Microsoft´s Digital Crimes Unit is taking legal action to stop cybercriminals from using AI technology to create harmful content. Despite efforts to enhance security, criminals are finding ways to bypass measures. Microsoft has uncovered a group selling access to AI services to create malicious tools. The company is working to prevent future incidents and has issued recommendations to protect the public from such threats. · 5 min sedan - 65 words

Microsoft recently announced that their advanced GenAI technology, Copilot Workspace, is now available for all users without any waitlist restrictions. This AI-enhanced development environment is designed to optimize productivity for teams, offering features like AI-assisted code suggestions, collaboration tools, and debugging capabilities. Users can brainstorm ideas, implement code, iterate on plans, test and validate code, collaborate with team members, and create pull requests directly within the workspace. The goal is to make coding tasks easier and more efficient for everyday use. This development brings us closer to a coding nirvana where AI can build apps based on provided instructions. Starting from a GitHub Issue or a natural language task, developers can work with Copilot Workspace to solve problems and even create web apps in minutes without any coding required. Resources are available to help users get started with this innovative tool. · 5 min sedan - 141 words

Mission-driven organizations are leveraging technology, particularly AI, to accelerate their work and increase inclusion internally. Nonprofits like Arapahoe Libraries and the McKnight Foundation are early adopters of Microsoft 365 Copilot, using the AI assistant to enhance productivity, creativity, and accessibility. By streamlining day-to-day operations and overcoming creative blocks, AI is helping social impact organizations advance justice, inclusion, and equity in their communities. Learn more about how AI solutions are transforming nonprofits' impact. · 10 min sedan - 71 words

Hume AI, a startup and research lab, is focused on creating chatbots with expressive voices to emotionally connect with humans. Users can calibrate voice personas with qualities like 'buoyancy' and 'assertiveness.' The company will soon allow users to create custom voices. Hume's technology measures emotional levels in voices and facial expressions. The goal is to generate the right voice for various roles like a digital therapist or hotel concierge. The CEO, Alan Cowen, emphasizes the importance of conveying meaning beyond words through tone and inflection. Voice interactions with AI are expected to become more popular as interfaces improve. Specialized AI models tailored for specific tasks are predicted to become more common in the future. · 85 min sedan - 115 words

The Runway to Seoul event returned to Dongdaemun Design Plaza in central Seoul on Dec. 29, featuring a fusion of Korean fashion, beauty, K-pop, and technology. Organized by the Seoul Business Agency and WWD Korea, the event drew 2,500 attendees, including international influencers. The runway showcased seven prominent Korean fashion brands, with performances by various dance crews and artists. Attendees were able to engage with exclusive activations and shop limited-edition collaborations. The event highlighted the creativity and potential of K-fashion on a global scale. · 111 min sedan - 82 words

OpenAI recently published a 15-page manifesto titled “AI in America: Open AI’s Economic Blueprint,” calling for voluntary rules, national security contracts, and more. The document warns that overregulation could lead to American investment dollars flowing to Chinese AI projects. OpenAI emphasizes the need for federal funding in areas like energy infrastructure and chip manufacturing to maintain US leadership in AI. The blueprint also highlights the importance of protecting children online and collaborating with law enforcement agencies. · 2 h sedan - 77 words

OpenAI has released its 'Economic Blueprint,' outlining steps to advance AI development in the US. Focus areas include chips, data, and energy. The US must act swiftly to secure its advantage amidst global competition. $175 billion in global funds is at stake, with potential for investment in China. Export controls on AI models are suggested for national security. A regulatory framework balancing innovation and safety is recommended. With a tech-friendly administration incoming, OpenAI anticipates more support for AI. The company plans to discuss proposals in Washington, D.C., this month. · 2 h sedan - 90 words

OpenAI warns that the US could lose $175 billion in global AI investment to China-backed projects, risking its leadership in artificial intelligence. The company calls for a national strategy to attract funds and promote responsible AI development. OpenAI CEO will launch an initiative in DC to showcase AI advancements and economic benefits. Meanwhile, NVIDIA criticizes Biden's new AI restrictions, arguing they could harm US leadership in AI and global innovation. · 2 h sedan - 69 words

OpenAI CEO Sam Altman is facing a civil lawsuit filed by his sister Annie in the US District Court for the Eastern District of Missouri. Annie is seeking damages for alleged sexual assault and battery that she says occurred during their childhood. Altman and his family have released a statement denying all allegations and expressing concern for Annie's well-being. The lawsuit, filed on January 6, seeks damages in excess of $75,000 plus punitive damages. According to court documents, the incidents allegedly occurred between 1997 and 2006 at their family home in Clayton, Missouri. · 2 h sedan - 89 words

