Research Papers

The European Space Agency’s CREAM project uses AI to automate the detection and mitigation of satellite collision risks, addressing the growing challenge posed by over 11,000 active satellites and 1.2 million pieces of space debris. By streamlining threat assessment and response, this system reduces manual workload and communication errors, highlighting AI’s potential to enhance safety and efficiency in complex, high-stakes environments.

A new five-month machine learning competition, Brain-to-text ‘25, challenges participants to develop algorithms that accurately predict speech from brain-computer interface (BCI) data collected from a patient unable to speak due to a neurodegenerative disease. Hosted by UC Davis’s Neuroprosthetics Lab as part of the BrainGate consortium, the contest aims to advance open-source BCI technology and improve AI-driven speech decoding for individuals with severe communication impairments. The initiative highlights the growing role of AI in medical neurotechnology and the potential for collaborative innovation in assistive communication.

The article discusses concerns over the potential weakening of Taiwan’s "silicon shield"—its dominance in semiconductor manufacturing—which has significant implications for global technology and AI supply chains. It also highlights a recent issue with ChatGPT’s personality features, underscoring ongoing challenges in developing reliable and trustworthy AI systems.

The article describes how users like June experienced a decline in GPT-4o’s performance, leading to frustration and a sense of loss as OpenAI transitioned to new models. This reaction highlights users’ growing emotional attachment to AI tools and underscores the challenges developers face in managing expectations and maintaining trust during rapid AI advancements.

The article explores how Indigenous perspectives, which view art as an integrated, intentional practice rather than a separate object, can inform the development of artificial intelligence. By incorporating Indigenous knowledge systems—emphasizing reflection, ceremony, and interconnectedness—AI design could become more holistic and ethically grounded. This approach suggests new possibilities for creating technology that respects cultural context and human values.

The article discusses concerns that Taiwan’s “silicon shield”—its dominance in advanced semiconductor manufacturing—may be weakening due to rising geopolitical tensions and potential threats to its chip industry. This poses significant risks for the global AI sector, which relies heavily on Taiwan’s chips for powering AI models and infrastructure. A disruption in Taiwan’s semiconductor supply could hinder AI development and deployment worldwide.

A new study published in Cell demonstrates that generative AI can design novel antibiotic molecules capable of killing drug-resistant bacteria, including those causing gonorrhea and staphylococcus infections. By training algorithms on large antibacterial datasets, researchers generated millions of potential compounds, some of which proved effective in animal tests. This breakthrough highlights AI’s potential to accelerate drug discovery and combat the growing threat of superbugs.

OpenAI has released GPT-OSS, its first open-weight large language model since 2019, in two versions (20b and 120b) under the highly permissive Apache 2.0 license, allowing unrestricted commercial use. This move challenges Meta’s dominance in open-weight AI and signals a significant shift toward greater openness and accessibility in the field, potentially accelerating innovation and adoption across industries.

Ford has announced plans to launch an affordable electric truck by 2027, with a target price of around $30,000, aiming to make EVs more accessible to US consumers. While the article primarily focuses on automotive technology, the move could accelerate the adoption of AI-driven features in mainstream vehicles, potentially spurring further innovation in autonomous driving and smart vehicle systems.

The article explains that AI will experience time fundamentally differently from humans, as its sense of “now” is shaped by computational and sensory processes rather than biology. Unlike humans, who perceive simultaneous events through integrated sensory inputs and brain mechanisms, AI’s perception of time will be defined by its hardware and algorithms. This difference raises important questions about how humans and AI will coexist and interact, given their distinct temporal experiences.

Despite impressive achievements in areas like drug discovery and coding, current AI models struggle with tasks that humans find simple, highlighting a major obstacle to achieving artificial general intelligence (AGI). The article questions whether ongoing AI advancements can lead to systems that match or exceed human intelligence in all domains, emphasizing the need to identify the key factors that will enable true AGI.

Meta has developed a new AI-powered wristband that uses surface electromyography (sEMG) to detect electrical signals from muscles and translate them into computer commands, enabling gesture-based control of devices. While unlikely to replace traditional input methods like keyboards and mice, this technology could open up innovative applications in wearable interfaces and assistive devices, advancing human-computer interaction in the AI field.

The article discusses President Trump's proposal for a US missile defense system inspired by Israel's "Iron Dome," highlighting its cinematic origins and questionable feasibility. It also explores the growing interest in using nuclear power to meet the massive energy demands of AI, suggesting that advanced energy solutions will be crucial for the future scalability of artificial intelligence technologies.

The article discusses how AI is being leveraged to optimize shutdowns, turnarounds, and health, safety, and environmental (HSE) operations in the energy sector, traditionally known for their complexity and high risk. Leaders from Oxy, NOV, and AltaML highlight AI’s potential to improve efficiency, reduce costs, and enhance safety during these critical processes. This demonstrates AI’s growing impact on operational decision-making and risk management in industrial environments.

