[29th June 2025] Interesting Things I Learnt This Week

1. The Illusion of Thinking - A recent paper from Apple's Machine Learning Research team challenges the idea that Large Reasoning Models (LRMs) can truly "think." The researchers argue that current methods for evaluating these models, which often use standard math and coding problems, are flawed due to potential data contamination and a focus on final answers rather than the reasoning process. Using controllable puzzles to test the models, the study found that even the most advanced LRMs experience a complete collapse in accuracy when the problem's complexity passes a certain point. Counterintuitively, as problems became more difficult, the models' "thinking" effort decreased, suggesting they "give up" rather than reason more deeply. The research identifies a significant gap between mimicking thought-like patterns and genuine, scalable reasoning, questioning the true capabilities of current AI.

 My Take: Thinking is a complex process, and each individual approaches analysis and problem-solving differently. It's a fundamental truth that Large Language Models (LLMs), despite their sophisticated outputs, do not truly "think" in the human sense. Instead, they operate by identifying patterns and features within data to generate responses. Many of us in the field didn't need a specific paper to understand this underlying mechanism; unfortunately, some AI influencers have propagated a skewed perception of LLMs. While their motivations may vary, any engineer with a solid grasp of LLMs recognizes the inaccuracies in much of the popular reporting. I commend Apple for releasing their paper, as it provides a much-needed dose of reality amidst the often-unreasonable narratives surrounding AI.

2. Japan to Begin Clinical Trials for Artificial Blood in 2025 -  In a major medical advancement, Japan is set to commence clinical trials for artificial blood by March 2025. This initiative, led by researchers at Nara Medical University, aims to address the critical blood shortages resulting from an aging population and a shrinking donor pool. The artificial blood, created by encasing hemoglobin from expired donor blood in protective shells, offers significant advantages. It eliminates the need for blood type matching, making it universally compatible for emergencies. Furthermore, it boasts a remarkable two-year shelf life at room temperature, a substantial improvement over the less than one-month storage period for donated blood. The trials will initially involve administering small amounts to healthy volunteers to confirm safety before expanding to test efficacy. If successful, this groundbreaking technology could be in practical use by 2030, potentially revolutionizing emergency medicine worldwide.

My Take: The development of artificial blood represents truly remarkable progress in medicine, offering a lifeline in situations where traditional blood transfusions are scarce or problematic. With the potential to bypass issues like blood type compatibility, storage limitations, and the risk of transmitting infections, this innovation could drastically improve outcomes for patients suffering from conditions such as severe blood loss due to trauma, or chronic disorders like Thalassemia and even complications from Dengue. While the current production costs are high, I am optimistic that ongoing advancements and increased efficiency in manufacturing processes will eventually make artificial blood more accessible and affordable, ultimately revolutionizing emergency care and significantly impacting global health.

 

3. World’s First Carbon Removal Plant Powered Directly by Wind Planned - A pioneering large-scale direct air capture (DAC) plant is planned for the Permian Basin in Texas, notable for being the first to be powered primarily by wind energy. Developed by the company Carbon Engineering, the facility is designed to capture up to 1 million metric tons of carbon dioxide from the atmosphere each year. While a significant step for carbon removal technology, the project has drawn scrutiny. The captured CO2 is intended for use in enhanced oil recovery (EOR), a process that injects the gas into oil wells to extract more crude oil. This link to the fossil fuel industry raises questions about the plant's overall climate benefit, as it will simultaneously help produce more of the hydrocarbons contributing to climate change.

My Take: Healing our planet necessitates the synergistic application of diverse technologies, and while direct air capture of carbon dioxide presents a promising solution, its current cost of approximately $500 per ton makes it a significant investment, especially when compared to the roughly one-third cost of bioenergy with carbon capture and storage. However, relying on a single approach would be shortsighted; what truly excites me is the ongoing exploration and development of various innovative methods for carbon removal, as a multi-pronged strategy will ultimately be essential to effectively combat climate change and restore ecological balance.

4. 42 Free and Open Source Projects Receive Funding to Reclaim the Public Nature of the Internet - NLnet has announced a new round of grants in collaboration with the Commons Fund to support open-source projects that contribute to an open, fair, and trustworthy internet. The call for proposals is now open, inviting individuals, organizations, and ad-hoc teams to apply for funding. This initiative aims to foster the development of digital commons and technologies that are freely available for everyone to use, study, modify, and share. Projects funded will cover a wide range of topics, including but not limited to privacy-enhancing technologies, open standards, and digital infrastructure. The collaboration with the Commons Fund allows for a larger pool of resources to be distributed, thereby increasing the potential impact of the funded projects. The deadline for applications and the specific thematic areas of focus are detailed in the announcement, encouraging innovative ideas that align with the mission of creating a more resilient and equitable internet for all.

My Take: Increased funding for Free and Open-Source Software (FOSS) projects is crucial, and I strongly advocate for all companies to financially contribute to the very projects they utilize. Consider the impact if companies paid even a modest $1 per seat per month for the FOSS they integrate into their operations. While companies like Oracle are notorious for high database licensing fees, the existence of robust open-source alternatives such as MariaDB and PostgreSQL has thankfully mitigated this arm-twisting. It's truly disappointing, however, that many companies, including well-funded unicorns, fail to adequately support these vital FOSS organizations. Initiatives like Zerodha's floss.fund are commendable, and we desperately need more such programs. Only through sustained financial support can FOSS development become a viable and respected career path globally, empowering more companies to build exceptional software without the pressures that led projects like Redis and Terraform to seek venture capital and alter their licensing models.