An AI startup called Subquadratic emerged from stealth last month with a huge claim: it had solved a mathematical bottleneck that had held back large language models for almost a decade. The purported breakthrough comes from slashing the number of computations transformers need to carry out to generate answers. The result is a faster and cheaper LLM that uses far less energy than any other model on the market. Many experts remain skeptical, but Subquadratic has started to share the receipts, suggesting their approach might be worth paying attention to.
The system works by optimizing the attention mechanism in transformers, the key component that allows models to weigh the importance of different input parts. Traditionally, this process requires a quadratic computational cost with respect to sequence length, limiting scalability. Subquadratic claims to achieve linear cost, paving the way for much larger and more efficient LLMs. However, researchers caution that the theory needs large-scale validation and that efficiency gains may not translate to real-world improvement in all scenarios.
Sponsored Protocol
BCI Trials Accelerate
This week also marked a milestone in brain-computer interface (BCI) research. Casey Harrell, a man with ALS, became the first "power user" of a brain implant. The device has enabled him to maintain an income, reconnect with friends and family, and read to his daughter. He described it as "nothing short of revolutionary." Over the past couple of years, the number of BCI trial volunteers has soared. This year, China became the first country to approve a BCI for medical use. Advances in technology are allowing engineers to provide more features than ever, and BCI research is properly taking off from the lab towards the market.
The implications are vast. BCIs could restore mobility and communication for people with severe neurological disorders, but they also raise ethical questions about neural data privacy and implant security. As companies like Neuralink and Synchron accelerate trials, regulation must keep pace.
Sponsored Protocol
Meanwhile, other tech stories unfold. Geopolitical tensions over AI, as seen in the SK Telecom and Anthropic case, highlight how innovation is intertwined with politics. Similarly, Adobe's expansion of Firefly shows generative AI permeating creative software.
Additionally, a recent study suggests AI may already be deskilling professionals: over-reliance appears to weaken doctors' and engineers' abilities. This serves as a cautionary note as AI integrates deeper into decision-making. The scientific community watches both promises and risks closely.