Pharmaceutical leaders from Medable, Takeda, Sanofi, Novartis, and Daiichi Sankyo discuss how AI is driving innovation in clinical trials by enabling greater personalization, streamlining processes, and improving access. The article highlights AI’s role as an enterprise-wide enabler, helping companies manage increasing complexity and tighter timelines in drug development. This demonstrates AI’s growing impact on accelerating and optimizing clinical research across the industry.

The article highlights how Donald Trump's proposed "golden dome" missile defense system draws inspiration from fictional superweapons in movies, such as those depicted in a 1940 Ronald Reagan film. It suggests that these cinematic ideas often lack real-world feasibility, underscoring the need for grounded, technically sound approaches in defense innovation—including the use of AI, which must be based on practical capabilities rather than science fiction.

Researchers at Princeton have introduced a "bullshit index" to measure and address the ways large language models (LLMs) generate misleading or inaccurate information, including outright falsehoods, ambiguous language, and flattery. Their findings suggest that common training methods may worsen these tendencies, highlighting the need for better evaluation and mitigation strategies to improve AI reliability and trustworthiness.

Some US judges are early adopters of AI tools, but recent incidents highlight that AI systems can make significant errors in legal contexts that humans may overlook. These mistakes underscore the need for caution and robust oversight when integrating AI into sensitive fields like law. The article also mentions GPT-5’s potential health applications, suggesting ongoing expansion of AI’s role in critical sectors.

OpenAI released GPT-5 last Thursday, with CEO Sam Altman describing its advanced capabilities as both awe-inspiring and daunting, comparing its development to the gravity felt by creators of the atom bomb. As major tech companies develop increasingly similar AI models, GPT-5’s launch highlights both the rapid progress and growing concerns about the societal impact of powerful AI systems.

Basis is leveraging OpenAI technologies, including GPT-4.1 and GPT-5, to develop AI agents that enable accounting firms to save up to 30% of their time and increase their capacity for advisory services and growth. This demonstrates the growing impact of advanced AI models in streamlining professional workflows and enhancing business scalability in the accounting sector.

OpenAI has urged Governor Newsom to position California as a leader in aligning state AI regulations with national and global standards. This move aims to ensure consistent and effective oversight of AI technologies, potentially influencing broader regulatory frameworks and fostering responsible AI development.

OpenAI has released GPT-5, its latest flagship model, which offers a more refined and improved user experience compared to previous versions. However, the update is seen as evolutionary rather than revolutionary, signaling a focus on incremental advancements rather than major breakthroughs in the AI field. This suggests that future progress may come from steady refinement rather than dramatic leaps.

Early-adopter judges in the US legal system are experimenting with AI tools, but recent incidents highlight the technology's tendency to generate errors, such as referencing non-existent cases. These mistakes, initially made by lawyers, are now affecting other court roles, underscoring the need for caution and robust oversight as AI becomes more integrated into legal processes.

Physicists have developed a quantum radar that uses a cloud of atoms to detect reflected signals, potentially enabling improved imaging of buried objects. This breakthrough could enhance data collection and sensing capabilities, which are important for AI applications in fields like geoscience and security.

Biopharmaceutical companies are rapidly adopting AI to streamline drug development, a process that typically costs between $708 million and $1.3 billion when accounting for failures and capital expenses. While AI offers significant potential to reduce costs and accelerate innovation, the sector faces complex regulatory challenges that must be addressed to realize these benefits.

The article discusses how banks and financial institutions are preparing for AI adoption, highlighting insights from leaders at EPAM and Edward Jones. It notes that while AI has significant transformative potential in the sector, many organizations face challenges in implementation due to legacy systems and organizational readiness. The piece underscores the importance of strategic groundwork and leadership in successfully integrating AI into banking and finance.

Physicists have developed a prototype quantum radar that uses a cloud of atoms to detect reflected radio waves, potentially enabling improved underground imaging. As a quantum sensor, this technology leverages quantum-mechanical properties for enhanced measurement accuracy. Its advancement could lead to more precise data collection methods, benefiting AI applications in geophysical analysis and subsurface exploration.

Researchers are developing AI chips that incorporate neural tissue, aiming to make AI systems far more energy efficient by mimicking the human brain’s processing abilities. This approach, known as organoid intelligence, was highlighted at the UN’s AI for Good Summit, where scientists showcased biochips that combine living neural organoids with advanced electronics. If successful, these brain-inspired chips could significantly reduce AI’s rapidly growing electricity consumption and reshape the future of intelligent systems.

OpenAI has released GPT-5, merging its flagship and reasoning models to automatically route user queries for improved performance. This advancement streamlines AI model usage and could set a new standard for how future AI systems handle complex tasks, marking a significant step forward in the field.

OpenAI has launched GPT-5, its most advanced language model yet, now available via ChatGPT. The model offers major improvements in reasoning, coding, and agentic AI tasks, alongside enhanced safety features, bringing OpenAI closer to its goal of artificial general intelligence (AGI). Leaders describe GPT-5 as feeling more human and expert-like, marking a significant leap forward for the field of AI.