general failing of modern AI compared to older systems. The event serves as a practical example that selecting the appropriate AI tool for a specific problem is crucial, as specialized algorithms often outperform general-purpose models in narrow domains.

Judge rejects AI chatbot free speech defense in teen's death lawsuit

In a significant ruling within a lawsuit over a teenager's death, a judge has rejected arguments that an AI chatbot's outputs are protected by First Amendment free speech rights, AP said in a report. This decision clarifies that AI companies may not easily use free speech as a defense against liability claims when AI-generated content is alleged to have caused harm, though the ruling does not determine the company's ultimate liability in the ongoing case. The judgment contributes to a growing number of lawsuits addressing AI liability and informs the broader legal and regulatory discussions surrounding AI accountability and the legal status of AI systems.

Industry leaders call on Senate to pass No Fakes Act addressing AI deepfake concerns

A group of industry leaders is advocating for the U.S. Senate to pass the 'No Fakes Act,' legislation designed to address the proliferation of artificial intelligence-generated deepfakes by providing legal recourse against unauthorized digital impersonations, AP says in a report. This call to action underscores increasing concerns about the potential for AI technology to be misused for creating deceptive content that can harm individuals' reputations or spread misinformation. The proposed bill aims to create a legal framework to counter the challenges posed by sophisticated AI-generated media.

Meta plans increased investment in human-generated data for AI development amid scarcity

Meta is reportedly planning a significant strategic investment in human-generated and curated data to advance its artificial intelligence models, as high-quality public training data becomes increasingly scarce, according to a report in Financial Times. This increased emphasis on human input highlights the growing critical importance and scale of such contributions for developing sophisticated AI, particularly for fine-tuning, alignment, and evaluation stages which require specific types of high-quality, diverse, and labeled data. The human-led efforts are expected to involve a range of tasks, including data labeling, ranking AI-generated outputs, providing corrective feedback, and curating datasets, aligning with a broader industry trend where human-in-the-loop processes are essential for enhancing AI model safety, helpfulness, and accuracy.

Judge considers sanctions against attorneys for AI use in prison case court filings

A judge is currently considering sanctions against attorneys following their use of artificial intelligence in court filings related to a prison case, AP says in a report. This situation highlights ongoing concerns within the legal profession about the reliability and proper use of AI tools, particularly regarding the potential for generative AI to produce inaccurate information or 'hallucinations,' such as non-existent case citations, which can undermine the integrity of legal proceedings. Courts and bar associations have been increasingly addressing these issues, with some jurisdictions implementing rules requiring disclosure of AI use and certification of accuracy for AI-generated content in legal documents.

New US supercomputer named for Jennifer Doudna to power AI and scientific research

The United States has unveiled a new supercomputer named after Nobel laureate Jennifer Doudna, intended to bolster AI capabilities and support a wide range of scientific research, AP says in a report. This high-performance computing system will provide crucial resources for training sophisticated AI models and running complex simulations, with an anticipated focus on genomics, life sciences, and other areas at the intersection of AI and biological research, aligning with Doudna's pioneering work. The development is seen as a key component of the US strategy to maintain leadership in AI and scientific innovation, contributing to the national AI ecosystem.

AI industry's "reasoning" claims under scrutiny as studies point to pattern-matching reality

The recent launch of AI models such as "o3-pro" is prompting a closer examination of what constitutes "reasoning" in artificial intelligence, with new studies indicating that these capabilities are largely based on advanced statistical pattern matching on vast datasets rather than genuine logical deduction, Ars Technica says in a report. This aligns with an ongoing debate within the AI community about the true mechanisms behind large language model (LLM) performance and whether current benchmarks adequately test for true reasoning. Consequently, there's an increasing research focus on AI interpretability to understand how models arrive at outputs, distinguishing between mimicking patterns and actual cognitive processes.

Klarna launches AI hotline with CEO's digital persona for customer feedback

Klarna has introduced an AI-powered phone hotline enabling users to interact with an interactive AI version of its CEO, Sebastian Siemiatkowski, to provide feedback on products, services, and suggest improvements, The Verge reports. This AI, trained on Siemiatkowski's real voice, insights, and experiences, also answers questions about Klarna's vision and history. After calls, transcripts and summaries are generated and then automatically analyzed by a large language model (LLM), populating an internal dashboard for Klarna's product and engineering teams to review for actionable insights, with the company claiming this could lead to rapid product enhancements. The service is currently available via phone numbers in the US and Sweden, and the AI primarily directs conversations towards gathering feedback.

Nvidia and Samsung plan investments in AI robotics startup Skild AI

Nvidia Corp. and Samsung Electronics Co. are planning to take minority stakes in Skild AI Inc., a startup operating in the emerging consumer robotics industry, Bloomberg said in a report. This strategic investment is aimed at bolstering the efforts of both technology companies as they seek to expand their work and influence in the field of artificial intelligence-driven consumer robotics.

New Apple study questions true reasoning capabilities in AI models, findings disputed

A new study from Apple, utilizing puzzle-based experiments, challenges the notion that current artificial intelligence models genuinely engage in reasoning, instead suggesting they predominantly simulate reasoning processes through sophisticated pattern matching, Ars Technica says in a report. The research indicates that while AI can produce outputs resembling logical steps, this may be achieved via statistical inference learned from data rather than an explicit internal logical process or true understanding, particularly when faced with tasks requiring complex deduction or multi-step planning. However, the findings and their implications are reportedly contested by other experts in the field, with disputes potentially arising from differing definitions of 'reasoning,' the specific methodologies of the puzzle tests employed, or arguments that different or more advanced AI models might yield different results.