Click on any of the titles to read the full piece!
DeepSeek is kinda out there, man. It’s only been a few weeks since the reasoning model was released, and the new AI has already made big waves in the community. The reason? DeepSeek is not necessarily better than ChatGPT or equivalent models from Meta. Rather, its phenomenon stems from its improvements in efficiency over competitors, allowing it to be trained for cheaper.
New (anecdotal) data from the MIT Technology Review suggests otherwise, however, as the energy saved while training DeepSeek may be neutralized by its requirement for more energy when answering prompts. With more and more companies pressured to adapt DeepSeek’s architecture, what will happen to the AI industry remains to be seen.
Stanford researchers have developed Evo, a new AI model designed to understand and work with genetic information at multiple biological scales. Trained using prokaryotic genomes, this 7-billion-parameter model can analyze and generate DNA sequences ranging from individual molecules to entire (still prokaryotic) genomes. This represents a significant advancement in understanding and manipulating biological systems.
This MIT study found that combining light and sound pulses at a specific frequency (40 Hz) can activate the brain's natural cleaning system in mice with Alzheimer 's-like conditions. This non-invasive treatment helps flush out harmful proteins by improving fluid flow through the brain, enhancing the function of specialized cleaning channels, and dilating waste-draining vessels. The findings suggest a promising new approach for treating Alzheimer's disease without drugs or surgery.
This University of Hong Kong study addresses a significant challenge faced by academics, students, and sick people alike: the limited accessibility, extreme energy consumption, and crippling costs of MRI technology worldwide. The researchers developed a revolutionary new type of MRI scanner that operates at ultra-low-field strength (0.05 Tesla), making it much simpler and more affordable than conventional MRI machines. It requires nothing more than a standard wall power outlet to work.
This UCLA study examined how the brain stores temporary memories by observing thousands of brain cells in mice during an odor memory task. Using high-resolution imaging, they saw that neural activity (which would eventually become a memory) was initially unstable and changed daily. However, these patterns became more fixed and reliable with continued practice, especially in deeper brain layers. The findings show how working memories "crystallize" with expertise.
In this University of Utah study, researchers developed a self-powered smart contact lens that generates electricity using two sources: tiny solar cells and the electrolytes in your tears, collected during blinking. The system works both indoors and outdoors, requiring no external power sources or accessories. This breakthrough can continuously power features like vision correction and health monitoring, drug delivery, and creating an ultra-immersive augmented reality experience.
Want to submit a piece? Or trying to write a piece and struggling? Check out the guides here!
Thank you for reading. Reminder: Byte Sized is open to everyone! Feel free to submit your piece. Please read the guides first though.
All submissions to berkan@usc.edu with the header “Byte Sized Submission” in Word Doc format please. Thank you!