Google DeepMind Audit Finds No Measurable AI Research Speedup
Edited by Steinar Vigdel Kolnes, Editor, Oslo
Google DeepMind's first internal audit of AI-assisted research found no measurable acceleration in AI progress, despite extensive use of AI-generated code.
Google DeepMind’s first internal review of research supported by artificial intelligence found no measurable increase in the pace of AI progress.
The finding contrasts with Google’s reported figure that AI now writes 75% of its code. That level of coding assistance has not translated into a measurable acceleration in AI development, according to the audit.
The review highlights a distinction between generating code and advancing AI research: widespread use of automated coding does not, by itself, show that AI is independently improving its own capabilities.
Source: Decrypt
