Welcome to Tech Beat, your daily read on the technology stories that matter.
A major research collaboration set out to test whether today's frontier AI systems could do real science, and the results were sobering. The agents could run experiments and process data just fine, but when it came to producing original work worthy of a top AI conference, they came up empty. Competent mechanics, zero genuine insight.
That finding lands with particular weight as the European Union announces an open call for seven AI gigafactories — massive computing installations designed to train the next generation of AI systems. The bloc is clearly betting that raw infrastructure is the missing ingredient, though critics might ask whether more compute alone addresses the creativity gap researchers just identified.
Meanwhile, Google is claiming a genuine win for AI in software security, saying it patched more bugs in Chrome during June alone than it had over the previous two years combined, all thanks to large language models scanning for vulnerabilities. It is a reminder that AI's most consequential near-term impact may not be headline-grabbing breakthroughs, but the quiet, unglamorous work of hardening the systems we already depend on every day.
Keep surfing. Tech Beat out.