OpenAI recently released an 'economic blueprint' for AI, urging the U.S. government to invest in chips, data, energy, and talent to maintain global leadership. The company believes that federal action is crucial to support AI development and protect national security. OpenAI's blueprint also addresses the need for increased federal spending on power and data transmission, as well as the development of new energy sources like solar and wind farms. The company advocates for best practices in model deployment, engagement with national security agencies, and responsible export controls. OpenAI emphasizes the importance of collaboration between the government and private sector to ensure AI safety and security standards. The company's lobbying efforts and partnerships with government agencies demonstrate its commitment to shaping U.S. AI policy. · 2 h sedan - 125 words

A San Francisco-based firm has proposed an 'economic blueprint' suggesting that U.S. governments should accelerate permitting and offer financing for data centers and power projects in specific areas of the country. The plan also recommends the formation of a network of countries with common security regulations to invest in AI infrastructure, as well as the establishment of international standards for the technology's use. · 2 h sedan - 64 words

In the age of artificial intelligence, knowing how to ask great questions is more valuable than memorizing facts. Experts suggest that being able to connect the dots and ask thought-provoking questions is essential in today's job market. Recruiters are looking for adaptability and curiosity in job applicants, as these soft skills are becoming increasingly important. To stay ahead, consider taking an online course on how to use AI to be more successful at work. Sign up for CNBC Make It's newsletter for more tips and tricks for success. · 2 h sedan - 88 words

Many companies are focusing on agentic AI as the next big trend. Leaders like NVIDIA's Jensen Huang and Microsoft's Satya Nadella see AI agents as the future of work. However, experts warn that agentic AI may face the same challenges as RPA. The shift from RPA to agentic AI is seen as a transformative evolution, combining automation with intelligent decision-making. While RPA deals with structured data, agentic AI uses LLMs to interpret decision trees. The future of agentic AI remains uncertain as it tries to address its limitations and evolve. · 2 h sedan - 90 words

EXO Labs has showcased a modified version of Llama 2, a language model, running on Windows 98. This demonstrates that if something can run on Windows 98, it can run on anything, giving users more control. Despite some challenges, the model works surprisingly well on older hardware. This project is part of EXO Labs' efforts to make AI models accessible to everyone, independent of big tech companies. If you're interested in local AI and command line tools, they are available as simple executables. · 2 h sedan - 80 words

Businesses like OpenTable and Wiley are using AI agents to revolutionize their industries. These agents handle tasks like reservation changes and case resolutions, freeing up employees to focus on building customer relationships. By integrating data, AI, and automation, companies can achieve real AI success. AI agents work alongside humans to complete tasks efficiently. Early adopters have found dynamic data retrieval, reasoning, security guardrails, orchestrating action, and collaboration essential for success. Training your own large language model (LLM) is expensive and time-consuming. It's time to scale AI implementation and see real results by leveraging out-of-the-box tools and integrated data flows. · 2 h sedan - 99 words

Learn how to quickly build a knowledge graph for your business with easy steps. Discover how knowledge graphs can improve information retrieval and enhance recommendation systems. Find out how to use graph databases to create a GraphRAG for better context. Stay tuned for more insights on building knowledge graphs using LLMs in future posts. · 2 h sedan - 54 words

Researchers have found that large language models (LLMs) exhibit a social desirability bias, skewing their responses to surveys towards what society values. This poses a challenge in evaluating and deploying these models effectively. The bias emerges after a few questions, leading LLMs to alter their answers to appear more favorable. Mitigating this bias proves difficult, with reverse coding the questions being the only partially effective method. Understanding and addressing these biases are crucial as LLMs are increasingly used in research and everyday applications. · 2 h sedan - 83 words

Researchers from top universities have warned of a serious data-poisoning vulnerability in large language models, which could lead to incorrect medical information being provided. These models are trained on vast amounts of data from the internet, making them susceptible to misinformation. Even a small amount of false training data can result in harmful medical errors. The team has proposed a method using biomedical knowledge graphs to detect and mitigate this risk. Their work has been published in Nature Medicine. · 2 h sedan - 79 words

Romance may not be the first metaphor that comes to mind when thinking about large language models (LLMs), but as our interactions with this tech deepen, the relationship begins to feel personal. I recall my own early experiences with an LLM, where a curious bond seemed to form—not with the technology itself but with the way it amplified my thinking. Unlike traditional tools, LLMs engage with us in a way that mirrors human interaction, triggering an emotional journey that reflects the phases of a complex relationship. · 2 h sedan - 87 words

