DNA computing

Researchers at Maynooth University have built a scaffolded DNA computer in which short DNA strands compete to attach along a long scaffold strand. Mismatched tiles are energetically unfavorable, so they detach and are replaced until the arrangement settles into the most stable configuration, a process that lets the molecular system compute through self-assembly.

Researchers have introduced a new computational tool that detects traces of human DNA hidden within microbiome sequencing data and uses them to predict a sample's biological sex. The approach taps genetic material that is typically discarded as contamination in microbiome studies, potentially aiding forensic and biomedical research while raising privacy questions about what sensitive information can be extracted from such datasets.

Researchers have reportedly created a functioning computer inside a single water droplet, using billions of DNA molecules to carry out computations. The work points to a new approach in molecular computing, where biological molecules instead of silicon chips process information. Details of the research, the team behind it, and what calculations the device actually performed remain sparse.

A new analysis by Epoch AI finds the cost of achieving a given level of AI model performance has dropped roughly 13-fold per year over the past five years. That decline is faster than historical cost curves for computing hardware, lithium batteries, DNA sequencing and electricity. The authors caution, however, that these steep efficiency gains may not continue indefinitely, prompting discussion about how long such rapid deflation in AI costs can last.