New AI News Shows 80 Percent Efficiency as Scale AI Develops

The year 2025 marked a significant acceleration in artificial intelligence, with Agentic AI emerging as a key development, capable of understanding complex goals and autonomously completing tasks. This advancement improved areas like customer service and software development. Concurrently, hardware innovation saw the rise of specialized AI chips such as TPUs and NPUs, alongside growth in neuromorphic and edge AI computing, laying groundwork for future discoveries. In response to this rapid evolution, the U.S. Department of Energy launched its Genesis Mission on November 24, 2025. This ambitious initiative aims to double American science and engineering output within a decade by establishing the American Science and Security Platform, which connects 17 national laboratories and leverages the expertise of 40,000 DOE staff. Experts emphasize that robust, open, high-performance data infrastructure is crucial for the success of such national AI goals, advocating for a shift away from limiting proprietary storage systems. AI tools are rapidly transforming various sectors. Kosmos, an AI agent developed by Edison Scientific, demonstrates remarkable efficiency, completing six months of doctoral-level research in just 12 hours and providing deep, interesting insights about 80 percent of the time. In the creative sphere, Higgsfield AI, founded in early 2024, empowers creators to generate cinematic videos from text or images, offering over 50 AI-powered camera movements without physical equipment. The economic impact of AI in 2025 was substantial, leading to over 50 individuals becoming AI billionaires. Lucy Guo, co-founder of Scale AI, notably became the youngest self-made woman billionaire during this period. Companies like Sierra, focusing on AI customer service, and ElevenLabs, known for voice generation, also experienced significant growth. Furthermore, Z.ai (formerly Zhipu AI) released GLM-4.5 in July 2025, an "agent-native" open-source large language model from China, which excels in complex tasks like multi-step planning and data visualization, ranking among the top global models. AI's integration extends even to leisure activities, with systems optimizing strategies for sports like curling and games such as poker. However, the rapid adoption of tools like ChatGPT also highlights critical concerns for organizations, particularly faith-based groups, regarding confidentiality, intellectual property, and plagiarism. Clear AI usage policies are recommended to manage these issues and ensure human judgment remains central, especially when handling sensitive data. In scientific publishing, the World Journal of Stem Cells proposes a new framework for evaluating research, moving beyond traditional citation counts. This updated approach focuses on transparency, the role of artificial intelligence, and an Economic Impact Factor, aiming to better value truly impactful research. The journal already practices transparent peer review, publishing reports alongside manuscripts to foster greater accountability in science.

Key Takeaways

• The U.S. Department of Energy launched the Genesis Mission on November 24, 2025, aiming to double American science and engineering output in ten years through the American Science and Security Platform.
• Strong, open, high-performance object storage is deemed crucial for national AI success, as current proprietary systems limit data movement and increase risk.
• 2025 saw the rise of Agentic AI, capable of understanding complex goals and completing tasks autonomously, alongside significant advancements in specialized AI hardware like TPUs and NPUs.
• Kosmos, an AI agent from Edison Scientific, can complete six months of doctoral-level research in 12 hours, providing deep insights about 80 percent of the time.
• Higgsfield AI, founded in early 2024, enables creators to make cinematic videos from text or images using over 50 AI-powered camera movements.
• Over 50 people became AI billionaires in 2025, with Lucy Guo, co-founder of Scale AI, briefly becoming the youngest self-made woman billionaire.
• Z.ai (formerly Zhipu AI) released GLM-4.5 in July 2025, an "agent-native" open-source large language model from China, excelling in reasoning, coding, and agent functions.
• AI is increasingly used to optimize leisure activities, such as finding strategies for curling and improving performance in games like poker.
• The rapid adoption of AI tools like ChatGPT raises concerns for organizations regarding confidentiality, intellectual property, and plagiarism, necessitating clear usage policies.
• The World Journal of Stem Cells proposes a new scientific publishing framework focusing on transparency, artificial intelligence, and an Economic Impact Factor.

DOE launches Genesis Mission to boost US science with AI

The U.S. Department of Energy (DOE) started the Genesis Mission on November 24, 2025, aiming to double American science and engineering output in ten years. This mission will create the American Science and Security Platform, connecting 17 national laboratories, supercomputers, and AI systems. U.S. Secretary of Energy Chris Wright announced this major effort, which involves 40,000 DOE staff and private companies. The platform will use high-performance computing and AI to advance energy, discovery science, and national security. Dr. Dar\u00edo Gil directs the mission, which also announced 24 new research partnerships on December 18.

Strong data infrastructure is key for AI success

The article argues that strong data infrastructure is crucial for artificial intelligence success, despite focus on AI models and chips. It states that AI often fails because the underlying data storage cannot handle the speed and scale needed. The U.S. government's AI Action Plan and Genesis Mission aim for AI leadership, but current proprietary storage systems limit data movement and increase risk. The author suggests the federal government should adopt open, high-performance object storage, similar to what the private sector has done. This change is vital to mobilize data securely and rapidly for national AI goals.