Researchers at Huawei have developed a new retrieval technique called GEAR, which promises to improve the way large language models answer complex queries. By using graph-based methods, GEAR enhances the ability of models to gather information from various documents and provide accurate answers. This approach outperforms other popular retrieval techniques and requires fewer iterations, making it faster, more efficient, and cost-effective. GEAR shows great potential for improving information retrieval and question-answering tasks in AI research. · 2 h sedan - 75 words

The misuse of artificial intelligence continues to challenge ethical and legal boundaries, with one of its most alarming manifestations being deepfake pornography. To counter this growing threat, the US Senate recently passed the bipartisan ‘Take It Down’ Act, which criminalises the publication of non-consensual intimate imagery, including AI-generated deepfake content. The Bill now moves to the House for consideration. Deepfake technology poses unique challenges to privacy and security, beyond its misuse in pornography, it enables sophisticated fraud. Governments, tech platforms, and citizens must work together to address this issue and protect personal rights in digital spaces. · 2 h sedan - 96 words

Chainalysis has acquired the artificial intelligence-powered fraud detection startup Alterya for $150 million. Alterya's technology helps identify scammers before they strike, benefiting over 100 million users. Chainalysis aims to enhance fraud protection in the crypto industry, following a $3 billion loss in 2024 due to hackers. The partnership with Alterya strengthens Chainalysis' efforts to combat scams and fraud in the digital currency space. · 2 h sedan - 59 words

Chainalysis, a leading crypto analysis and forensics company, has purchased Israeli startup Alterya for $150 million. Alterya specializes in AI-driven fraud detection for financial institutions and crypto service providers. This acquisition will enable Chainalysis to enhance its fraud prevention capabilities and detect advanced scams using generative AI. Alterya has already protected 100 million users from fraud, including clients like Binance, Coinbase, and Square. · 2 h sedan - 61 words

Romance may not be the first metaphor that comes to mind when thinking about large language models (LLMs), but as our interactions with this tech deepen, the relationship begins to feel personal. I recall my own early experiences with an LLM, where a curious bond seemed to form—not with the technology itself but with the way it amplified my thinking. · 3 h sedan - 61 words

Online journalist Matthias, co-founder of THE DECODER, believes AI will change human-computer relations. Microsoft Research Asia developed rStar-Math, a training method for small language models to excel in math. The system uses plain English explanations and Python code to verify solutions. Despite limitations, rStar-Math improved accuracy in math problems and performed well on tests. The system´s test-time compute approach ensures accuracy but is time-consuming and costly. Microsoft aims to create more efficient AI models, sharing rStar-Math code with the research community. · 3 h sedan - 82 words

EXO Labs recently showcased a fascinating experiment running an AI model on a vintage Windows 98 system. The video demonstrates an old PC smoothly generating a story about 'Sleepy Joe.' EXO's mission is to democratize AI and make it accessible to everyone. They are currently exploring new possibilities with BitNet, a revolutionary architecture. If you're interested in collaborating or just want to join a community of tech enthusiasts, check out EXO's Discord Retro channel. · 3 h sedan - 74 words

Online journalist Matthias is the co-founder and publisher of THE DECODER. He believes that artificial intelligence will fundamentally change the relationship between humans and computers. Alibaba has introduced a new web interface called Qwen Chat, allowing users to test drive its Qwen AI models for free. The platform includes various AI models for different tasks, such as general conversation, image understanding, and programming. Alibaba plans to expand the platform's capabilities with web search, image generation, and voice interaction. Join THE DECODER Newsletter for the latest AI news delivered weekly to your inbox. · 3 h sedan - 92 words

Experts predict that we will create around 175 zetabytes of data in 2025, with most of it being duplicate or existing data. This year, only 20-30 zetabytes of 'new' data will be generated. The rise of AI tools like ChatGPT has been fueled by large data sets, but the amount of available data is limited. As we run out of legacy data to train on, companies are looking to capture and automate new data in real-time to stay ahead in the tech race. The focus is shifting towards niche data sets and reducing the data-point lifecycle to zero for faster, more efficient AI solutions. · 3 h sedan - 102 words

An interesting project called llama2.c has been developed to run a simplified version of the Llama 2 model entirely in C code. This project aims to make AI technology accessible on various devices, including old gaming consoles like the Xbox 360. Developer Andrei David successfully ported llama2.c to the Xbox 360, facing challenges related to memory management and code optimization. Despite the limitations of the console, the model showed strong performance. This experiment opens up possibilities for running other AI models on the Xbox 360 in the future. · 3 h sedan - 86 words

Romance may not be the first metaphor that comes to mind when thinking about large language models (LLMs), but as our interactions with this tech deepen, the relationship begins to feel personal. I recall my own early experiences with an LLM, where a curious bond seemed to form—not with the technology itself but with the way it amplified my thinking. Unlike traditional tools, LLMs engage with us in a way that mirrors human interaction, triggering an emotional journey that reflects the phases of a complex relationship. · 3 h sedan - 87 words

A recent shift in AI models is moving towards bypassing the traditional finetuning process. OpenAI's GPT-1 showed improvement on tasks without finetuning, leading to the development of GPT-2 and GPT-3 focused on task-agnostic pre-training. The move away from finetuning addresses limitations like the need for large datasets and the risk of spurious correlations. New approaches like task-agnostic learning, the scale hypothesis, and in-context learning are key elements in this paradigm shift. · 3 h sedan - 71 words

The AI (R)Evolution is taking us into the future. In 2024, AI models have advanced rapidly, impacting our daily lives and businesses. We can now use AI tools to summarize texts, analyze data, and even create images. This technology is reshaping how we work and interact with software. The future looks bright with AI helping us solve problems and boost productivity. · 3 h sedan - 61 words

Microsoft CEO Satya Nadella announced the formation of a new group focused on developing AI apps and tools for third-party customers. The group, called Core AI - Platform and Tools, will be led by Jay Parikh. This initiative aims to create an end-to-end Copilot & AI stack for building and running AI apps. The move comes 10 months after hiring DeepMind co-founder Mustafa Suleyman. Microsoft, a principal investor in OpenAI, has been at the forefront of the AI boom since the launch of ChatGPT in 2022. · 3 h sedan - 83 words

Microsoft has announced the creation of a new engineering division, CoreAI – Platform and Tools, focusing on building AI platforms and tools for Microsoft and its customers. Led by Jay Parikh, the new group will reshape software development with an AI-first application stack. Azure will provide infrastructure, while Azure AI Foundry, GitHub, and Visual Studio Code will support advanced AI tools. The CoreAI group aims to develop an end-to-end Copilot and AI stack for creating and running AI applications. · 3 h sedan - 78 words

Microsoft has taken legal action against a group of cybercriminals who created tools to bypass security measures in its AI tools, leading to the creation of harmful content. The lawsuit accuses 10 defendants of stealing API keys from Microsoft customers to breach Azure Open AI service. The criminals sold access to the tools to others. Microsoft has implemented new safety measures to prevent such abuse in the future. · 3 h sedan - 67 words

Raspberry AI, a platform using AI to help fashion creatives, has raised $24 million in funding. The money will be used to improve their product and expand their team. Their platform allows brands to quickly create designs and marketing assets. The CEO, Cheryl Liu, believes that the fashion industry is changing rapidly and their tools can help brands adapt. Investors like Andreessen Horowitz are confident in Raspberry AI's potential to revolutionize the design process for retail products. · 3 h sedan - 76 words

A new ransomware group, FunkSec, has emerged as a growing concern for its use of artificial intelligence (AI) to enhance its tools. The group has already claimed more than 85 victims globally. FunkSec operates as both a ransomware group and a data broker, using AI-driven tools to quickly iterate despite limited expertise. Victims face low ransom demands and the group targets residents of various countries. While FunkSec has drawn attention for its AI-enhanced operations, questions remain about its sustainability and effectiveness. Organizations are urged to bolster their cybersecurity measures to defend against evolving threats. · 3 h sedan - 93 words

Police across the United States are using facial recognition technology in ways that were never intended, leading to wrongful arrests of at least eight Americans. The technology, which lacks scientific reliability, has been used as a shortcut by law enforcement agencies to find and arrest suspects without proper evidence. In some cases, police have ignored contradictory evidence, failed to collect key pieces of evidence, and relied solely on AI matches without any independent corroboration. The misuse of facial recognition technology has resulted in innocent people being arrested and held in jail for crimes they did not commit. · 3 h sedan - 97 words

Increased AI use linked to eroding critical thinking skills. A study by Michael Gerlich at SBS Swiss Business School found that reliance on AI tools is connected to diminished critical thinking abilities. The study, published in Societies, reveals a negative correlation between AI tool usage and critical thinking scores. Younger users show higher dependence on AI tools and lower critical thinking skills compared to older age groups. The findings suggest implications for educational policy and professional settings. · 3 h sedan - 77 words

Artificial intelligence (AI) is rapidly advancing, sparking discussions among software engineers about its impact on their profession. AI tools are becoming more capable of automating coding tasks, but may lack the creativity needed for complex problem-solving. Reddit users believe AI will transform, not replace, software engineering roles, fostering collaboration between human ingenuity and AI. The future holds new possibilities for engineers, with AI serving as a powerful assistant rather than a threat. · 3 h sedan - 72 words