2025 saw Agentic AI and new hardware redefine technology

In 2025, technology advanced significantly with the rise of Agentic AI, which can understand complex goals and complete tasks on its own. This AI, powered by large language models, improved customer service and software development. Hardware also saw major changes, with new specialized AI chips like TPUs and NPUs, plus growth in neuromorphic and edge AI computing. Other breakthroughs included personalized medicine, sustainable technology, and more realistic immersive experiences with AR/VR. Quantum computing also made important progress, setting the stage for future discoveries.

AI is changing how we play and enjoy leisure

Artificial intelligence is increasingly being used to optimize fun activities, from sports to games. For example, University of Alberta academics developed an AI to find the best strategies for curling, which top teams now use. AI has also become very good at games like poker, with a Carnegie Mellon University bot beating pros in 2019. Even horse betting and fantasy sports use AI to spot patterns for better outcomes. While AI improves performance, the article questions if we lose the joy of discovery and personal effort when machines optimize our leisure.

New plan for science publishing uses AI and economic impact

The World Journal of Stem Cells proposes a new way to evaluate scientific publishing after 16 years of evolution. This new framework moves beyond traditional citation counts to focus on transparency, artificial intelligence, and an Economic Impact Factor. The journal has published over 1200 articles, with top works like Kyurkchiev et al. and Casteilla et al. receiving many citations. It also supports important clinical research, such as Tsang et al.'s stem cell stroke trial. The journal already practices transparent peer review by publishing reports with manuscripts, aiming to value what truly matters in science.

AI offers benefits but needs clear rules

Artificial intelligence, with tools like ChatGPT seeing rapid adoption, brings both opportunities and challenges for organizations, especially faith-based groups. Key concerns include confidentiality, as sensitive data entered into AI might not remain private. There are also issues with intellectual property, as AI-generated content may lack clear ownership or accidentally infringe on others' rights. Organizations must also watch for plagiarism and ensure human judgment is not replaced by AI, upholding fiduciary duties. Faith-based groups should also consider the moral impact of AI. To manage these issues, clear AI usage policies are recommended.

Higgsfield AI creates cinematic videos from text and images

Higgsfield AI is a new platform that helps creators make cinematic videos from text, images, or sketches. It offers over 50 AI-powered camera movements, like dolly zooms and crane shots, without needing physical equipment. The platform combines video generation with editing tools like Canvas and SOUL Inpaint for precise visual adjustments. Founded in early 2024 by Alex Mashrabov and Yerzat Dulat, Higgsfield AI also includes a Turbo model for faster work and can upscale videos to 8K. It is useful for ads, product demos, and social media content, but users should check its terms for commercial use.

AI tool Kosmos speeds up scientific research

Many leaders claim AI will revolutionize science, but questions remain about its actual impact. Sam Rodriques, a scientist and technologist, created Kosmos, an AI agent from Edison Scientific. Kosmos can complete six months of doctoral-level research in just 12 hours, providing deep and interesting insights about 80 percent of the time. Rodriques explains that while AI tools like Kosmos can greatly speed up data analysis, some claims about AI quickly curing diseases are unrealistic. Human involvement remains crucial, and certain bottlenecks still prevent a fully AI-accelerated scientific future.

Over 50 people became AI billionaires in 2025

In 2025, more than 50 people became billionaires due to the growth of artificial intelligence. The focus shifted from grand AI predictions to creating immense wealth. Many of these new billionaires founded companies that offer enterprise software, data labeling services, or AI agents to replace human workers. Examples include Bret Taylor and Clay Bavor, who founded Sierra for AI customer service, and Mati Staniszewski and Piotr Dabkowski from ElevenLabs, known for voice generation. Lucy Guo, co-founder of Scale AI, also briefly became the youngest self-made woman billionaire.

Zai launches GLM-4.5 a powerful new AI model

Z.ai, formerly Zhipu AI, released GLM-4.5 in July 2025, a new open-source large language model from China. This model is the first "agent-native" open-source AI, combining reasoning, coding, and agent functions into one system. It can handle complex tasks like multi-step planning and data visualization without extra fine-tuning. GLM-4.5 performs very well on benchmarks, ranking among the top globally and first for open-source models. It comes in two versions, GLM-4.5 and GLM-4.5-Air, offering powerful AI at a low cost. Users can access it through a free web chat, API, or download its open weights for commercial use.

Sources

• DOE’s ‘Genesis Mission’ Enlists AI to Double U.S. Research Productivity in a Decade
• AI leadership begins — and fails — with the foundation no one talks about
• Tech in 2025: Rise of Agentic AI, hardware shifts and breakthroughs that redefined the year - The Times of India
• What Happens When We Insist on Optimizing Fun?
• Evolving target: A 16-year progressive framework for shifting the rubric of scientific publishing toward transparency, artificial intelligence, and the Economic Impact Factor for impact that matters
• Artificial Intelligence — A Blessing With Boundaries (Video)
• Higgsfield AIレビュー：シネマティックなテキスト→動画生成の完全ガイド
• Where Is All the A.I.-Driven Scientific Progress?
• More Than 50 People Became AI Billionaires in 2025
• Z.AI：GLM-4.5無料オンラインチャット