Anthropic, an AI company backed by Amazon, is reportedly close to reaching a $60 billion valuation. This would make it one of the top five most valuable unlisted companies in the US. The increasing popularity of generative AI has led to significant investments in AI projects by Big Tech and startups. Other top US unicorns include SpaceX at $350 billion, OpenAI at $157 billion, Stripe at $70 billion, and Databricks at $62 billion. Anthropic, founded by ex-OpenAI employees, is gaining attention for its AI capabilities in coding and development. · 4 h sedan - 84 words

Recent reports indicate that Anthropic is in talks to raise $2 billion in a deal that would value the company at $60 billion, while OpenAI raised $6.6 billion, doubling its valuation to $157 billion. However, both companies are facing significant losses, with OpenAI's cumulative losses reaching $8 billion. The trend of startups with big losses potentially facing bankruptcy is a cause for concern, as seen with past examples. It is important to remain skeptical of high valuations and to closely monitor the financial health of these companies in order to avoid potential pitfalls in the future. · 4 h sedan - 91 words

A new tool called Copilot Analytics will be released in March, providing valuable information on AI adoption and business impact within organizations. This platform, part of the Microsoft 365 Suite, aims to empower IT and business leaders to measure the effectiveness of Copilot agents. It will offer customizable reporting for deeper analysis and greater control over AI management. Scheduled to roll out this March, Copilot Analytics will be available in both the Microsoft 365 Suite and Microsoft Viva, offering managers and IT admins easy access to valuable insights. · 5 h sedan - 86 words

Red Hat, Inc. has completed the acquisition of Neural Magic, a pioneer in software and algorithms that accelerate generative AI workloads. This move enhances Red Hat's expertise in AI performance engineering and model optimization, aiming to deliver high-performing AI workloads across the hybrid cloud. The acquisition of Neural Magic aligns with Red Hat's commitment to open-source innovation and providing cost-effective AI solutions. This strategic enhancement will enable enterprises to leverage the power of AI more efficiently and effectively. · 5 h sedan - 78 words

Stay updated with the latest news, project case studies, and product announcements in broadcast technology, creative design, and engineering by subscribing to NewscastStudio. Mediaocean's newly released '2025 H1 Advertising Outlook Report' highlights key trends in the advertising industry, focusing on generative AI and automation. The report emphasizes the increasing importance of digital channels like social media, digital display, video, and Connected TV (CTV). Marketers are planning to boost spending on these platforms, with CTV emerging as a top choice for ad investments. Automation is also on the rise, streamlining workflows and creating new opportunities for consumer engagement. Subscribe now for more updates! · 5 h sedan - 102 words

HCL Technologies, India's third-largest IT services firm, reported a 5% increase in revenue for Q3 FY25, reaching ₹29,890 crore. Net profit also rose by 5.5% to ₹4,591 crore. The company secured AI and GenAI-led deals in various sectors and announced plans for new labs and partnerships. HCLTech raised its revenue growth guidance for FY25 and declared an interim dividend. CEO C. Vijayakumar highlighted solid growth and strong client confidence in digital and AI services. Analysts noted healthy deal wins and steady growth, positioning HCLTech for industry-leading growth in FY25. In contrast, Tata Consultancy Services reported a 6% increase in revenue for Q3 FY25, with signs of recovery in key sectors. HCLTech saw a slight rise in attrition, while TCS reported a decline in employees due to geopolitical uncertainties. · 5 h sedan - 123 words

The use of Generative Artificial Intelligence (GenAI) in businesses is revolutionizing workflows, offering opportunities for growth. AI adoption in India is rising, with spending projected to reach $6 billion by 2027. A hybrid multicloud approach can enhance performance and reduce costs for GenAI deployments. Companies are increasingly opting for hybrid cloud solutions to modernize tech stacks and drive innovation. Elevating the skills of current employees is crucial for successful GenAI implementation. Data security and governance are top priorities for businesses using GenAI. Businesses should carefully craft a hybrid multicloud approach to meet their specific needs and thrive in the digital realm. · 5 h sedan - 99 words

Dutch fashion retailer Omoda has partnered with Google Cloud to improve its online shopping platform using AI. They have introduced an AI-powered chatbot called Omoda Stylist and a Discover the Look function that suggests complete outfits to customers. This new technology aims to make online shopping more personalized and engaging. Omoda plans to further enhance the shopping experience by adding avatar-based outfit previews and integrating with users' wardrobes. · 5 h sedan - 68 words